ClusterCockpit/cc-backend
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-backend synced 2025-10-30 15:45:05 +01:00
Code Issues Packages Releases Activity

Merge branch 'dev' into add_editmeta_byrequest

Browse Source
This commit is contained in:
Christoph Kluge
2025-07-03 14:02:54 +02:00
parent 8dfaa78a83 60ec7e54f5
commit d35d00a18e
242 changed files with 15413 additions and 15625 deletions
Download Patch File Download Diff File Expand all files Collapse all files

441
api/schema.graphqls
View File

@@ -4,61 +4,78 @@ scalar Any
scalar NullableFloat
scalar MetricScope
scalar JobState
scalar NodeState
scalar MonitoringState
type Node {
id: ID!
hostname: String!
cluster: String!
subCluster: String!
nodeState: NodeState!
HealthState: MonitoringState!
metaData: Any
}
type NodeStats {
state: String!
count: Int!
}
type Job {
id: ID!
jobId: Int!
user: String!
project: String!
cluster: String!
subCluster: String!
startTime: Time!
duration: Int!
walltime: Int!
numNodes: Int!
numHWThreads: Int!
numAcc: Int!
energy: Float!
SMT: Int!
exclusive: Int!
partition: String!
arrayJobId: Int!
id: ID!
jobId: Int!
user: String!
project: String!
cluster: String!
subCluster: String!
startTime: Time!
duration: Int!
walltime: Int!
numNodes: Int!
numHWThreads: Int!
numAcc: Int!
energy: Float!
SMT: Int!
exclusive: Int!
partition: String!
arrayJobId: Int!
monitoringStatus: Int!
state: JobState!
tags: [Tag!]!
resources: [Resource!]!
concurrentJobs: JobLinkResultList
footprint: [FootprintValue]
energyFootprint: [EnergyFootprintValue]
metaData: Any
userData: User
state: JobState!
tags: [Tag!]!
resources: [Resource!]!
concurrentJobs: JobLinkResultList
footprint: [FootprintValue]
energyFootprint: [EnergyFootprintValue]
metaData: Any
userData: User
}
type JobLink {
id: ID!
jobId: Int!
id: ID!
jobId: Int!
}
type Cluster {
name: String!
partitions: [String!]! # Slurm partitions
subClusters: [SubCluster!]! # Hardware partitions/subclusters
name: String!
partitions: [String!]! # Slurm partitions
subClusters: [SubCluster!]! # Hardware partitions/subclusters
}
type SubCluster {
name: String!
nodes: String!
numberOfNodes: Int!
processorType: String!
socketsPerNode: Int!
coresPerSocket: Int!
threadsPerCore: Int!
flopRateScalar: MetricValue!
flopRateSimd: MetricValue!
name: String!
nodes: String!
numberOfNodes: Int!
processorType: String!
socketsPerNode: Int!
coresPerSocket: Int!
threadsPerCore: Int!
flopRateScalar: MetricValue!
flopRateSimd: MetricValue!
memoryBandwidth: MetricValue!
topology: Topology!
metricConfig: [MetricConfig!]!
footprint: [String!]!
topology: Topology!
metricConfig: [MetricConfig!]!
footprint: [String!]!
}
type FootprintValue {
@@ -80,99 +97,112 @@ type MetricValue {
}
type Topology {
node: [Int!]
socket: [[Int!]!]
node: [Int!]
socket: [[Int!]!]
memoryDomain: [[Int!]!]
die: [[Int!]!]
core: [[Int!]!]
die: [[Int!]!]
core: [[Int!]!]
accelerators: [Accelerator!]
}
type Accelerator {
id: String!
type: String!
id: String!
type: String!
model: String!
}
type SubClusterConfig {
name: String!
peak: Float
normal: Float
name: String!
peak: Float
normal: Float
caution: Float
alert: Float
remove: Boolean
alert: Float
remove: Boolean
}
type MetricConfig {
name: String!
unit: Unit!
scope: MetricScope!
name: String!
unit: Unit!
scope: MetricScope!
aggregation: String!
timestep: Int!
peak: Float!
normal: Float
timestep: Int!
peak: Float!
normal: Float
caution: Float!
alert: Float!
alert: Float!
lowerIsBetter: Boolean
subClusters: [SubClusterConfig!]!
}
type Tag {
id: ID!
id: ID!
type: String!
name: String!
scope: String!
}
type Resource {
hostname: String!
hwthreads: [Int!]
accelerators: [String!]
hostname: String!
hwthreads: [Int!]
accelerators: [String!]
configuration: String
}
type JobMetricWithName {
name: String!
scope: MetricScope!
name: String!
scope: MetricScope!
metric: JobMetric!
}
type JobMetric {
unit: Unit
timestep: Int!
series: [Series!]
unit: Unit
timestep: Int!
series: [Series!]
statisticsSeries: StatsSeries
}
type Series {
hostname: String!
id: String
hostname: String!
id: String
statistics: MetricStatistics
data: [NullableFloat!]!
data: [NullableFloat!]!
}
type StatsSeries {
mean: [NullableFloat!]!
mean: [NullableFloat!]!
median: [NullableFloat!]!
min: [NullableFloat!]!
max: [NullableFloat!]!
min: [NullableFloat!]!
max: [NullableFloat!]!
}
type JobStatsWithScope {
name: String!
scope: MetricScope!
stats: [ScopedStats!]!
type NamedStatsWithScope {
name: String!
scope: MetricScope!
stats: [ScopedStats!]!
}
type ScopedStats {
hostname: String!
id: String
data: MetricStatistics!
hostname: String!
id: String
data: MetricStatistics!
}
type JobStats {
name: String!
stats: MetricStatistics!
id: Int!
jobId: String!
startTime: Int!
duration: Int!
cluster: String!
subCluster: String!
numNodes: Int!
numHWThreads: Int
numAccelerators: Int
stats: [NamedStats!]!
}
type NamedStats {
name: String!
data: MetricStatistics!
}
type Unit {
@@ -188,12 +218,12 @@ type MetricStatistics {
type MetricFootprints {
metric: String!
data: [NullableFloat!]!
data: [NullableFloat!]!
}
type Footprints {
timeWeights: TimeWeights!
metrics: [MetricFootprints!]!
metrics: [MetricFootprints!]!
}
type TimeWeights {
@@ -202,20 +232,33 @@ type TimeWeights {
coreHours: [NullableFloat!]!
}
enum Aggregate { USER, PROJECT, CLUSTER }
enum SortByAggregate { TOTALWALLTIME, TOTALJOBS, TOTALNODES, TOTALNODEHOURS, TOTALCORES, TOTALCOREHOURS, TOTALACCS, TOTALACCHOURS }
enum Aggregate {
USER
PROJECT
CLUSTER
}
enum SortByAggregate {
TOTALWALLTIME
TOTALJOBS
TOTALNODES
TOTALNODEHOURS
TOTALCORES
TOTALCOREHOURS
TOTALACCS
TOTALACCHOURS
}
type NodeMetrics {
host: String!
host: String!
subCluster: String!
metrics: [JobMetricWithName!]!
metrics: [JobMetricWithName!]!
}
type NodesResultList {
items: [NodeMetrics!]!
items: [NodeMetrics!]!
offset: Int
limit: Int
count: Int
limit: Int
count: Int
totalNodes: Int
hasNextPage: Boolean
}
@@ -234,14 +277,14 @@ type GlobalMetricListItem {
}
type Count {
name: String!
name: String!
count: Int!
}
type User {
username: String!
name: String!
email: String!
name: String!
email: String!
}
input MetricStatItem {
@@ -250,26 +293,81 @@ input MetricStatItem {
}
type Query {
clusters: [Cluster!]! # List of all clusters
tags: [Tag!]! # List of all tags
globalMetrics: [GlobalMetricListItem!]!
clusters: [Cluster!]! # List of all clusters
tags: [Tag!]! # List of all tags
globalMetrics: [GlobalMetricListItem!]!
user(username: String!): User
allocatedNodes(cluster: String!): [Count!]!
node(id: ID!): Node
nodes(filter: [NodeFilter!], order: OrderByInput): NodeStateResultList!
nodeStats(filter: [NodeFilter!]): [NodeStats!]!
job(id: ID!): Job
jobMetrics(id: ID!, metrics: [String!], scopes: [MetricScope!], resolution: Int): [JobMetricWithName!]!
jobStats(id: ID!, metrics: [String!]): [JobStats!]!
scopedJobStats(id: ID!, metrics: [String!], scopes: [MetricScope!]): [JobStatsWithScope!]!
jobMetrics(
id: ID!
metrics: [String!]
scopes: [MetricScope!]
resolution: Int
): [JobMetricWithName!]!
jobStats(id: ID!, metrics: [String!]): [NamedStats!]!
scopedJobStats(
id: ID!
metrics: [String!]
scopes: [MetricScope!]
): [NamedStatsWithScope!]!
jobs(
filter: [JobFilter!]
page: PageRequest
order: OrderByInput
): JobResultList!
jobsStatistics(
filter: [JobFilter!]
metrics: [String!]
page: PageRequest
sortBy: SortByAggregate
groupBy: Aggregate
numDurationBins: String
numMetricBins: Int
): [JobsStatistics!]!
jobsMetricStats(filter: [JobFilter!], metrics: [String!]): [JobStats!]!
jobsFootprints(filter: [JobFilter!], metrics: [String!]!): Footprints
jobs(filter: [JobFilter!], page: PageRequest, order: OrderByInput): JobResultList!
jobsStatistics(filter: [JobFilter!], metrics: [String!], page: PageRequest, sortBy: SortByAggregate, groupBy: Aggregate, numDurationBins: String, numMetricBins: Int): [JobsStatistics!]!
rooflineHeatmap(
filter: [JobFilter!]!
rows: Int!
cols: Int!
minX: Float!
minY: Float!
maxX: Float!
maxY: Float!
): [[Float!]!]!
rooflineHeatmap(filter: [JobFilter!]!, rows: Int!, cols: Int!, minX: Float!, minY: Float!, maxX: Float!, maxY: Float!): [[Float!]!]!
nodeMetrics(cluster: String!, nodes: [String!], scopes: [MetricScope!], metrics: [String!], from: Time!, to: Time!): [NodeMetrics!]!
nodeMetricsList(cluster: String!, subCluster: String!, nodeFilter: String!, scopes: [MetricScope!], metrics: [String!], from: Time!, to: Time!, page: PageRequest, resolution: Int): NodesResultList!
nodeMetrics(
cluster: String!
nodes: [String!]
scopes: [MetricScope!]
metrics: [String!]
from: Time!
to: Time!
): [NodeMetrics!]!
nodeMetricsList(
cluster: String!
subCluster: String!
nodeFilter: String!
scopes: [MetricScope!]
metrics: [String!]
from: Time!
to: Time!
page: PageRequest
resolution: Int
): NodesResultList!
}
type Mutation {
@@ -282,37 +380,52 @@ type Mutation {
updateConfiguration(name: String!, value: String!): String
}
type IntRangeOutput { from: Int!, to: Int! }
type TimeRangeOutput { range: String, from: Time!, to: Time! }
type IntRangeOutput {
from: Int!
to: Int!
}
type TimeRangeOutput {
range: String
from: Time!
to: Time!
}
input NodeFilter {
hostname: StringInput
cluster: StringInput
nodeState: NodeState
healthState: MonitoringState
}
input JobFilter {
tags: [ID!]
jobId: StringInput
arrayJobId: Int
user: StringInput
project: StringInput
jobName: StringInput
cluster: StringInput
partition: StringInput
duration: IntRange
energy: FloatRange
tags: [ID!]
dbId: [ID!]
jobId: StringInput
arrayJobId: Int
user: StringInput
project: StringInput
jobName: StringInput
cluster: StringInput
partition: StringInput
duration: IntRange
energy: FloatRange
minRunningFor: Int
numNodes: IntRange
numNodes: IntRange
numAccelerators: IntRange
numHWThreads: IntRange
numHWThreads: IntRange
startTime: TimeRange
state: [JobState!]
startTime: TimeRange
state: [JobState!]
metricStats: [MetricStatItem!]
exclusive: Int
node: StringInput
exclusive: Int
node: StringInput
}
input OrderByInput {
field: String!
type: String!,
type: String!
order: SortDirectionEnum! = ASC
}
@@ -322,34 +435,46 @@ enum SortDirectionEnum {
}
input StringInput {
eq: String
neq: String
contains: String
eq: String
neq: String
contains: String
startsWith: String
endsWith: String
in: [String!]
endsWith: String
in: [String!]
}
input IntRange { from: Int!, to: Int! }
input TimeRange { range: String, from: Time, to: Time }
input IntRange {
from: Int!
to: Int!
}
input TimeRange {
range: String
from: Time
to: Time
}
input FloatRange {
from: Float!
to: Float!
}
type NodeStateResultList {
items: [Node!]!
count: Int
}
type JobResultList {
items: [Job!]!
items: [Job!]!
offset: Int
limit: Int
count: Int
limit: Int
count: Int
hasNextPage: Boolean
}
type JobLinkResultList {
listQuery: String
items: [JobLink!]!
count: Int
items: [JobLink!]!
count: Int
}
type HistoPoint {
@@ -371,27 +496,27 @@ type MetricHistoPoint {
max: Int
}
type JobsStatistics {
id: ID! # If `groupBy` was used, ID of the user/project/cluster
name: String! # if User-Statistics: Given Name of Account (ID) Owner
totalJobs: Int! # Number of jobs
runningJobs: Int! # Number of running jobs
shortJobs: Int! # Number of jobs with a duration of less than duration
totalWalltime: Int! # Sum of the duration of all matched jobs in hours
totalNodes: Int! # Sum of the nodes of all matched jobs
totalNodeHours: Int! # Sum of the node hours of all matched jobs
totalCores: Int! # Sum of the cores of all matched jobs
totalCoreHours: Int! # Sum of the core hours of all matched jobs
totalAccs: Int! # Sum of the accs of all matched jobs
totalAccHours: Int! # Sum of the gpu hours of all matched jobs
histDuration: [HistoPoint!]! # value: hour, count: number of jobs with a rounded duration of value
histNumNodes: [HistoPoint!]! # value: number of nodes, count: number of jobs with that number of nodes
histNumCores: [HistoPoint!]! # value: number of cores, count: number of jobs with that number of cores
histNumAccs: [HistoPoint!]! # value: number of accs, count: number of jobs with that number of accs
histMetrics: [MetricHistoPoints!]! # metric: metricname, data array of histopoints: value: metric average bin, count: number of jobs with that metric average
type JobsStatistics {
id: ID! # If `groupBy` was used, ID of the user/project/cluster
name: String! # if User-Statistics: Given Name of Account (ID) Owner
totalJobs: Int! # Number of jobs
runningJobs: Int! # Number of running jobs
shortJobs: Int! # Number of jobs with a duration of less than duration
totalWalltime: Int! # Sum of the duration of all matched jobs in hours
totalNodes: Int! # Sum of the nodes of all matched jobs
totalNodeHours: Int! # Sum of the node hours of all matched jobs
totalCores: Int! # Sum of the cores of all matched jobs
totalCoreHours: Int! # Sum of the core hours of all matched jobs
totalAccs: Int! # Sum of the accs of all matched jobs
totalAccHours: Int! # Sum of the gpu hours of all matched jobs
histDuration: [HistoPoint!]! # value: hour, count: number of jobs with a rounded duration of value
histNumNodes: [HistoPoint!]! # value: number of nodes, count: number of jobs with that number of nodes
histNumCores: [HistoPoint!]! # value: number of cores, count: number of jobs with that number of cores
histNumAccs: [HistoPoint!]! # value: number of accs, count: number of jobs with that number of accs
histMetrics: [MetricHistoPoints!]! # metric: metricname, data array of histopoints: value: metric average bin, count: number of jobs with that metric average
}
input PageRequest {
itemsPerPage: Int!
page: Int!
page: Int!
}

273
api/swagger.json
View File

@@ -201,7 +201,7 @@
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/api.DefaultJobApiResponse"
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
@@ -271,7 +271,7 @@
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/api.DefaultJobApiResponse"
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
@@ -341,7 +341,7 @@
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/api.DefaultJobApiResponse"
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
@@ -523,7 +523,7 @@
"ApiKeyAuth": []
}
],
"description": "Job specified in request body will be saved to database as \"running\" with new DB ID.\nJob specifications follow the 'JobMeta' scheme, API will fail to execute if requirements are not met.",
"description": "Job specified in request body will be saved to database as \"running\" with new DB ID.\nJob specifications follow the 'Job' scheme, API will fail to execute if requirements are not met.",
"consumes": [
"application/json"
],
@@ -541,7 +541,7 @@
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/schema.JobMeta"
"$ref": "#/definitions/schema.Job"
}
}
],
@@ -549,7 +549,7 @@
"201": {
"description": "Job added successfully",
"schema": {
"$ref": "#/definitions/api.DefaultJobApiResponse"
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
@@ -592,7 +592,7 @@
"ApiKeyAuth": []
}
],
"description": "Job to stop is specified by request body. All fields are required in this case.\nReturns full job resource information according to 'JobMeta' scheme.",
"description": "Job to stop is specified by request body. All fields are required in this case.\nReturns full job resource information according to 'Job' scheme.",
"produces": [
"application/json"
],
@@ -615,7 +615,7 @@
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/schema.JobMeta"
"$ref": "#/definitions/schema.Job"
}
},
"400": {
@@ -737,7 +737,7 @@
"ApiKeyAuth": []
}
],
"description": "Job to get is specified by database ID\nReturns full job resource information according to 'JobMeta' scheme and all metrics according to 'JobData'.",
"description": "Job to get is specified by database ID\nReturns full job resource information according to 'Job' scheme and all metrics according to 'JobData'.",
"produces": [
"application/json"
],
@@ -811,7 +811,7 @@
"ApiKeyAuth": []
}
],
"description": "Job to get is specified by database ID\nReturns full job resource information according to 'JobMeta' scheme and all metrics according to 'JobData'.",
"description": "Job to get is specified by database ID\nReturns full job resource information according to 'Job' scheme and all metrics according to 'JobData'.",
"consumes": [
"application/json"
],
@@ -889,6 +889,66 @@
}
}
},
"/api/nodestats/": {
"post": {
"security": [
{
"ApiKeyAuth": []
}
],
"description": "Returns a JSON-encoded list of users.\nRequired query-parameter defines if all users or only users with additional special roles are returned.",
"produces": [
"application/json"
],
"tags": [
"Nodestates"
],
"summary": "Deliver updated Slurm node states",
"parameters": [
{
"description": "Request body containing nodes and their states",
"name": "request",
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/api.UpdateNodeStatesRequest"
}
}
],
"responses": {
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
"description": "Bad Request",
"schema": {
"$ref": "#/definitions/api.ErrorResponse"
}
},
"401": {
"description": "Unauthorized",
"schema": {
"$ref": "#/definitions/api.ErrorResponse"
}
},
"403": {
"description": "Forbidden",
"schema": {
"$ref": "#/definitions/api.ErrorResponse"
}
},
"500": {
"description": "Internal Server Error",
"schema": {
"$ref": "#/definitions/api.ErrorResponse"
}
}
}
}
},
"/api/users/": {
"get": {
"security": [
@@ -1137,7 +1197,7 @@
}
}
},
"api.DefaultJobApiResponse": {
"api.DefaultApiResponse": {
"type": "object",
"properties": {
"msg": {
@@ -1231,7 +1291,7 @@
"description": "Array of jobs",
"type": "array",
"items": {
"$ref": "#/definitions/schema.JobMeta"
"$ref": "#/definitions/schema.Job"
}
},
"page": {
@@ -1285,6 +1345,20 @@
}
}
},
"api.Node": {
"type": "object",
"properties": {
"hostname": {
"type": "string"
},
"states": {
"type": "array",
"items": {
"type": "string"
}
}
}
},
"api.StopJobApiRequest": {
"type": "object",
"required": [
@@ -1318,6 +1392,21 @@
}
}
},
"api.UpdateNodeStatesRequest": {
"type": "object",
"properties": {
"cluster": {
"type": "string",
"example": "fritz"
},
"nodes": {
"type": "array",
"items": {
"$ref": "#/definitions/api.Node"
}
}
}
},
"schema.Accelerator": {
"type": "object",
"properties": {
@@ -1353,7 +1442,6 @@
}
},
"schema.Job": {
"description": "Information of a HPC job.",
"type": "object",
"properties": {
"arrayJobId": {
@@ -1451,6 +1539,12 @@
"type": "string",
"example": "abcd200"
},
"requestedMemory": {
"description": "in MB",
"type": "integer",
"minimum": 1,
"example": 128000
},
"resources": {
"type": "array",
"items": {
@@ -1462,7 +1556,14 @@
"example": 4
},
"startTime": {
"type": "string"
"type": "integer",
"example": 1649723812
},
"statistics": {
"type": "object",
"additionalProperties": {
"$ref": "#/definitions/schema.JobStatistics"
}
},
"subCluster": {
"type": "string",
@@ -1510,147 +1611,6 @@
}
}
},
"schema.JobMeta": {
"description": "Meta data information of a HPC job.",
"type": "object",
"properties": {
"arrayJobId": {
"type": "integer",
"example": 123000
},
"cluster": {
"type": "string",
"example": "fritz"
},
"concurrentJobs": {
"$ref": "#/definitions/schema.JobLinkResultList"
},
"duration": {
"type": "integer",
"minimum": 1,
"example": 43200
},
"energy": {
"type": "number"
},
"energyFootprint": {
"type": "object",
"additionalProperties": {
"type": "number"
}
},
"exclusive": {
"type": "integer",
"maximum": 2,
"minimum": 0,
"example": 1
},
"footprint": {
"type": "object",
"additionalProperties": {
"type": "number"
}
},
"id": {
"type": "integer"
},
"jobId": {
"type": "integer",
"example": 123000
},
"jobState": {
"enum": [
"completed",
"failed",
"cancelled",
"stopped",
"timeout",
"out_of_memory"
],
"allOf": [
{
"$ref": "#/definitions/schema.JobState"
}
],
"example": "completed"
},
"metaData": {
"type": "object",
"additionalProperties": {
"type": "string"
}
},
"monitoringStatus": {
"type": "integer",
"maximum": 3,
"minimum": 0,
"example": 1
},
"numAcc": {
"type": "integer",
"minimum": 1,
"example": 2
},
"numHwthreads": {
"type": "integer",
"minimum": 1,
"example": 20
},
"numNodes": {
"type": "integer",
"minimum": 1,
"example": 2
},
"partition": {
"type": "string",
"example": "main"
},
"project": {
"type": "string",
"example": "abcd200"
},
"resources": {
"type": "array",
"items": {
"$ref": "#/definitions/schema.Resource"
}
},
"smt": {
"type": "integer",
"example": 4
},
"startTime": {
"type": "integer",
"minimum": 1,
"example": 1649723812
},
"statistics": {
"type": "object",
"additionalProperties": {
"$ref": "#/definitions/schema.JobStatistics"
}
},
"subCluster": {
"type": "string",
"example": "main"
},
"tags": {
"type": "array",
"items": {
"$ref": "#/definitions/schema.Tag"
}
},
"user": {
"type": "string",
"example": "abcd100h"
},
"walltime": {
"type": "integer",
"minimum": 1,
"example": 86400
}
}
},
"schema.JobMetric": {
"type": "object",
"properties": {
@@ -1978,6 +1938,9 @@
},
"remove": {
"type": "boolean"
},
"unit": {
"$ref": "#/definitions/schema.Unit"
}
}
},

201
api/swagger.yaml
View File

@@ -31,7 +31,7 @@ definitions:
example: Debug
type: string
type: object
api.DefaultJobApiResponse:
api.DefaultApiResponse:
properties:
msg:
type: string
@@ -96,7 +96,7 @@ definitions:
jobs:
description: Array of jobs
items:
$ref: '#/definitions/schema.JobMeta'
$ref: '#/definitions/schema.Job'
type: array
page:
description: Page id returned
@@ -132,6 +132,15 @@ definitions:
scope:
$ref: '#/definitions/schema.MetricScope'
type: object
api.Node:
properties:
hostname:
type: string
states:
items:
type: string
type: array
type: object
api.StopJobApiRequest:
properties:
cluster:
@@ -154,6 +163,16 @@ definitions:
- jobState
- stopTime
type: object
api.UpdateNodeStatesRequest:
properties:
cluster:
example: fritz
type: string
nodes:
items:
$ref: '#/definitions/api.Node'
type: array
type: object
schema.Accelerator:
properties:
id:
@@ -177,7 +196,6 @@ definitions:
type: array
type: object
schema.Job:
description: Information of a HPC job.
properties:
arrayJobId:
example: 123000
@@ -249,6 +267,11 @@ definitions:
project:
example: abcd200
type: string
requestedMemory:
description: in MB
example: 128000
minimum: 1
type: integer
resources:
items:
$ref: '#/definitions/schema.Resource'
@@ -257,7 +280,12 @@ definitions:
example: 4
type: integer
startTime:
type: string
example: 1649723812
type: integer
statistics:
additionalProperties:
$ref: '#/definitions/schema.JobStatistics'
type: object
subCluster:
example: main
type: string
@@ -289,109 +317,6 @@ definitions:
$ref: '#/definitions/schema.JobLink'
type: array
type: object
schema.JobMeta:
description: Meta data information of a HPC job.
properties:
arrayJobId:
example: 123000
type: integer
cluster:
example: fritz
type: string
concurrentJobs:
$ref: '#/definitions/schema.JobLinkResultList'
duration:
example: 43200
minimum: 1
type: integer
energy:
type: number
energyFootprint:
additionalProperties:
type: number
type: object
exclusive:
example: 1
maximum: 2
minimum: 0
type: integer
footprint:
additionalProperties:
type: number
type: object
id:
type: integer
jobId:
example: 123000
type: integer
jobState:
allOf:
- $ref: '#/definitions/schema.JobState'
enum:
- completed
- failed
- cancelled
- stopped
- timeout
- out_of_memory
example: completed
metaData:
additionalProperties:
type: string
type: object
monitoringStatus:
example: 1
maximum: 3
minimum: 0
type: integer
numAcc:
example: 2
minimum: 1
type: integer
numHwthreads:
example: 20
minimum: 1
type: integer
numNodes:
example: 2
minimum: 1
type: integer
partition:
example: main
type: string
project:
example: abcd200
type: string
resources:
items:
$ref: '#/definitions/schema.Resource'
type: array
smt:
example: 4
type: integer
startTime:
example: 1649723812
minimum: 1
type: integer
statistics:
additionalProperties:
$ref: '#/definitions/schema.JobStatistics'
type: object
subCluster:
example: main
type: string
tags:
items:
$ref: '#/definitions/schema.Tag'
type: array
user:
example: abcd100h
type: string
walltime:
example: 86400
minimum: 1
type: integer
type: object
schema.JobMetric:
properties:
series:
@@ -620,6 +545,8 @@ definitions:
type: number
remove:
type: boolean
unit:
$ref: '#/definitions/schema.Unit'
type: object
schema.Tag:
description: Defines a tag using name and type.
@@ -797,7 +724,7 @@ paths:
get:
description: |-
Job to get is specified by database ID
Returns full job resource information according to 'JobMeta' scheme and all metrics according to 'JobData'.
Returns full job resource information according to 'Job' scheme and all metrics according to 'JobData'.
parameters:
- description: Database ID of Job
in: path
@@ -850,7 +777,7 @@ paths:
- application/json
description: |-
Job to get is specified by database ID
Returns full job resource information according to 'JobMeta' scheme and all metrics according to 'JobData'.
Returns full job resource information according to 'Job' scheme and all metrics according to 'JobData'.
parameters:
- description: Database ID of Job
in: path
@@ -921,7 +848,7 @@ paths:
"200":
description: Success message
schema:
$ref: '#/definitions/api.DefaultJobApiResponse'
$ref: '#/definitions/api.DefaultApiResponse'
"400":
description: Bad Request
schema:
@@ -968,7 +895,7 @@ paths:
"200":
description: Success message
schema:
$ref: '#/definitions/api.DefaultJobApiResponse'
$ref: '#/definitions/api.DefaultApiResponse'
"400":
description: Bad Request
schema:
@@ -1015,7 +942,7 @@ paths:
"200":
description: Success message
schema:
$ref: '#/definitions/api.DefaultJobApiResponse'
$ref: '#/definitions/api.DefaultApiResponse'
"400":
description: Bad Request
schema:
@@ -1141,21 +1068,21 @@ paths:
- application/json
description: |-
Job specified in request body will be saved to database as "running" with new DB ID.
Job specifications follow the 'JobMeta' scheme, API will fail to execute if requirements are not met.
Job specifications follow the 'Job' scheme, API will fail to execute if requirements are not met.
parameters:
- description: Job to add
in: body
name: request
required: true
schema:
$ref: '#/definitions/schema.JobMeta'
$ref: '#/definitions/schema.Job'
produces:
- application/json
responses:
"201":
description: Job added successfully
schema:
$ref: '#/definitions/api.DefaultJobApiResponse'
$ref: '#/definitions/api.DefaultApiResponse'
"400":
description: Bad Request
schema:
@@ -1186,7 +1113,7 @@ paths:
post:
description: |-
Job to stop is specified by request body. All fields are required in this case.
Returns full job resource information according to 'JobMeta' scheme.
Returns full job resource information according to 'Job' scheme.
parameters:
- description: All fields required
in: body
@@ -1200,7 +1127,7 @@ paths:
"200":
description: Success message
schema:
$ref: '#/definitions/schema.JobMeta'
$ref: '#/definitions/schema.Job'
"400":
description: Bad Request
schema:
@@ -1280,6 +1207,46 @@ paths:
summary: Adds one or more tags to a job
tags:
- Job add and modify
/api/nodestats/:
post:
description: |-
Returns a JSON-encoded list of users.
Required query-parameter defines if all users or only users with additional special roles are returned.
parameters:
- description: Request body containing nodes and their states
in: body
name: request
required: true
schema:
$ref: '#/definitions/api.UpdateNodeStatesRequest'
produces:
- application/json
responses:
"200":
description: Success message
schema:
$ref: '#/definitions/api.DefaultApiResponse'
"400":
description: Bad Request
schema:
$ref: '#/definitions/api.ErrorResponse'
"401":
description: Unauthorized
schema:
$ref: '#/definitions/api.ErrorResponse'
"403":
description: Forbidden
schema:
$ref: '#/definitions/api.ErrorResponse'
"500":
description: Internal Server Error
schema:
$ref: '#/definitions/api.ErrorResponse'
security:
- ApiKeyAuth: []
summary: Deliver updated Slurm node states
tags:
- Nodestates
/api/users/:
get:
description: |-

8
cmd/cc-backend/cli.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package main
@@ -7,8 +7,9 @@ package main
import "flag"
var (
flagReinitDB, flagInit, flagServer, flagSyncLDAP, flagGops, flagMigrateDB, flagRevertDB, flagForceDB, flagDev, flagVersion, flagLogDateTime bool
flagNewUser, flagDelUser, flagGenJWT, flagConfigFile, flagImportJob, flagLogLevel string
flagReinitDB, flagInit, flagServer, flagSyncLDAP, flagGops, flagMigrateDB, flagRevertDB,
flagForceDB, flagDev, flagVersion, flagLogDateTime, flagApplyTags bool
flagNewUser, flagDelUser, flagGenJWT, flagConfigFile, flagImportJob, flagLogLevel string
)
func cliInit() {
@@ -21,6 +22,7 @@ func cliInit() {
flag.BoolVar(&flagVersion, "version", false, "Show version information and exit")
flag.BoolVar(&flagMigrateDB, "migrate-db", false, "Migrate database to supported version and exit")
flag.BoolVar(&flagRevertDB, "revert-db", false, "Migrate database to previous version and exit")
flag.BoolVar(&flagApplyTags, "apply-tags", false, "Run taggers on all completed jobs and exit")
flag.BoolVar(&flagForceDB, "force-db", false, "Force database version, clear dirty flag and exit")
flag.BoolVar(&flagLogDateTime, "logdate", false, "Set this flag to add date and time to log messages")
flag.StringVar(&flagConfigFile, "config", "./config.json", "Specify alternative path to `config.json`")

16
cmd/cc-backend/init.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package main
@@ -8,8 +8,8 @@ import (
"os"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/internal/util"
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/util"
)
const envString = `
@@ -73,23 +73,23 @@ const configString = `
func initEnv() {
if util.CheckFileExists("var") {
log.Exit("Directory ./var already exists. Cautiously exiting application initialization.")
cclog.Exit("Directory ./var already exists. Cautiously exiting application initialization.")
}
if err := os.WriteFile("config.json", []byte(configString), 0o666); err != nil {
log.Abortf("Could not write default ./config.json with permissions '0o666'. Application initialization failed, exited.\nError: %s\n", err.Error())
cclog.Abortf("Could not write default ./config.json with permissions '0o666'. Application initialization failed, exited.\nError: %s\n", err.Error())
}
if err := os.WriteFile(".env", []byte(envString), 0o666); err != nil {
log.Abortf("Could not write default ./.env file with permissions '0o666'. Application initialization failed, exited.\nError: %s\n", err.Error())
cclog.Abortf("Could not write default ./.env file with permissions '0o666'. Application initialization failed, exited.\nError: %s\n", err.Error())
}
if err := os.Mkdir("var", 0o777); err != nil {
log.Abortf("Could not create default ./var folder with permissions '0o777'. Application initialization failed, exited.\nError: %s\n", err.Error())
cclog.Abortf("Could not create default ./var folder with permissions '0o777'. Application initialization failed, exited.\nError: %s\n", err.Error())
}
err := repository.MigrateDB("sqlite3", "./var/job.db")
if err != nil {
log.Abortf("Could not initialize default sqlite3 database as './var/job.db'. Application initialization failed, exited.\nError: %s\n", err.Error())
cclog.Abortf("Could not initialize default sqlite3 database as './var/job.db'. Application initialization failed, exited.\nError: %s\n", err.Error())
}
}

88
cmd/cc-backend/main.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package main
@@ -19,12 +19,15 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/importer"
"github.com/ClusterCockpit/cc-backend/internal/metricdata"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/internal/tagger"
"github.com/ClusterCockpit/cc-backend/internal/taskManager"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/runtimeEnv"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/ClusterCockpit/cc-lib/util"
"github.com/google/gops/agent"
"github.com/joho/godotenv"
_ "github.com/go-sql-driver/mysql"
_ "github.com/mattn/go-sqlite3"
@@ -58,13 +61,12 @@ func main() {
os.Exit(0)
}
// Apply config flags for pkg/log
log.Init(flagLogLevel, flagLogDateTime)
cclog.Init(flagLogLevel, flagLogDateTime)
// If init flag set, run tasks here before any file dependencies cause errors
if flagInit {
initEnv()
log.Exit("Successfully setup environment!\n" +
cclog.Exit("Successfully setup environment!\n" +
"Please review config.json and .env and adjust it to your needs.\n" +
"Add your job-archive at ./var/job-archive.")
}
@@ -72,12 +74,13 @@ func main() {
// See https://github.com/google/gops (Runtime overhead is almost zero)
if flagGops {
if err := agent.Listen(agent.Options{}); err != nil {
log.Abortf("Could not start gops agent with 'gops/agent.Listen(agent.Options{})'. Application startup failed, exited.\nError: %s\n", err.Error())
cclog.Abortf("Could not start gops agent with 'gops/agent.Listen(agent.Options{})'. Application startup failed, exited.\nError: %s\n", err.Error())
}
}
if err := runtimeEnv.LoadEnv("./.env"); err != nil && !os.IsNotExist(err) {
log.Abortf("Could not parse existing .env file at location './.env'. Application startup failed, exited.\nError: %s\n", err.Error())
err := godotenv.Load()
if err != nil {
cclog.Abortf("Could not parse existing .env file at location './.env'. Application startup failed, exited.\nError: %s\n", err.Error())
}
// Initialize sub-modules and handle command line flags.
@@ -95,25 +98,25 @@ func main() {
if flagMigrateDB {
err := repository.MigrateDB(config.Keys.DBDriver, config.Keys.DB)
if err != nil {
log.Abortf("MigrateDB Failed: Could not migrate '%s' database at location '%s' to version %d.\nError: %s\n", config.Keys.DBDriver, config.Keys.DB, repository.Version, err.Error())
cclog.Abortf("MigrateDB Failed: Could not migrate '%s' database at location '%s' to version %d.\nError: %s\n", config.Keys.DBDriver, config.Keys.DB, repository.Version, err.Error())
}
log.Exitf("MigrateDB Success: Migrated '%s' database at location '%s' to version %d.\n", config.Keys.DBDriver, config.Keys.DB, repository.Version)
cclog.Exitf("MigrateDB Success: Migrated '%s' database at location '%s' to version %d.\n", config.Keys.DBDriver, config.Keys.DB, repository.Version)
}
if flagRevertDB {
err := repository.RevertDB(config.Keys.DBDriver, config.Keys.DB)
if err != nil {
log.Abortf("RevertDB Failed: Could not revert '%s' database at location '%s' to version %d.\nError: %s\n", config.Keys.DBDriver, config.Keys.DB, (repository.Version - 1), err.Error())
cclog.Abortf("RevertDB Failed: Could not revert '%s' database at location '%s' to version %d.\nError: %s\n", config.Keys.DBDriver, config.Keys.DB, (repository.Version - 1), err.Error())
}
log.Exitf("RevertDB Success: Reverted '%s' database at location '%s' to version %d.\n", config.Keys.DBDriver, config.Keys.DB, (repository.Version - 1))
cclog.Exitf("RevertDB Success: Reverted '%s' database at location '%s' to version %d.\n", config.Keys.DBDriver, config.Keys.DB, (repository.Version - 1))
}
if flagForceDB {
err := repository.ForceDB(config.Keys.DBDriver, config.Keys.DB)
if err != nil {
log.Abortf("ForceDB Failed: Could not force '%s' database at location '%s' to version %d.\nError: %s\n", config.Keys.DBDriver, config.Keys.DB, repository.Version, err.Error())
cclog.Abortf("ForceDB Failed: Could not force '%s' database at location '%s' to version %d.\nError: %s\n", config.Keys.DBDriver, config.Keys.DB, repository.Version, err.Error())
}
log.Exitf("ForceDB Success: Forced '%s' database at location '%s' to version %d.\n", config.Keys.DBDriver, config.Keys.DB, repository.Version)
cclog.Exitf("ForceDB Success: Forced '%s' database at location '%s' to version %d.\n", config.Keys.DBDriver, config.Keys.DB, repository.Version)
}
repository.Connect(config.Keys.DBDriver, config.Keys.DB)
@@ -125,7 +128,7 @@ func main() {
if flagNewUser != "" {
parts := strings.SplitN(flagNewUser, ":", 3)
if len(parts) != 3 || len(parts[0]) == 0 {
log.Abortf("Add User: Could not parse supplied argument format: No changes.\n"+
cclog.Abortf("Add User: Could not parse supplied argument format: No changes.\n"+
"Want: <username>:[admin,support,manager,api,user]:<password>\n"+
"Have: %s\n", flagNewUser)
}
@@ -134,18 +137,18 @@ func main() {
if err := ur.AddUser(&schema.User{
Username: parts[0], Projects: make([]string, 0), Password: parts[2], Roles: strings.Split(parts[1], ","),
}); err != nil {
log.Abortf("Add User: Could not add new user authentication for '%s' and roles '%s'.\nError: %s\n", parts[0], parts[1], err.Error())
cclog.Abortf("Add User: Could not add new user authentication for '%s' and roles '%s'.\nError: %s\n", parts[0], parts[1], err.Error())
} else {
log.Printf("Add User: Added new user '%s' with roles '%s'.\n", parts[0], parts[1])
cclog.Printf("Add User: Added new user '%s' with roles '%s'.\n", parts[0], parts[1])
}
}
if flagDelUser != "" {
ur := repository.GetUserRepository()
if err := ur.DelUser(flagDelUser); err != nil {
log.Abortf("Delete User: Could not delete user '%s' from DB.\nError: %s\n", flagDelUser, err.Error())
cclog.Abortf("Delete User: Could not delete user '%s' from DB.\nError: %s\n", flagDelUser, err.Error())
} else {
log.Printf("Delete User: Deleted user '%s' from DB.\n", flagDelUser)
cclog.Printf("Delete User: Deleted user '%s' from DB.\n", flagDelUser)
}
}
@@ -153,67 +156,78 @@ func main() {
if flagSyncLDAP {
if authHandle.LdapAuth == nil {
log.Abort("Sync LDAP: LDAP authentication is not configured, could not synchronize. No changes, exited.")
cclog.Abort("Sync LDAP: LDAP authentication is not configured, could not synchronize. No changes, exited.")
}
if err := authHandle.LdapAuth.Sync(); err != nil {
log.Abortf("Sync LDAP: Could not synchronize, failed with error.\nError: %s\n", err.Error())
cclog.Abortf("Sync LDAP: Could not synchronize, failed with error.\nError: %s\n", err.Error())
}
log.Print("Sync LDAP: LDAP synchronization successfull.")
cclog.Print("Sync LDAP: LDAP synchronization successfull.")
}
if flagGenJWT != "" {
ur := repository.GetUserRepository()
user, err := ur.GetUser(flagGenJWT)
if err != nil {
log.Abortf("JWT: Could not get supplied user '%s' from DB. No changes, exited.\nError: %s\n", flagGenJWT, err.Error())
cclog.Abortf("JWT: Could not get supplied user '%s' from DB. No changes, exited.\nError: %s\n", flagGenJWT, err.Error())
}
if !user.HasRole(schema.RoleApi) {
log.Warnf("JWT: User '%s' does not have the role 'api'. REST API endpoints will return error!\n", user.Username)
cclog.Warnf("JWT: User '%s' does not have the role 'api'. REST API endpoints will return error!\n", user.Username)
}
jwt, err := authHandle.JwtAuth.ProvideJWT(user)
if err != nil {
log.Abortf("JWT: User '%s' found in DB, but failed to provide JWT.\nError: %s\n", user.Username, err.Error())
cclog.Abortf("JWT: User '%s' found in DB, but failed to provide JWT.\nError: %s\n", user.Username, err.Error())
}
log.Printf("JWT: Successfully generated JWT for user '%s': %s\n", user.Username, jwt)
cclog.Printf("JWT: Successfully generated JWT for user '%s': %s\n", user.Username, jwt)
}
} else if flagNewUser != "" || flagDelUser != "" {
log.Abort("Error: Arguments '--add-user' and '--del-user' can only be used if authentication is enabled. No changes, exited.")
cclog.Abort("Error: Arguments '--add-user' and '--del-user' can only be used if authentication is enabled. No changes, exited.")
}
if err := archive.Init(config.Keys.Archive, config.Keys.DisableArchive); err != nil {
log.Abortf("Init: Failed to initialize archive.\nError: %s\n", err.Error())
cclog.Abortf("Init: Failed to initialize archive.\nError: %s\n", err.Error())
}
if err := metricdata.Init(); err != nil {
log.Abortf("Init: Failed to initialize metricdata repository.\nError %s\n", err.Error())
cclog.Abortf("Init: Failed to initialize metricdata repository.\nError %s\n", err.Error())
}
if flagReinitDB {
if err := importer.InitDB(); err != nil {
log.Abortf("Init DB: Failed to re-initialize repository DB.\nError: %s\n", err.Error())
cclog.Abortf("Init DB: Failed to re-initialize repository DB.\nError: %s\n", err.Error())
} else {
log.Print("Init DB: Sucessfully re-initialized repository DB.")
cclog.Print("Init DB: Sucessfully re-initialized repository DB.")
}
}
if flagImportJob != "" {
if err := importer.HandleImportFlag(flagImportJob); err != nil {
log.Abortf("Import Job: Job import failed.\nError: %s\n", err.Error())
cclog.Abortf("Import Job: Job import failed.\nError: %s\n", err.Error())
} else {
log.Printf("Import Job: Imported Job '%s' into DB.\n", flagImportJob)
cclog.Printf("Import Job: Imported Job '%s' into DB.\n", flagImportJob)
}
}
if config.Keys.EnableJobTaggers {
tagger.Init()
}
if flagApplyTags {
if err := tagger.RunTaggers(); err != nil {
cclog.Abortf("Running job taggers.\nError: %s\n", err.Error())
}
}
if !flagServer {
log.Exit("No errors, server flag not set. Exiting cc-backend.")
cclog.Exit("No errors, server flag not set. Exiting cc-backend.")
}
archiver.Start(repository.GetJobRepository())
taskManager.Start()
serverInit()
@@ -235,6 +249,8 @@ func main() {
serverShutdown()
util.FsWatcherShutdown()
taskManager.Shutdown()
}()
@@ -243,5 +259,5 @@ func main() {
}
runtimeEnv.SystemdNotifiy(true, "running")
wg.Wait()
log.Print("Graceful shutdown completed!")
cclog.Print("Graceful shutdown completed!")
}

24
cmd/cc-backend/server.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package main
@@ -27,9 +27,9 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/graph"
"github.com/ClusterCockpit/cc-backend/internal/graph/generated"
"github.com/ClusterCockpit/cc-backend/internal/routerConfig"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/runtimeEnv"
"github.com/ClusterCockpit/cc-backend/web"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/gorilla/handlers"
"github.com/gorilla/mux"
httpSwagger "github.com/swaggo/http-swagger"
@@ -101,7 +101,7 @@ func serverInit() {
router.HandleFunc("/login", func(rw http.ResponseWriter, r *http.Request) {
rw.Header().Add("Content-Type", "text/html; charset=utf-8")
log.Debugf("##%v##", info)
cclog.Debugf("##%v##", info)
web.RenderTemplate(rw, "login.tmpl", &web.Page{Title: "Login", Build: buildInfo, Infos: info})
}).Methods(http.MethodGet)
router.HandleFunc("/imprint", func(rw http.ResponseWriter, r *http.Request) {
@@ -237,7 +237,7 @@ func serverInit() {
if config.Keys.EmbedStaticFiles {
if i, err := os.Stat("./var/img"); err == nil {
if i.IsDir() {
log.Info("Use local directory for static images")
cclog.Info("Use local directory for static images")
router.PathPrefix("/img/").Handler(http.StripPrefix("/img/", http.FileServer(http.Dir("./var/img"))))
}
}
@@ -258,12 +258,12 @@ func serverInit() {
func serverStart() {
handler := handlers.CustomLoggingHandler(io.Discard, router, func(_ io.Writer, params handlers.LogFormatterParams) {
if strings.HasPrefix(params.Request.RequestURI, "/api/") {
log.Debugf("%s %s (%d, %.02fkb, %dms)",
cclog.Debugf("%s %s (%d, %.02fkb, %dms)",
params.Request.Method, params.URL.RequestURI(),
params.StatusCode, float32(params.Size)/1024,
time.Since(params.TimeStamp).Milliseconds())
} else {
log.Debugf("%s %s (%d, %.02fkb, %dms)",
cclog.Debugf("%s %s (%d, %.02fkb, %dms)",
params.Request.Method, params.URL.RequestURI(),
params.StatusCode, float32(params.Size)/1024,
time.Since(params.TimeStamp).Milliseconds())
@@ -280,7 +280,7 @@ func serverStart() {
// Start http or https server
listener, err := net.Listen("tcp", config.Keys.Addr)
if err != nil {
log.Abortf("Server Start: Starting http listener on '%s' failed.\nError: %s\n", config.Keys.Addr, err.Error())
cclog.Abortf("Server Start: Starting http listener on '%s' failed.\nError: %s\n", config.Keys.Addr, err.Error())
}
if !strings.HasSuffix(config.Keys.Addr, ":80") && config.Keys.RedirectHttpTo != "" {
@@ -293,7 +293,7 @@ func serverStart() {
cert, err := tls.LoadX509KeyPair(
config.Keys.HttpsCertFile, config.Keys.HttpsKeyFile)
if err != nil {
log.Abortf("Server Start: Loading X509 keypair failed. Check options 'https-cert-file' and 'https-key-file' in 'config.json'.\nError: %s\n", err.Error())
cclog.Abortf("Server Start: Loading X509 keypair failed. Check options 'https-cert-file' and 'https-key-file' in 'config.json'.\nError: %s\n", err.Error())
}
listener = tls.NewListener(listener, &tls.Config{
Certificates: []tls.Certificate{cert},
@@ -304,20 +304,20 @@ func serverStart() {
MinVersion: tls.VersionTLS12,
PreferServerCipherSuites: true,
})
log.Printf("HTTPS server listening at %s...\n", config.Keys.Addr)
cclog.Printf("HTTPS server listening at %s...\n", config.Keys.Addr)
} else {
log.Printf("HTTP server listening at %s...\n", config.Keys.Addr)
cclog.Printf("HTTP server listening at %s...\n", config.Keys.Addr)
}
//
// Because this program will want to bind to a privileged port (like 80), the listener must
// be established first, then the user can be changed, and after that,
// the actual http server can be started.
if err := runtimeEnv.DropPrivileges(config.Keys.Group, config.Keys.User); err != nil {
log.Abortf("Server Start: Error while preparing server start.\nError: %s\n", err.Error())
cclog.Abortf("Server Start: Error while preparing server start.\nError: %s\n", err.Error())
}
if err = server.Serve(listener); err != nil && err != http.ErrServerClosed {
log.Abortf("Server Start: Starting server failed.\nError: %s\n", err.Error())
cclog.Abortf("Server Start: Starting server failed.\nError: %s\n", err.Error())
}
}

35
go.mod
View File

@@ -6,9 +6,10 @@ toolchain go1.24.1
require (
github.com/99designs/gqlgen v0.17.66
github.com/ClusterCockpit/cc-units v0.4.0
github.com/ClusterCockpit/cc-lib v0.3.0
github.com/Masterminds/squirrel v1.5.4
github.com/coreos/go-oidc/v3 v3.12.0
github.com/expr-lang/expr v1.17.3
github.com/go-co-op/gocron/v2 v2.16.0
github.com/go-ldap/ldap/v3 v3.4.10
github.com/go-sql-driver/mysql v1.9.0
@@ -18,18 +19,17 @@ require (
github.com/gorilla/handlers v1.5.2
github.com/gorilla/mux v1.8.1
github.com/gorilla/sessions v1.4.0
github.com/influxdata/influxdb-client-go/v2 v2.14.0
github.com/jmoiron/sqlx v1.4.0
github.com/joho/godotenv v1.5.1
github.com/mattn/go-sqlite3 v1.14.24
github.com/prometheus/client_golang v1.21.0
github.com/prometheus/common v0.62.0
github.com/prometheus/client_golang v1.22.0
github.com/prometheus/common v0.63.0
github.com/qustavo/sqlhooks/v2 v2.1.0
github.com/santhosh-tekuri/jsonschema/v5 v5.3.1
github.com/swaggo/http-swagger v1.3.4
github.com/swaggo/swag v1.16.4
github.com/vektah/gqlparser/v2 v2.5.22
golang.org/x/crypto v0.35.0
golang.org/x/exp v0.0.0-20250218142911-aa4b98e5adaa
golang.org/x/crypto v0.37.0
golang.org/x/oauth2 v0.27.0
golang.org/x/time v0.5.0
)
@@ -39,11 +39,11 @@ require (
github.com/Azure/go-ntlmssp v0.0.0-20221128193559-754e69321358 // indirect
github.com/KyleBanks/depth v1.2.1 // indirect
github.com/agnivade/levenshtein v1.2.1 // indirect
github.com/apapsch/go-jsonmerge/v2 v2.0.0 // indirect
github.com/beorn7/perks v1.0.1 // indirect
github.com/cespare/xxhash/v2 v2.3.0 // indirect
github.com/cpuguy83/go-md2man/v2 v2.0.6 // indirect
github.com/felixge/httpsnoop v1.0.4 // indirect
github.com/fsnotify/fsnotify v1.9.0 // indirect
github.com/go-asn1-ber/asn1-ber v1.5.7 // indirect
github.com/go-jose/go-jose/v4 v4.0.5 // indirect
github.com/go-openapi/jsonpointer v0.21.0 // indirect
@@ -57,7 +57,6 @@ require (
github.com/hashicorp/errwrap v1.1.0 // indirect
github.com/hashicorp/go-multierror v1.1.1 // indirect
github.com/hashicorp/golang-lru/v2 v2.0.7 // indirect
github.com/influxdata/line-protocol v0.0.0-20210922203350-b1ad95c89adf // indirect
github.com/jonboulle/clockwork v0.5.0 // indirect
github.com/josharian/intern v1.0.0 // indirect
github.com/jpillora/backoff v1.0.0 // indirect
@@ -69,9 +68,8 @@ require (
github.com/modern-go/reflect2 v1.0.2 // indirect
github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
github.com/mwitkow/go-conntrack v0.0.0-20190716064945-2f068394615f // indirect
github.com/oapi-codegen/runtime v1.1.1 // indirect
github.com/prometheus/client_model v0.6.1 // indirect
github.com/prometheus/procfs v0.15.1 // indirect
github.com/prometheus/client_model v0.6.2 // indirect
github.com/prometheus/procfs v0.16.1 // indirect
github.com/robfig/cron/v3 v3.0.1 // indirect
github.com/russross/blackfriday/v2 v2.1.0 // indirect
github.com/sosodev/duration v1.3.1 // indirect
@@ -79,13 +77,14 @@ require (
github.com/urfave/cli/v2 v2.27.5 // indirect
github.com/xrash/smetrics v0.0.0-20240521201337-686a1a2994c1 // indirect
go.uber.org/atomic v1.11.0 // indirect
golang.org/x/mod v0.23.0 // indirect
golang.org/x/net v0.36.0 // indirect
golang.org/x/sync v0.11.0 // indirect
golang.org/x/sys v0.30.0 // indirect
golang.org/x/text v0.22.0 // indirect
golang.org/x/tools v0.30.0 // indirect
google.golang.org/protobuf v1.36.5 // indirect
golang.org/x/exp v0.0.0-20250408133849-7e4ce0ab07d0 // indirect
golang.org/x/mod v0.24.0 // indirect
golang.org/x/net v0.39.0 // indirect
golang.org/x/sync v0.13.0 // indirect
golang.org/x/sys v0.32.0 // indirect
golang.org/x/text v0.24.0 // indirect
golang.org/x/tools v0.32.0 // indirect
google.golang.org/protobuf v1.36.6 // indirect
gopkg.in/yaml.v2 v2.4.0 // indirect
gopkg.in/yaml.v3 v3.0.1 // indirect
sigs.k8s.io/yaml v1.4.0 // indirect

79
go.sum
View File

@@ -6,8 +6,8 @@ github.com/Azure/go-ansiterm v0.0.0-20230124172434-306776ec8161 h1:L/gRVlceqvL25
github.com/Azure/go-ansiterm v0.0.0-20230124172434-306776ec8161/go.mod h1:xomTg63KZ2rFqZQzSB4Vz2SUXa1BpHTVz9L5PTmPC4E=
github.com/Azure/go-ntlmssp v0.0.0-20221128193559-754e69321358 h1:mFRzDkZVAjdal+s7s0MwaRv9igoPqLRdzOLzw/8Xvq8=
github.com/Azure/go-ntlmssp v0.0.0-20221128193559-754e69321358/go.mod h1:chxPXzSsl7ZWRAuOIE23GDNzjWuZquvFlgA8xmpunjU=
github.com/ClusterCockpit/cc-units v0.4.0 h1:zP5DOu99GmErW0tCDf0gcLrlWt42RQ9dpoONEOh4cI0=
github.com/ClusterCockpit/cc-units v0.4.0/go.mod h1:3S3PAhAayS3pbgcT4q9Vn9VJw22Op51X0YimtG77zBw=
github.com/ClusterCockpit/cc-lib v0.3.0 h1:HEWOgnzRM01U10ZFfpiUWMzkLHg5nPdXZqdsiI2q4x0=
github.com/ClusterCockpit/cc-lib v0.3.0/go.mod h1:7CuXVNIJdynMZf6B9v4m54VCbbFg3ZD0tvLw2bVxN0A=
github.com/KyleBanks/depth v1.2.1 h1:5h8fQADFrWtarTdtDudMmGsC7GPbOAu6RVB3ffsVFHc=
github.com/KyleBanks/depth v1.2.1/go.mod h1:jzSb9d0L43HxTQfT+oSA1EEp2q+ne2uh6XgeJcm8brE=
github.com/Masterminds/squirrel v1.5.4 h1:uUcX/aBc8O7Fg9kaISIUsHXdKuqehiXAMQTYX8afzqM=
@@ -16,7 +16,6 @@ github.com/Microsoft/go-winio v0.6.2 h1:F2VQgta7ecxGYO8k3ZZz3RS8fVIXVxONVUPlNERo
github.com/Microsoft/go-winio v0.6.2/go.mod h1:yd8OoFMLzJbo9gZq8j5qaps8bJ9aShtEA8Ipt1oGCvU=
github.com/PuerkitoBio/goquery v1.9.3 h1:mpJr/ikUA9/GNJB/DBZcGeFDXUtosHRyRrwh7KGdTG0=
github.com/PuerkitoBio/goquery v1.9.3/go.mod h1:1ndLHPdTz+DyQPICCWYlYQMPl0oXZj0G6D4LCYA6u4U=
github.com/RaveNoX/go-jsoncommentstrip v1.0.0/go.mod h1:78ihd09MekBnJnxpICcwzCMzGrKSKYe4AqU6PDYYpjk=
github.com/agnivade/levenshtein v1.2.1 h1:EHBY3UOn1gwdy/VbFwgo4cxecRznFk7fKWN1KOX7eoM=
github.com/agnivade/levenshtein v1.2.1/go.mod h1:QVVI16kDrtSuwcpd0p1+xMC6Z/VfhtCyDIjcwga4/DU=
github.com/alexbrainman/sspi v0.0.0-20231016080023-1a75b4708caa h1:LHTHcTQiSGT7VVbI0o4wBRNQIgn917usHWOd6VAffYI=
@@ -25,13 +24,10 @@ github.com/andreyvit/diff v0.0.0-20170406064948-c7f18ee00883 h1:bvNMNQO63//z+xNg
github.com/andreyvit/diff v0.0.0-20170406064948-c7f18ee00883/go.mod h1:rCTlJbsFo29Kk6CurOXKm700vrz8f0KW0JNfpkRJY/8=
github.com/andybalholm/cascadia v1.3.2 h1:3Xi6Dw5lHF15JtdcmAHD3i1+T8plmv7BQ/nsViSLyss=
github.com/andybalholm/cascadia v1.3.2/go.mod h1:7gtRlve5FxPPgIgX36uWBX58OdBsSS6lUvCFb+h7KvU=
github.com/apapsch/go-jsonmerge/v2 v2.0.0 h1:axGnT1gRIfimI7gJifB699GoE/oq+F2MU7Dml6nw9rQ=
github.com/apapsch/go-jsonmerge/v2 v2.0.0/go.mod h1:lvDnEdqiQrp0O42VQGgmlKpxL1AP2+08jFMw88y4klk=
github.com/arbovm/levenshtein v0.0.0-20160628152529-48b4e1c0c4d0 h1:jfIu9sQUG6Ig+0+Ap1h4unLjW6YQJpKZVmUzxsD4E/Q=
github.com/arbovm/levenshtein v0.0.0-20160628152529-48b4e1c0c4d0/go.mod h1:t2tdKJDJF9BV14lnkjHmOQgcvEKgtqs5a1N3LNdJhGE=
github.com/beorn7/perks v1.0.1 h1:VlbKKnNfV8bJzeqoa4cOKqO6bYr3WgKZxO8Z16+hsOM=
github.com/beorn7/perks v1.0.1/go.mod h1:G2ZrVWU2WbWT9wwq4/hrbKbnv/1ERSJQ0ibhJ6rlkpw=
github.com/bmatcuk/doublestar v1.1.1/go.mod h1:UD6OnuiIn0yFxxA2le/rnRU1G4RaI4UvFv1sNto9p6w=
github.com/cespare/xxhash/v2 v2.3.0 h1:UL815xU9SqsFlibzuggzjXhog7bL6oX9BbNZnL2UFvs=
github.com/cespare/xxhash/v2 v2.3.0/go.mod h1:VGX0DQ3Q6kWi7AoAeZDth3/j3BFtOZR5XLFGgcrjCOs=
github.com/coreos/go-oidc/v3 v3.12.0 h1:sJk+8G2qq94rDI6ehZ71Bol3oUHy63qNYmkiSjrc/Jo=
@@ -53,8 +49,12 @@ github.com/docker/go-connections v0.5.0 h1:USnMq7hx7gwdVZq1L49hLXaFtUdTADjXGp+uj
github.com/docker/go-connections v0.5.0/go.mod h1:ov60Kzw0kKElRwhNs9UlUHAE/F9Fe6GLaXnqyDdmEXc=
github.com/docker/go-units v0.5.0 h1:69rxXcBk27SvSaaxTtLh/8llcHD8vYHT7WSdRZ/jvr4=
github.com/docker/go-units v0.5.0/go.mod h1:fgPhTUdO+D/Jk86RDLlptpiXQzgHJF7gydDDbaIK4Dk=
github.com/expr-lang/expr v1.17.3 h1:myeTTuDFz7k6eFe/JPlep/UsiIjVhG61FMHFu63U7j0=
github.com/expr-lang/expr v1.17.3/go.mod h1:8/vRC7+7HBzESEqt5kKpYXxrxkr31SaO8r40VO/1IT4=
github.com/felixge/httpsnoop v1.0.4 h1:NFTV2Zj1bL4mc9sqWACXbQFVBBg2W3GPvqp8/ESS2Wg=
github.com/felixge/httpsnoop v1.0.4/go.mod h1:m8KPJKqk1gH5J9DgRY2ASl2lWCfGKXixSwevea8zH2U=
github.com/fsnotify/fsnotify v1.9.0 h1:2Ml+OJNzbYCTzsxtv8vKSFD9PbJjmhYF14k/jKC7S9k=
github.com/fsnotify/fsnotify v1.9.0/go.mod h1:8jBTzvmWwFyi3Pb8djgCCO5IBqzKJ/Jwo8TRcHyHii0=
github.com/go-asn1-ber/asn1-ber v1.5.7 h1:DTX+lbVTWaTw1hQ+PbZPlnDZPEIs0SS/GCZAl535dDk=
github.com/go-asn1-ber/asn1-ber v1.5.7/go.mod h1:hEBeB/ic+5LoWskz+yKT7vGhhPYkProFKoKdwZRWMe0=
github.com/go-co-op/gocron/v2 v2.16.0 h1:uqUF6WFZ4enRU45pWFNcn1xpDLc+jBOTKhPQI16Z1xs=
@@ -88,8 +88,9 @@ github.com/golang-jwt/jwt/v5 v5.2.2/go.mod h1:pqrtFR0X4osieyHYxtmOUWsAWrfe1Q5UVI
github.com/golang-migrate/migrate/v4 v4.18.2 h1:2VSCMz7x7mjyTXx3m2zPokOY82LTRgxK1yQYKo6wWQ8=
github.com/golang-migrate/migrate/v4 v4.18.2/go.mod h1:2CM6tJvn2kqPXwnXO/d3rAQYiyoIm180VsO8PRX6Rpk=
github.com/google/go-cmp v0.5.9/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
github.com/google/go-cmp v0.6.0 h1:ofyhxvXcZhMsU5ulbFiLKl/XBFqE1GSq7atu8tAmTRI=
github.com/google/go-cmp v0.6.0/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
github.com/google/go-cmp v0.7.0 h1:wk8382ETsv4JYUZwIsn6YpYiWiBsYLSJiTsyBybVuN8=
github.com/google/go-cmp v0.7.0/go.mod h1:pXiqmnSA92OHEEa9HXL2W4E7lf9JzCmGVUdgjX3N/iU=
github.com/google/gofuzz v1.0.0/go.mod h1:dBl0BpW6vV/+mYPU4Po3pmUjxk6FQPldtuIdl/M65Eg=
github.com/google/gofuzz v1.2.0 h1:xRy4A+RhZaiKjJ1bPfwQ8sedCA+YS2YcCHW6ec7JMi0=
github.com/google/gofuzz v1.2.0/go.mod h1:dBl0BpW6vV/+mYPU4Po3pmUjxk6FQPldtuIdl/M65Eg=
@@ -119,10 +120,6 @@ github.com/hashicorp/go-uuid v1.0.3 h1:2gKiV6YVmrJ1i2CKKa9obLvRieoRGviZFL26PcT/C
github.com/hashicorp/go-uuid v1.0.3/go.mod h1:6SBZvOh/SIDV7/2o3Jml5SYk/TvGqwFJ/bN7x4byOro=
github.com/hashicorp/golang-lru/v2 v2.0.7 h1:a+bsQ5rvGLjzHuww6tVxozPZFVghXaHOwFs4luLUK2k=
github.com/hashicorp/golang-lru/v2 v2.0.7/go.mod h1:QeFd9opnmA6QUJc5vARoKUSoFhyfM2/ZepoAG6RGpeM=
github.com/influxdata/influxdb-client-go/v2 v2.14.0 h1:AjbBfJuq+QoaXNcrova8smSjwJdUHnwvfjMF71M1iI4=
github.com/influxdata/influxdb-client-go/v2 v2.14.0/go.mod h1:Ahpm3QXKMJslpXl3IftVLVezreAUtBOTZssDrjZEFHI=
github.com/influxdata/line-protocol v0.0.0-20210922203350-b1ad95c89adf h1:7JTmneyiNEwVBOHSjoMxiWAqB992atOeepeFYegn5RU=
github.com/influxdata/line-protocol v0.0.0-20210922203350-b1ad95c89adf/go.mod h1:xaLFMmpvUxqXtVkUJfg9QmT88cDaCJ3ZKgdZ78oO8Qo=
github.com/jcmturner/aescts/v2 v2.0.0 h1:9YKLH6ey7H4eDBXW8khjYslgyqG2xZikXP0EQFKrle8=
github.com/jcmturner/aescts/v2 v2.0.0/go.mod h1:AiaICIRyfYg35RUkr8yESTqvSy7csK90qZ5xfvvsoNs=
github.com/jcmturner/dnsutils/v2 v2.0.0 h1:lltnkeZGL0wILNvrNiVCR6Ro5PGU/SeBvVO/8c/iPbo=
@@ -137,6 +134,8 @@ github.com/jcmturner/rpc/v2 v2.0.3 h1:7FXXj8Ti1IaVFpSAziCZWNzbNuZmnvw/i6CqLNdWfZ
github.com/jcmturner/rpc/v2 v2.0.3/go.mod h1:VUJYCIDm3PVOEHw8sgt091/20OJjskO/YJki3ELg/Hc=
github.com/jmoiron/sqlx v1.4.0 h1:1PLqN7S1UYp5t4SrVVnt4nUVNemrDAtxlulVe+Qgm3o=
github.com/jmoiron/sqlx v1.4.0/go.mod h1:ZrZ7UsYB/weZdl2Bxg6jCRO9c3YHl8r3ahlKmRT4JLY=
github.com/joho/godotenv v1.5.1 h1:7eLL/+HRGLY0ldzfGMeQkb7vMd0as4CfYvUVzLqw0N0=
github.com/joho/godotenv v1.5.1/go.mod h1:f4LDr5Voq0i2e/R5DDNOoa2zzDfwtkZa6DnEwAbqwq4=
github.com/jonboulle/clockwork v0.5.0 h1:Hyh9A8u51kptdkR+cqRpT1EebBwTn1oK9YfGYbdFz6I=
github.com/jonboulle/clockwork v0.5.0/go.mod h1:3mZlmanh0g2NDKO5TWZVJAfofYk64M7XN3SzBPjZF60=
github.com/josharian/intern v1.0.0 h1:vlS4z54oSdjm0bgjRigI+G1HpF+tI+9rE5LLzOg8HmY=
@@ -145,9 +144,6 @@ github.com/jpillora/backoff v1.0.0 h1:uvFg412JmmHBHw7iwprIxkPMI+sGQ4kzOWsMeHnm2E
github.com/jpillora/backoff v1.0.0/go.mod h1:J/6gKK9jxlEcS3zixgDgUAsiuZ7yrSoa/FX5e0EB2j4=
github.com/json-iterator/go v1.1.12 h1:PV8peI4a0ysnczrg+LtxykD8LfKY9ML6u2jnxaEnrnM=
github.com/json-iterator/go v1.1.12/go.mod h1:e30LSqwooZae/UwlEbR2852Gd8hjQvJoHmT4TnhNGBo=
github.com/juju/gnuflag v0.0.0-20171113085948-2ce1bb71843d/go.mod h1:2PavIy+JPciBPrBUjwbNvtwB6RQlve+hkpll6QSNmOE=
github.com/klauspost/compress v1.17.11 h1:In6xLpyWOi1+C7tXUUWv2ot1QvBjxevKAaI6IXrJmUc=
github.com/klauspost/compress v1.17.11/go.mod h1:pMDklpSncoRMuLFrf1W9Ss9KT+0rH90U12bZKk7uwG0=
github.com/kr/pretty v0.3.1 h1:flRD4NNwYAUpkphVc1HcthR4KEIFJ65n8Mw5qdRn3LE=
github.com/kr/pretty v0.3.1/go.mod h1:hoEshYVHaxMs3cyo3Yncou5ZscifuDolrwPKZanG3xk=
github.com/kr/text v0.2.0 h1:5Nx0Ya0ZqY2ygV366QzturHI13Jq95ApcVaJBhpS+AY=
@@ -180,8 +176,6 @@ github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 h1:C3w9PqII01/Oq
github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822/go.mod h1:+n7T8mK8HuQTcFwEeznm/DIxMOiR9yIdICNftLE1DvQ=
github.com/mwitkow/go-conntrack v0.0.0-20190716064945-2f068394615f h1:KUppIJq7/+SVif2QVs3tOP0zanoHgBEVAwHxUSIzRqU=
github.com/mwitkow/go-conntrack v0.0.0-20190716064945-2f068394615f/go.mod h1:qRWi+5nqEBWmkhHvq77mSJWrCKwh8bxhgT7d/eI7P4U=
github.com/oapi-codegen/runtime v1.1.1 h1:EXLHh0DXIJnWhdRPN2w4MXAzFyE4CskzhNLUmtpMYro=
github.com/oapi-codegen/runtime v1.1.1/go.mod h1:SK9X900oXmPWilYR5/WKPzt3Kqxn/uS/+lbpREv+eCg=
github.com/opencontainers/go-digest v1.0.0 h1:apOUWs51W5PlhuyGyz9FCeeBIOUDA/6nW8Oi/yOhh5U=
github.com/opencontainers/go-digest v1.0.0/go.mod h1:0JzlMkj0TRzQZfJkVvzbP0HBR3IKzErnv2BNG4W4MAM=
github.com/opencontainers/image-spec v1.1.0 h1:8SG7/vwALn54lVB/0yZ/MMwhFrPYtpEHQb2IpWsCzug=
@@ -191,14 +185,14 @@ github.com/pkg/errors v0.9.1 h1:FEBLx1zS214owpjy7qsBeixbURkuhQAwrK5UwLGTwt4=
github.com/pkg/errors v0.9.1/go.mod h1:bwawxfHBFNV+L2hUp1rHADufV3IMtnDRdf1r5NINEl0=
github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
github.com/prometheus/client_golang v1.21.0 h1:DIsaGmiaBkSangBgMtWdNfxbMNdku5IK6iNhrEqWvdA=
github.com/prometheus/client_golang v1.21.0/go.mod h1:U9NM32ykUErtVBxdvD3zfi+EuFkkaBvMb09mIfe0Zgg=
github.com/prometheus/client_model v0.6.1 h1:ZKSh/rekM+n3CeS952MLRAdFwIKqeY8b62p8ais2e9E=
github.com/prometheus/client_model v0.6.1/go.mod h1:OrxVMOVHjw3lKMa8+x6HeMGkHMQyHDk9E3jmP2AmGiY=
github.com/prometheus/common v0.62.0 h1:xasJaQlnWAeyHdUBeGjXmutelfJHWMRr+Fg4QszZ2Io=
github.com/prometheus/common v0.62.0/go.mod h1:vyBcEuLSvWos9B1+CyL7JZ2up+uFzXhkqml0W5zIY1I=
github.com/prometheus/procfs v0.15.1 h1:YagwOFzUgYfKKHX6Dr+sHT7km/hxC76UB0learggepc=
github.com/prometheus/procfs v0.15.1/go.mod h1:fB45yRUv8NstnjriLhBQLuOUt+WW4BsoGhij/e3PBqk=
github.com/prometheus/client_golang v1.22.0 h1:rb93p9lokFEsctTys46VnV1kLCDpVZ0a/Y92Vm0Zc6Q=
github.com/prometheus/client_golang v1.22.0/go.mod h1:R7ljNsLXhuQXYZYtw6GAE9AZg8Y7vEW5scdCXrWRXC0=
github.com/prometheus/client_model v0.6.2 h1:oBsgwpGs7iVziMvrGhE53c/GrLUsZdHnqNwqPLxwZyk=
github.com/prometheus/client_model v0.6.2/go.mod h1:y3m2F6Gdpfy6Ut/GBsUqTWZqCUvMVzSfMLjcu6wAwpE=
github.com/prometheus/common v0.63.0 h1:YR/EIY1o3mEFP/kZCD7iDMnLPlGyuU2Gb3HIcXnA98k=
github.com/prometheus/common v0.63.0/go.mod h1:VVFF/fBIoToEnWRVkYoXEkq3R3paCoxG9PXP74SnV18=
github.com/prometheus/procfs v0.16.1 h1:hZ15bTNuirocR6u0JZ6BAHHmwS1p8B4P6MRqxtzMyRg=
github.com/prometheus/procfs v0.16.1/go.mod h1:teAbpZRB1iIAJYREa1LsoWUXykVXA1KlTmWl8x/U+Is=
github.com/qustavo/sqlhooks/v2 v2.1.0 h1:54yBemHnGHp/7xgT+pxwmIlMSDNYKx5JW5dfRAiCZi0=
github.com/qustavo/sqlhooks/v2 v2.1.0/go.mod h1:aMREyKo7fOKTwiLuWPsaHRXEmtqG4yREztO0idF83AU=
github.com/robfig/cron/v3 v3.0.1 h1:WdRxkvbJztn8LMz/QEvLN5sBU+xKpSqwwUO1Pjr4qDs=
@@ -213,7 +207,6 @@ github.com/sergi/go-diff v1.3.1 h1:xkr+Oxo4BOQKmkn/B9eMK0g5Kg/983T9DqqPHwYqD+8=
github.com/sergi/go-diff v1.3.1/go.mod h1:aMJSSKb2lpPvRNec0+w3fl7LP9IOFzdc9Pa4NFbPK1I=
github.com/sosodev/duration v1.3.1 h1:qtHBDMQ6lvMQsL15g4aopM4HEfOaYuhWBw3NPTtlqq4=
github.com/sosodev/duration v1.3.1/go.mod h1:RQIBBX0+fMLc/D9+Jb/fwvVmo0eZvDDEERAikUR6SDg=
github.com/spkg/bom v0.0.0-20160624110644-59b7046e48ad/go.mod h1:qLr4V1qq6nMqFKkMo8ZTx3f+BZEkzsRUY10Xsm2mwU0=
github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
github.com/stretchr/objx v0.4.0/go.mod h1:YvHI0jy2hoMjB+UWwv71VJQ9isScKT/TqJzVSSt89Yw=
github.com/stretchr/objx v0.5.0/go.mod h1:Yh+to48EsGEfYuaHDzXPcE3xhTkx73EhmCGUpEOglKo=
@@ -257,17 +250,17 @@ golang.org/x/crypto v0.13.0/go.mod h1:y6Z2r+Rw4iayiXXAIxJIDAJ1zMW4yaTpebo8fPOliY
golang.org/x/crypto v0.19.0/go.mod h1:Iy9bg/ha4yyC70EfRS8jz+B6ybOBKMaSxLj6P6oBDfU=
golang.org/x/crypto v0.23.0/go.mod h1:CKFgDieR+mRhux2Lsu27y0fO304Db0wZe70UKqHu0v8=
golang.org/x/crypto v0.31.0/go.mod h1:kDsLvtWBEx7MV9tJOj9bnXsPbxwJQ6csT/x4KIN4Ssk=
golang.org/x/crypto v0.35.0 h1:b15kiHdrGCHrP6LvwaQ3c03kgNhhiMgvlhxHQhmg2Xs=
golang.org/x/crypto v0.35.0/go.mod h1:dy7dXNW32cAb/6/PRuTNsix8T+vJAqvuIy5Bli/x0YQ=
golang.org/x/exp v0.0.0-20250218142911-aa4b98e5adaa h1:t2QcU6V556bFjYgu4L6C+6VrCPyJZ+eyRsABUPs1mz4=
golang.org/x/exp v0.0.0-20250218142911-aa4b98e5adaa/go.mod h1:BHOTPb3L19zxehTsLoJXVaTktb06DFgmdW6Wb9s8jqk=
golang.org/x/crypto v0.37.0 h1:kJNSjF/Xp7kU0iB2Z+9viTPMW4EqqsrywMXLJOOsXSE=
golang.org/x/crypto v0.37.0/go.mod h1:vg+k43peMZ0pUMhYmVAWysMK35e6ioLh3wB8ZCAfbVc=
golang.org/x/exp v0.0.0-20250408133849-7e4ce0ab07d0 h1:R84qjqJb5nVJMxqWYb3np9L5ZsaDtB+a39EqjV0JSUM=
golang.org/x/exp v0.0.0-20250408133849-7e4ce0ab07d0/go.mod h1:S9Xr4PYopiDyqSyp5NjCrhFrqg6A5zA2E/iPHPhqnS8=
golang.org/x/mod v0.6.0-dev.0.20220419223038-86c51ed26bb4/go.mod h1:jJ57K6gSWd91VN4djpZkiMVwK6gcyfeH4XE8wZrZaV4=
golang.org/x/mod v0.8.0/go.mod h1:iBbtSCu2XBx23ZKBPSOrRkjjQPZFPuis4dIYUhu/chs=
golang.org/x/mod v0.12.0/go.mod h1:iBbtSCu2XBx23ZKBPSOrRkjjQPZFPuis4dIYUhu/chs=
golang.org/x/mod v0.15.0/go.mod h1:hTbmBsO62+eylJbnUtE2MGJUyE7QWk4xUqPFrRgJ+7c=
golang.org/x/mod v0.17.0/go.mod h1:hTbmBsO62+eylJbnUtE2MGJUyE7QWk4xUqPFrRgJ+7c=
golang.org/x/mod v0.23.0 h1:Zb7khfcRGKk+kqfxFaP5tZqCnDZMjC5VtUBs87Hr6QM=
golang.org/x/mod v0.23.0/go.mod h1:6SkKJ3Xj0I0BrPOZoBy3bdMptDDU9oJrpohJ3eWZ1fY=
golang.org/x/mod v0.24.0 h1:ZfthKaKaT4NrhGVZHO1/WDTwGES4De8KtWO0SIbNJMU=
golang.org/x/mod v0.24.0/go.mod h1:IXM97Txy2VM4PJ3gI61r1YEk/gAj6zAHN3AdZt6S9Ww=
golang.org/x/net v0.0.0-20190620200207-3b0461eec859/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
golang.org/x/net v0.0.0-20200114155413-6afb5195e5aa/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
@@ -279,8 +272,8 @@ golang.org/x/net v0.15.0/go.mod h1:idbUs1IY1+zTqbi8yxTbhexhEEk5ur9LInksu6HrEpk=
golang.org/x/net v0.21.0/go.mod h1:bIjVDfnllIU7BJ2DNgfnXvpSvtn8VRwhlsaeUTyUS44=
golang.org/x/net v0.25.0/go.mod h1:JkAGAh7GEvH74S6FOH42FLoXpXbE/aqXSrIQjXgsiwM=
golang.org/x/net v0.33.0/go.mod h1:HXLR5J+9DxmrqMwG9qjGCxZ+zKXxBru04zlTvWlWuN4=
golang.org/x/net v0.36.0 h1:vWF2fRbw4qslQsQzgFqZff+BItCvGFQqKzKIzx1rmoA=
golang.org/x/net v0.36.0/go.mod h1:bFmbeoIPfrw4sMHNhb4J9f6+tPziuGjq7Jk/38fxi1I=
golang.org/x/net v0.39.0 h1:ZCu7HMWDxpXpaiKdhzIfaltL9Lp31x/3fCP11bc6/fY=
golang.org/x/net v0.39.0/go.mod h1:X7NRbYVEA+ewNkCNyJ513WmMdQ3BineSwVtN2zD/d+E=
golang.org/x/oauth2 v0.27.0 h1:da9Vo7/tDv5RH/7nZDz1eMGS/q1Vv1N/7FCrBhI9I3M=
golang.org/x/oauth2 v0.27.0/go.mod h1:onh5ek6nERTohokkhCD/y2cV4Do3fxFHFuAejCkRWT8=
golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
@@ -290,8 +283,8 @@ golang.org/x/sync v0.3.0/go.mod h1:FU7BRWz2tNW+3quACPkgCx/L+uEAv1htQ0V83Z9Rj+Y=
golang.org/x/sync v0.6.0/go.mod h1:Czt+wKu1gCyEFDUtn0jG5QVvpJ6rzVqr5aXyt9drQfk=
golang.org/x/sync v0.7.0/go.mod h1:Czt+wKu1gCyEFDUtn0jG5QVvpJ6rzVqr5aXyt9drQfk=
golang.org/x/sync v0.10.0/go.mod h1:Czt+wKu1gCyEFDUtn0jG5QVvpJ6rzVqr5aXyt9drQfk=
golang.org/x/sync v0.11.0 h1:GGz8+XQP4FvTTrjZPzNKTMFtSXH80RAzG+5ghFPgK9w=
golang.org/x/sync v0.11.0/go.mod h1:Czt+wKu1gCyEFDUtn0jG5QVvpJ6rzVqr5aXyt9drQfk=
golang.org/x/sync v0.13.0 h1:AauUjRAJ9OSnvULf/ARrrVywoJDy0YS2AwQ98I37610=
golang.org/x/sync v0.13.0/go.mod h1:1dzgHSNfp02xaA81J2MS99Qcpr2w7fw1gpm99rleRqA=
golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
golang.org/x/sys v0.0.0-20210615035016-665e8c7367d1/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
@@ -303,8 +296,8 @@ golang.org/x/sys v0.12.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
golang.org/x/sys v0.17.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
golang.org/x/sys v0.20.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
golang.org/x/sys v0.28.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
golang.org/x/sys v0.30.0 h1:QjkSwP/36a20jFYWkSue1YwXzLmsV5Gfq7Eiy72C1uc=
golang.org/x/sys v0.30.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
golang.org/x/sys v0.32.0 h1:s77OFDvIQeibCmezSnk/q6iAfkdiQaJi4VzroCFrN20=
golang.org/x/sys v0.32.0/go.mod h1:BJP2sWEmIv4KK5OTEluFJCKSidICx8ciO85XgH3Ak8k=
golang.org/x/telemetry v0.0.0-20240228155512-f48c80bd79b2/go.mod h1:TeRTkGYfJXctD9OcfyVLyj2J3IxLnKwHJR8f4D8a3YE=
golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
golang.org/x/term v0.0.0-20210927222741-03fcf44c2211/go.mod h1:jbD1KX2456YbFQfuXm/mYQcufACuNUgVhRMnK/tPxf8=
@@ -323,8 +316,8 @@ golang.org/x/text v0.13.0/go.mod h1:TvPlkZtksWOMsz7fbANvkp4WM8x/WCo/om8BMLbz+aE=
golang.org/x/text v0.14.0/go.mod h1:18ZOQIKpY8NJVqYksKHtTdi31H5itFRjB5/qKTNYzSU=
golang.org/x/text v0.15.0/go.mod h1:18ZOQIKpY8NJVqYksKHtTdi31H5itFRjB5/qKTNYzSU=
golang.org/x/text v0.21.0/go.mod h1:4IBbMaMmOPCJ8SecivzSH54+73PCFmPWxNTLm+vZkEQ=
golang.org/x/text v0.22.0 h1:bofq7m3/HAFvbF51jz3Q9wLg3jkvSPuiZu/pD1XwgtM=
golang.org/x/text v0.22.0/go.mod h1:YRoo4H8PVmsu+E3Ou7cqLVH8oXWIHVoX0jqUWALQhfY=
golang.org/x/text v0.24.0 h1:dd5Bzh4yt5KYA8f9CJHCP4FB4D51c2c6JvN37xJJkJ0=
golang.org/x/text v0.24.0/go.mod h1:L8rBsPeo2pSS+xqN0d5u2ikmjtmoJbDBT1b7nHvFCdU=
golang.org/x/time v0.5.0 h1:o7cqy6amK/52YcAKIPlM3a+Fpj35zvRj2TP+e1xFSfk=
golang.org/x/time v0.5.0/go.mod h1:3BpzKBy/shNhVucY/MWOyx10tF3SFh9QdLuxbVysPQM=
golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
@@ -333,11 +326,11 @@ golang.org/x/tools v0.1.12/go.mod h1:hNGJHUnrk76NpqgfD5Aqm5Crs+Hm0VOH/i9J2+nxYbc
golang.org/x/tools v0.6.0/go.mod h1:Xwgl3UAJ/d3gWutnCtw505GrjyAbvKui8lOU390QaIU=
golang.org/x/tools v0.13.0/go.mod h1:HvlwmtVNQAhOuCjW7xxvovg8wbNq7LwfXh/k7wXUl58=
golang.org/x/tools v0.21.1-0.20240508182429-e35e4ccd0d2d/go.mod h1:aiJjzUbINMkxbQROHiO6hDPo2LHcIPhhQsa9DLh0yGk=
golang.org/x/tools v0.30.0 h1:BgcpHewrV5AUp2G9MebG4XPFI1E2W41zU1SaqVA9vJY=
golang.org/x/tools v0.30.0/go.mod h1:c347cR/OJfw5TI+GfX7RUPNMdDRRbjvYTS0jPyvsVtY=
golang.org/x/tools v0.32.0 h1:Q7N1vhpkQv7ybVzLFtTjvQya2ewbwNDZzUgfXGqtMWU=
golang.org/x/tools v0.32.0/go.mod h1:ZxrU41P/wAbZD8EDa6dDCa6XfpkhJ7HFMjHJXfBDu8s=
golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
google.golang.org/protobuf v1.36.5 h1:tPhr+woSbjfYvY6/GPufUoYizxw1cF/yFoxJ2fmpwlM=
google.golang.org/protobuf v1.36.5/go.mod h1:9fA7Ob0pmnwhb644+1+CVWFRbNajQ6iRojtC/QF5bRE=
google.golang.org/protobuf v1.36.6 h1:z1NpPI8ku2WgiWnf+t9wTPsn6eP1L7ksHUlkfLvd9xY=
google.golang.org/protobuf v1.36.6/go.mod h1:jduwjTPXsFjZGTmRluh+L6NjiWu7pchiJ2/5YcXBHnY=
gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c h1:Hei/4ADfdWqJk1ZMxUNpqntNwaWcugrBjAiHlqqRiVk=
gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c/go.mod h1:JHkPIbrfpd72SG/EVd6muEfDQjcINNoR0C8j2r3qZ4Q=

75
gqlgen.yml
View File

@@ -51,61 +51,52 @@ models:
- github.com/99designs/gqlgen/graphql.Int64
- github.com/99designs/gqlgen/graphql.Int32
Job:
model: "github.com/ClusterCockpit/cc-backend/pkg/schema.Job"
model: "github.com/ClusterCockpit/cc-lib/schema.Job"
fields:
tags:
resolver: true
metaData:
resolver: true
Cluster:
model: "github.com/ClusterCockpit/cc-backend/pkg/schema.Cluster"
model: "github.com/ClusterCockpit/cc-lib/schema.Cluster"
fields:
partitions:
resolver: true
NullableFloat:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.Float" }
MetricScope:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.MetricScope" }
MetricValue:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.MetricValue" }
Node:
model: "github.com/ClusterCockpit/cc-lib/schema.Node"
fields:
metaData:
resolver: true
NullableFloat: { model: "github.com/ClusterCockpit/cc-lib/schema.Float" }
MetricScope: { model: "github.com/ClusterCockpit/cc-lib/schema.MetricScope" }
MetricValue: { model: "github.com/ClusterCockpit/cc-lib/schema.MetricValue" }
JobStatistics:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.JobStatistics" }
{ model: "github.com/ClusterCockpit/cc-lib/schema.JobStatistics" }
GlobalMetricListItem:
{
model: "github.com/ClusterCockpit/cc-backend/pkg/schema.GlobalMetricListItem",
}
{ model: "github.com/ClusterCockpit/cc-lib/schema.GlobalMetricListItem" }
ClusterSupport:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.ClusterSupport" }
Tag: { model: "github.com/ClusterCockpit/cc-backend/pkg/schema.Tag" }
Resource:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.Resource" }
JobState:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.JobState" }
TimeRange:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.TimeRange" }
IntRange:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.IntRange" }
JobMetric:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.JobMetric" }
Series: { model: "github.com/ClusterCockpit/cc-backend/pkg/schema.Series" }
{ model: "github.com/ClusterCockpit/cc-lib/schema.ClusterSupport" }
Tag: { model: "github.com/ClusterCockpit/cc-lib/schema.Tag" }
Resource: { model: "github.com/ClusterCockpit/cc-lib/schema.Resource" }
JobState: { model: "github.com/ClusterCockpit/cc-lib/schema.JobState" }
MonitoringState:
{ model: "github.com/ClusterCockpit/cc-lib/schema.NodeState" }
HealthState:
{ model: "github.com/ClusterCockpit/cc-lib/schema.MonitoringState" }
TimeRange: { model: "github.com/ClusterCockpit/cc-lib/schema.TimeRange" }
IntRange: { model: "github.com/ClusterCockpit/cc-lib/schema.IntRange" }
JobMetric: { model: "github.com/ClusterCockpit/cc-lib/schema.JobMetric" }
Series: { model: "github.com/ClusterCockpit/cc-lib/schema.Series" }
MetricStatistics:
{
model: "github.com/ClusterCockpit/cc-backend/pkg/schema.MetricStatistics",
}
{ model: "github.com/ClusterCockpit/cc-lib/schema.MetricStatistics" }
MetricConfig:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.MetricConfig" }
{ model: "github.com/ClusterCockpit/cc-lib/schema.MetricConfig" }
SubClusterConfig:
{
model: "github.com/ClusterCockpit/cc-backend/pkg/schema.SubClusterConfig",
}
Accelerator:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.Accelerator" }
Topology:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.Topology" }
{ model: "github.com/ClusterCockpit/cc-lib/schema.SubClusterConfig" }
Accelerator: { model: "github.com/ClusterCockpit/cc-lib/schema.Accelerator" }
Topology: { model: "github.com/ClusterCockpit/cc-lib/schema.Topology" }
FilterRanges:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.FilterRanges" }
SubCluster:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.SubCluster" }
StatsSeries:
{ model: "github.com/ClusterCockpit/cc-backend/pkg/schema.StatsSeries" }
Unit: { model: "github.com/ClusterCockpit/cc-backend/pkg/schema.Unit" }
{ model: "github.com/ClusterCockpit/cc-lib/schema.FilterRanges" }
SubCluster: { model: "github.com/ClusterCockpit/cc-lib/schema.SubCluster" }
StatsSeries: { model: "github.com/ClusterCockpit/cc-lib/schema.StatsSeries" }
Unit: { model: "github.com/ClusterCockpit/cc-lib/schema.Unit" }

37
internal/api/api_test.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package api_test
@@ -27,8 +27,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/metricdata"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/gorilla/mux"
_ "github.com/mattn/go-sqlite3"
@@ -116,14 +116,14 @@ func setup(t *testing.T) *api.RestApi {
]
}`
log.Init("info", true)
cclog.Init("info", true)
tmpdir := t.TempDir()
jobarchive := filepath.Join(tmpdir, "job-archive")
if err := os.Mkdir(jobarchive, 0777); err != nil {
t.Fatal(err)
}
if err := os.WriteFile(filepath.Join(jobarchive, "version.txt"), []byte(fmt.Sprintf("%d", 2)), 0666); err != nil {
if err := os.WriteFile(filepath.Join(jobarchive, "version.txt"), fmt.Appendf(nil, "%d", 2), 0666); err != nil {
t.Fatal(err)
}
@@ -204,11 +204,11 @@ func TestRestApi(t *testing.T) {
restapi.MountApiRoutes(r)
var TestJobId int64 = 123
var TestClusterName string = "testcluster"
TestClusterName := "testcluster"
var TestStartTime int64 = 123456789
const startJobBody string = `{
"jobId": 123,
"jobId": 123,
"user": "testuser",
"project": "testproj",
"cluster": "testcluster",
@@ -221,7 +221,6 @@ func TestRestApi(t *testing.T) {
"exclusive": 1,
"monitoringStatus": 1,
"smt": 1,
"tags": [{ "type": "testTagType", "name": "testTagName", "scope": "testuser" }],
"resources": [
{
"hostname": "host123",
@@ -252,16 +251,17 @@ func TestRestApi(t *testing.T) {
if response.StatusCode != http.StatusCreated {
t.Fatal(response.Status, recorder.Body.String())
}
resolver := graph.GetResolverInstance()
// resolver := graph.GetResolverInstance()
restapi.JobRepository.SyncJobs()
job, err := restapi.JobRepository.Find(&TestJobId, &TestClusterName, &TestStartTime)
if err != nil {
t.Fatal(err)
}
job.Tags, err = resolver.Job().Tags(ctx, job)
if err != nil {
t.Fatal(err)
}
// job.Tags, err = resolver.Job().Tags(ctx, job)
// if err != nil {
// t.Fatal(err)
// }
if job.JobID != 123 ||
job.User != "testuser" ||
@@ -278,13 +278,13 @@ func TestRestApi(t *testing.T) {
job.MonitoringStatus != 1 ||
job.SMT != 1 ||
!reflect.DeepEqual(job.Resources, []*schema.Resource{{Hostname: "host123", HWThreads: []int{0, 1, 2, 3, 4, 5, 6, 7}}}) ||
job.StartTime.Unix() != 123456789 {
job.StartTime != 123456789 {
t.Fatalf("unexpected job properties: %#v", job)
}
if len(job.Tags) != 1 || job.Tags[0].Type != "testTagType" || job.Tags[0].Name != "testTagName" || job.Tags[0].Scope != "testuser" {
t.Fatalf("unexpected tags: %#v", job.Tags)
}
// if len(job.Tags) != 1 || job.Tags[0].Type != "testTagType" || job.Tags[0].Name != "testTagName" || job.Tags[0].Scope != "testuser" {
// t.Fatalf("unexpected tags: %#v", job.Tags)
// }
}); !ok {
return
}
@@ -352,7 +352,7 @@ func TestRestApi(t *testing.T) {
t.Run("CheckDoubleStart", func(t *testing.T) {
// Starting a job with the same jobId and cluster should only be allowed if the startTime is far appart!
body := strings.Replace(startJobBody, `"startTime": 123456789`, `"startTime": 123456790`, -1)
body := strings.ReplaceAll(startJobBody, `"startTime": 123456789`, `"startTime": 123456790`)
req := httptest.NewRequest(http.MethodPost, "/jobs/start_job/", bytes.NewBuffer([]byte(body)))
recorder := httptest.NewRecorder()
@@ -402,6 +402,7 @@ func TestRestApi(t *testing.T) {
}
time.Sleep(1 * time.Second)
restapi.JobRepository.SyncJobs()
const stopJobBodyFailed string = `{
"jobId": 12345,

70
internal/api/cluster.go Normal file
View File

@@ -0,0 +1,70 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package api
import (
"bufio"
"encoding/json"
"fmt"
"net/http"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-lib/schema"
)
// GetClustersApiResponse model
type GetClustersApiResponse struct {
Clusters []*schema.Cluster `json:"clusters"` // Array of clusters
}
// getClusters godoc
// @summary Lists all cluster configs
// @tags Cluster query
// @description Get a list of all cluster configs. Specific cluster can be requested using query parameter.
// @produce json
// @param cluster query string false "Job Cluster"
// @success 200 {object} api.GetClustersApiResponse "Array of clusters"
// @failure 400 {object} api.ErrorResponse "Bad Request"
// @failure 401 {object} api.ErrorResponse "Unauthorized"
// @failure 403 {object} api.ErrorResponse "Forbidden"
// @failure 500 {object} api.ErrorResponse "Internal Server Error"
// @security ApiKeyAuth
// @router /api/clusters/ [get]
func (api *RestApi) getClusters(rw http.ResponseWriter, r *http.Request) {
if user := repository.GetUserFromContext(r.Context()); user != nil &&
!user.HasRole(schema.RoleApi) {
handleError(fmt.Errorf("missing role: %v", schema.GetRoleString(schema.RoleApi)), http.StatusForbidden, rw)
return
}
rw.Header().Add("Content-Type", "application/json")
bw := bufio.NewWriter(rw)
defer bw.Flush()
var clusters []*schema.Cluster
if r.URL.Query().Has("cluster") {
name := r.URL.Query().Get("cluster")
cluster := archive.GetCluster(name)
if cluster == nil {
handleError(fmt.Errorf("unknown cluster: %s", name), http.StatusBadRequest, rw)
return
}
clusters = append(clusters, cluster)
} else {
clusters = archive.Clusters
}
payload := GetClustersApiResponse{
Clusters: clusters,
}
if err := json.NewEncoder(bw).Encode(payload); err != nil {
handleError(err, http.StatusInternalServerError, rw)
return
}
}

273
internal/api/docs.go
View File

@@ -208,7 +208,7 @@ const docTemplate = `{
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/api.DefaultJobApiResponse"
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
@@ -278,7 +278,7 @@ const docTemplate = `{
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/api.DefaultJobApiResponse"
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
@@ -348,7 +348,7 @@ const docTemplate = `{
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/api.DefaultJobApiResponse"
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
@@ -530,7 +530,7 @@ const docTemplate = `{
"ApiKeyAuth": []
}
],
"description": "Job specified in request body will be saved to database as \"running\" with new DB ID.\nJob specifications follow the 'JobMeta' scheme, API will fail to execute if requirements are not met.",
"description": "Job specified in request body will be saved to database as \"running\" with new DB ID.\nJob specifications follow the 'Job' scheme, API will fail to execute if requirements are not met.",
"consumes": [
"application/json"
],
@@ -548,7 +548,7 @@ const docTemplate = `{
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/schema.JobMeta"
"$ref": "#/definitions/schema.Job"
}
}
],
@@ -556,7 +556,7 @@ const docTemplate = `{
"201": {
"description": "Job added successfully",
"schema": {
"$ref": "#/definitions/api.DefaultJobApiResponse"
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
@@ -599,7 +599,7 @@ const docTemplate = `{
"ApiKeyAuth": []
}
],
"description": "Job to stop is specified by request body. All fields are required in this case.\nReturns full job resource information according to 'JobMeta' scheme.",
"description": "Job to stop is specified by request body. All fields are required in this case.\nReturns full job resource information according to 'Job' scheme.",
"produces": [
"application/json"
],
@@ -622,7 +622,7 @@ const docTemplate = `{
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/schema.JobMeta"
"$ref": "#/definitions/schema.Job"
}
},
"400": {
@@ -744,7 +744,7 @@ const docTemplate = `{
"ApiKeyAuth": []
}
],
"description": "Job to get is specified by database ID\nReturns full job resource information according to 'JobMeta' scheme and all metrics according to 'JobData'.",
"description": "Job to get is specified by database ID\nReturns full job resource information according to 'Job' scheme and all metrics according to 'JobData'.",
"produces": [
"application/json"
],
@@ -818,7 +818,7 @@ const docTemplate = `{
"ApiKeyAuth": []
}
],
"description": "Job to get is specified by database ID\nReturns full job resource information according to 'JobMeta' scheme and all metrics according to 'JobData'.",
"description": "Job to get is specified by database ID\nReturns full job resource information according to 'Job' scheme and all metrics according to 'JobData'.",
"consumes": [
"application/json"
],
@@ -896,6 +896,66 @@ const docTemplate = `{
}
}
},
"/api/nodestats/": {
"post": {
"security": [
{
"ApiKeyAuth": []
}
],
"description": "Returns a JSON-encoded list of users.\nRequired query-parameter defines if all users or only users with additional special roles are returned.",
"produces": [
"application/json"
],
"tags": [
"Nodestates"
],
"summary": "Deliver updated Slurm node states",
"parameters": [
{
"description": "Request body containing nodes and their states",
"name": "request",
"in": "body",
"required": true,
"schema": {
"$ref": "#/definitions/api.UpdateNodeStatesRequest"
}
}
],
"responses": {
"200": {
"description": "Success message",
"schema": {
"$ref": "#/definitions/api.DefaultApiResponse"
}
},
"400": {
"description": "Bad Request",
"schema": {
"$ref": "#/definitions/api.ErrorResponse"
}
},
"401": {
"description": "Unauthorized",
"schema": {
"$ref": "#/definitions/api.ErrorResponse"
}
},
"403": {
"description": "Forbidden",
"schema": {
"$ref": "#/definitions/api.ErrorResponse"
}
},
"500": {
"description": "Internal Server Error",
"schema": {
"$ref": "#/definitions/api.ErrorResponse"
}
}
}
}
},
"/api/users/": {
"get": {
"security": [
@@ -1144,7 +1204,7 @@ const docTemplate = `{
}
}
},
"api.DefaultJobApiResponse": {
"api.DefaultApiResponse": {
"type": "object",
"properties": {
"msg": {
@@ -1238,7 +1298,7 @@ const docTemplate = `{
"description": "Array of jobs",
"type": "array",
"items": {
"$ref": "#/definitions/schema.JobMeta"
"$ref": "#/definitions/schema.Job"
}
},
"page": {
@@ -1292,6 +1352,20 @@ const docTemplate = `{
}
}
},
"api.Node": {
"type": "object",
"properties": {
"hostname": {
"type": "string"
},
"states": {
"type": "array",
"items": {
"type": "string"
}
}
}
},
"api.StopJobApiRequest": {
"type": "object",
"required": [
@@ -1325,6 +1399,21 @@ const docTemplate = `{
}
}
},
"api.UpdateNodeStatesRequest": {
"type": "object",
"properties": {
"cluster": {
"type": "string",
"example": "fritz"
},
"nodes": {
"type": "array",
"items": {
"$ref": "#/definitions/api.Node"
}
}
}
},
"schema.Accelerator": {
"type": "object",
"properties": {
@@ -1360,7 +1449,6 @@ const docTemplate = `{
}
},
"schema.Job": {
"description": "Information of a HPC job.",
"type": "object",
"properties": {
"arrayJobId": {
@@ -1458,6 +1546,12 @@ const docTemplate = `{
"type": "string",
"example": "abcd200"
},
"requestedMemory": {
"description": "in MB",
"type": "integer",
"minimum": 1,
"example": 128000
},
"resources": {
"type": "array",
"items": {
@@ -1469,7 +1563,14 @@ const docTemplate = `{
"example": 4
},
"startTime": {
"type": "string"
"type": "integer",
"example": 1649723812
},
"statistics": {
"type": "object",
"additionalProperties": {
"$ref": "#/definitions/schema.JobStatistics"
}
},
"subCluster": {
"type": "string",
@@ -1517,147 +1618,6 @@ const docTemplate = `{
}
}
},
"schema.JobMeta": {
"description": "Meta data information of a HPC job.",
"type": "object",
"properties": {
"arrayJobId": {
"type": "integer",
"example": 123000
},
"cluster": {
"type": "string",
"example": "fritz"
},
"concurrentJobs": {
"$ref": "#/definitions/schema.JobLinkResultList"
},
"duration": {
"type": "integer",
"minimum": 1,
"example": 43200
},
"energy": {
"type": "number"
},
"energyFootprint": {
"type": "object",
"additionalProperties": {
"type": "number"
}
},
"exclusive": {
"type": "integer",
"maximum": 2,
"minimum": 0,
"example": 1
},
"footprint": {
"type": "object",
"additionalProperties": {
"type": "number"
}
},
"id": {
"type": "integer"
},
"jobId": {
"type": "integer",
"example": 123000
},
"jobState": {
"enum": [
"completed",
"failed",
"cancelled",
"stopped",
"timeout",
"out_of_memory"
],
"allOf": [
{
"$ref": "#/definitions/schema.JobState"
}
],
"example": "completed"
},
"metaData": {
"type": "object",
"additionalProperties": {
"type": "string"
}
},
"monitoringStatus": {
"type": "integer",
"maximum": 3,
"minimum": 0,
"example": 1
},
"numAcc": {
"type": "integer",
"minimum": 1,
"example": 2
},
"numHwthreads": {
"type": "integer",
"minimum": 1,
"example": 20
},
"numNodes": {
"type": "integer",
"minimum": 1,
"example": 2
},
"partition": {
"type": "string",
"example": "main"
},
"project": {
"type": "string",
"example": "abcd200"
},
"resources": {
"type": "array",
"items": {
"$ref": "#/definitions/schema.Resource"
}
},
"smt": {
"type": "integer",
"example": 4
},
"startTime": {
"type": "integer",
"minimum": 1,
"example": 1649723812
},
"statistics": {
"type": "object",
"additionalProperties": {
"$ref": "#/definitions/schema.JobStatistics"
}
},
"subCluster": {
"type": "string",
"example": "main"
},
"tags": {
"type": "array",
"items": {
"$ref": "#/definitions/schema.Tag"
}
},
"user": {
"type": "string",
"example": "abcd100h"
},
"walltime": {
"type": "integer",
"minimum": 1,
"example": 86400
}
}
},
"schema.JobMetric": {
"type": "object",
"properties": {
@@ -1985,6 +1945,9 @@ const docTemplate = `{
},
"remove": {
"type": "boolean"
},
"unit": {
"$ref": "#/definitions/schema.Unit"
}
}
},

1038
internal/api/job.go Normal file
View File

File diff suppressed because it is too large Load Diff

80
internal/api/node.go Normal file
View File

@@ -0,0 +1,80 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package api
import (
"fmt"
"net/http"
"strings"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-lib/schema"
)
type Node struct {
Name string `json:"hostname"`
States []string `json:"states"`
CpusAllocated int `json:"cpusAllocated"`
CpusTotal int `json:"cpusTotal"`
MemoryAllocated int `json:"memoryAllocated"`
MemoryTotal int `json:"memoryTotal"`
GpusAllocated int `json:"gpusAllocated"`
GpusTotal int `json:"gpusTotal"`
}
// updateNodeStatesRequest model
type UpdateNodeStatesRequest struct {
Nodes []Node `json:"nodes"`
Cluster string `json:"cluster" example:"fritz"`
}
// this routine assumes that only one of them exists per node
func determineState(states []string) schema.NodeState {
for _, state := range states {
switch strings.ToLower(state) {
case "allocated":
return schema.NodeStateAllocated
case "reserved":
return schema.NodeStateReserved
case "idle":
return schema.NodeStateIdle
case "down":
return schema.NodeStateDown
case "mixed":
return schema.NodeStateMixed
}
}
return schema.NodeStateUnknown
}
// updateNodeStates godoc
// @summary Deliver updated Slurm node states
// @tags Nodestates
// @description Returns a JSON-encoded list of users.
// @description Required query-parameter defines if all users or only users with additional special roles are returned.
// @produce json
// @param request body UpdateNodeStatesRequest true "Request body containing nodes and their states"
// @success 200 {object} api.DefaultApiResponse "Success message"
// @failure 400 {object} api.ErrorResponse "Bad Request"
// @failure 401 {object} api.ErrorResponse "Unauthorized"
// @failure 403 {object} api.ErrorResponse "Forbidden"
// @failure 500 {object} api.ErrorResponse "Internal Server Error"
// @security ApiKeyAuth
// @router /api/nodestats/ [post]
func (api *RestApi) updateNodeStates(rw http.ResponseWriter, r *http.Request) {
// Parse request body
req := UpdateNodeStatesRequest{}
if err := decode(r.Body, &req); err != nil {
handleError(fmt.Errorf("parsing request body failed: %w", err), http.StatusBadRequest, rw)
return
}
repo := repository.GetNodeRepository()
for _, node := range req.Nodes {
state := determineState(node.States)
repo.UpdateNodeState(node.Name, req.Cluster, &state)
}
}

1238
internal/api/rest.go
View File

File diff suppressed because it is too large Load Diff

159
internal/api/user.go Normal file
View File

@@ -0,0 +1,159 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package api
import (
"encoding/json"
"fmt"
"net/http"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/gorilla/mux"
)
type ApiReturnedUser struct {
Username string `json:"username"`
Name string `json:"name"`
Roles []string `json:"roles"`
Email string `json:"email"`
Projects []string `json:"projects"`
}
// getUsers godoc
// @summary Returns a list of users
// @tags User
// @description Returns a JSON-encoded list of users.
// @description Required query-parameter defines if all users or only users with additional special roles are returned.
// @produce json
// @param not-just-user query bool true "If returned list should contain all users or only users with additional special roles"
// @success 200 {array} api.ApiReturnedUser "List of users returned successfully"
// @failure 400 {string} string "Bad Request"
// @failure 401 {string} string "Unauthorized"
// @failure 403 {string} string "Forbidden"
// @failure 500 {string} string "Internal Server Error"
// @security ApiKeyAuth
// @router /api/users/ [get]
func (api *RestApi) getUsers(rw http.ResponseWriter, r *http.Request) {
// SecuredCheck() only worked with TokenAuth: Removed
if user := repository.GetUserFromContext(r.Context()); !user.HasRole(schema.RoleAdmin) {
http.Error(rw, "Only admins are allowed to fetch a list of users", http.StatusForbidden)
return
}
users, err := repository.GetUserRepository().ListUsers(r.URL.Query().Get("not-just-user") == "true")
if err != nil {
http.Error(rw, err.Error(), http.StatusInternalServerError)
return
}
json.NewEncoder(rw).Encode(users)
}
func (api *RestApi) updateUser(rw http.ResponseWriter, r *http.Request) {
// SecuredCheck() only worked with TokenAuth: Removed
if user := repository.GetUserFromContext(r.Context()); !user.HasRole(schema.RoleAdmin) {
http.Error(rw, "Only admins are allowed to update a user", http.StatusForbidden)
return
}
// Get Values
newrole := r.FormValue("add-role")
delrole := r.FormValue("remove-role")
newproj := r.FormValue("add-project")
delproj := r.FormValue("remove-project")
// TODO: Handle anything but roles...
if newrole != "" {
if err := repository.GetUserRepository().AddRole(r.Context(), mux.Vars(r)["id"], newrole); err != nil {
http.Error(rw, err.Error(), http.StatusUnprocessableEntity)
return
}
rw.Write([]byte("Add Role Success"))
} else if delrole != "" {
if err := repository.GetUserRepository().RemoveRole(r.Context(), mux.Vars(r)["id"], delrole); err != nil {
http.Error(rw, err.Error(), http.StatusUnprocessableEntity)
return
}
rw.Write([]byte("Remove Role Success"))
} else if newproj != "" {
if err := repository.GetUserRepository().AddProject(r.Context(), mux.Vars(r)["id"], newproj); err != nil {
http.Error(rw, err.Error(), http.StatusUnprocessableEntity)
return
}
rw.Write([]byte("Add Project Success"))
} else if delproj != "" {
if err := repository.GetUserRepository().RemoveProject(r.Context(), mux.Vars(r)["id"], delproj); err != nil {
http.Error(rw, err.Error(), http.StatusUnprocessableEntity)
return
}
rw.Write([]byte("Remove Project Success"))
} else {
http.Error(rw, "Not Add or Del [role|project]?", http.StatusInternalServerError)
}
}
func (api *RestApi) createUser(rw http.ResponseWriter, r *http.Request) {
// SecuredCheck() only worked with TokenAuth: Removed
rw.Header().Set("Content-Type", "text/plain")
me := repository.GetUserFromContext(r.Context())
if !me.HasRole(schema.RoleAdmin) {
http.Error(rw, "Only admins are allowed to create new users", http.StatusForbidden)
return
}
username, password, role, name, email, project := r.FormValue("username"),
r.FormValue("password"), r.FormValue("role"), r.FormValue("name"),
r.FormValue("email"), r.FormValue("project")
if len(password) == 0 && role != schema.GetRoleString(schema.RoleApi) {
http.Error(rw, "Only API users are allowed to have a blank password (login will be impossible)", http.StatusBadRequest)
return
}
if len(project) != 0 && role != schema.GetRoleString(schema.RoleManager) {
http.Error(rw, "only managers require a project (can be changed later)",
http.StatusBadRequest)
return
} else if len(project) == 0 && role == schema.GetRoleString(schema.RoleManager) {
http.Error(rw, "managers require a project to manage (can be changed later)",
http.StatusBadRequest)
return
}
if err := repository.GetUserRepository().AddUser(&schema.User{
Username: username,
Name: name,
Password: password,
Email: email,
Projects: []string{project},
Roles: []string{role},
}); err != nil {
http.Error(rw, err.Error(), http.StatusUnprocessableEntity)
return
}
fmt.Fprintf(rw, "User %v successfully created!\n", username)
}
func (api *RestApi) deleteUser(rw http.ResponseWriter, r *http.Request) {
// SecuredCheck() only worked with TokenAuth: Removed
if user := repository.GetUserFromContext(r.Context()); !user.HasRole(schema.RoleAdmin) {
http.Error(rw, "Only admins are allowed to delete a user", http.StatusForbidden)
return
}
username := r.FormValue("username")
if err := repository.GetUserRepository().DelUser(username); err != nil {
http.Error(rw, err.Error(), http.StatusUnprocessableEntity)
return
}
rw.WriteHeader(http.StatusOK)
}

30
internal/archiver/archiveWorker.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package archiver
@@ -10,8 +10,8 @@ import (
"time"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
sq "github.com/Masterminds/squirrel"
)
@@ -40,8 +40,8 @@ func archivingWorker() {
// not using meta data, called to load JobMeta into Cache?
// will fail if job meta not in repository
if _, err := jobRepo.FetchMetadata(job); err != nil {
log.Errorf("archiving job (dbid: %d) failed at check metadata step: %s", job.ID, err.Error())
jobRepo.UpdateMonitoringStatus(job.ID, schema.MonitoringStatusArchivingFailed)
cclog.Errorf("archiving job (dbid: %d) failed at check metadata step: %s", job.ID, err.Error())
jobRepo.UpdateMonitoringStatus(*job.ID, schema.MonitoringStatusArchivingFailed)
continue
}
@@ -49,30 +49,34 @@ func archivingWorker() {
// TODO: Maybe use context with cancel/timeout here
jobMeta, err := ArchiveJob(job, context.Background())
if err != nil {
log.Errorf("archiving job (dbid: %d) failed at archiving job step: %s", job.ID, err.Error())
jobRepo.UpdateMonitoringStatus(job.ID, schema.MonitoringStatusArchivingFailed)
cclog.Errorf("archiving job (dbid: %d) failed at archiving job step: %s", job.ID, err.Error())
jobRepo.UpdateMonitoringStatus(*job.ID, schema.MonitoringStatusArchivingFailed)
continue
}
stmt := sq.Update("job").Where("job.id = ?", job.ID)
if stmt, err = jobRepo.UpdateFootprint(stmt, jobMeta); err != nil {
log.Errorf("archiving job (dbid: %d) failed at update Footprint step: %s", job.ID, err.Error())
cclog.Errorf("archiving job (dbid: %d) failed at update Footprint step: %s", job.ID, err.Error())
continue
}
if stmt, err = jobRepo.UpdateEnergy(stmt, jobMeta); err != nil {
log.Errorf("archiving job (dbid: %d) failed at update Energy step: %s", job.ID, err.Error())
cclog.Errorf("archiving job (dbid: %d) failed at update Energy step: %s", job.ID, err.Error())
continue
}
// Update the jobs database entry one last time:
stmt = jobRepo.MarkArchived(stmt, schema.MonitoringStatusArchivingSuccessful)
if err := jobRepo.Execute(stmt); err != nil {
log.Errorf("archiving job (dbid: %d) failed at db execute: %s", job.ID, err.Error())
cclog.Errorf("archiving job (dbid: %d) failed at db execute: %s", job.ID, err.Error())
continue
}
log.Debugf("archiving job %d took %s", job.JobID, time.Since(start))
log.Printf("archiving job (dbid: %d) successful", job.ID)
cclog.Debugf("archiving job %d took %s", job.JobID, time.Since(start))
cclog.Printf("archiving job (dbid: %d) successful", job.ID)
repository.CallJobStopHooks(job)
archivePending.Done()
default:
continue
}
}
}
@@ -80,7 +84,7 @@ func archivingWorker() {
// Trigger async archiving
func TriggerArchiving(job *schema.Job) {
if archiveChannel == nil {
log.Fatal("Cannot archive without archiving channel. Did you Start the archiver?")
cclog.Fatal("Cannot archive without archiving channel. Did you Start the archiver?")
}
archivePending.Add(1)

22
internal/archiver/archiver.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package archiver
@@ -11,12 +11,12 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/metricDataDispatcher"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
// Writes a running job to the job-archive
func ArchiveJob(job *schema.Job, ctx context.Context) (*schema.JobMeta, error) {
func ArchiveJob(job *schema.Job, ctx context.Context) (*schema.Job, error) {
allMetrics := make([]string, 0)
metricConfigs := archive.GetCluster(job.Cluster).MetricConfig
for _, mc := range metricConfigs {
@@ -36,15 +36,11 @@ func ArchiveJob(job *schema.Job, ctx context.Context) (*schema.JobMeta, error) {
jobData, err := metricDataDispatcher.LoadData(job, allMetrics, scopes, ctx, 0) // 0 Resulotion-Value retrieves highest res (60s)
if err != nil {
log.Error("Error wile loading job data for archiving")
cclog.Error("Error wile loading job data for archiving")
return nil, err
}
jobMeta := &schema.JobMeta{
BaseJob: job.BaseJob,
StartTime: job.StartTime.Unix(),
Statistics: make(map[string]schema.JobStatistics),
}
job.Statistics = make(map[string]schema.JobStatistics)
for metric, data := range jobData {
avg, min, max := 0.0, math.MaxFloat32, -math.MaxFloat32
@@ -61,7 +57,7 @@ func ArchiveJob(job *schema.Job, ctx context.Context) (*schema.JobMeta, error) {
}
// Round AVG Result to 2 Digits
jobMeta.Statistics[metric] = schema.JobStatistics{
job.Statistics[metric] = schema.JobStatistics{
Unit: schema.Unit{
Prefix: archive.GetMetricConfig(job.Cluster, metric).Unit.Prefix,
Base: archive.GetMetricConfig(job.Cluster, metric).Unit.Base,
@@ -76,8 +72,8 @@ func ArchiveJob(job *schema.Job, ctx context.Context) (*schema.JobMeta, error) {
// only return the JobMeta structure as the
// statistics in there are needed.
if config.Keys.DisableArchive {
return jobMeta, nil
return job, nil
}
return jobMeta, archive.GetHandle().ImportJob(jobMeta, &jobData)
return job, archive.GetHandle().ImportJob(job, &jobData)
}

90
internal/auth/auth.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package auth
@@ -22,9 +22,9 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/internal/util"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/ClusterCockpit/cc-lib/util"
"github.com/gorilla/sessions"
)
@@ -66,7 +66,7 @@ func (auth *Authentication) AuthViaSession(
) (*schema.User, error) {
session, err := auth.sessionStore.Get(r, "session")
if err != nil {
log.Error("Error while getting session store")
cclog.Error("Error while getting session store")
return nil, err
}
@@ -93,16 +93,16 @@ func Init() {
sessKey := os.Getenv("SESSION_KEY")
if sessKey == "" {
log.Warn("environment variable 'SESSION_KEY' not set (will use non-persistent random key)")
cclog.Warn("environment variable 'SESSION_KEY' not set (will use non-persistent random key)")
bytes := make([]byte, 32)
if _, err := rand.Read(bytes); err != nil {
log.Fatal("Error while initializing authentication -> failed to generate random bytes for session key")
cclog.Fatal("Error while initializing authentication -> failed to generate random bytes for session key")
}
authInstance.sessionStore = sessions.NewCookieStore(bytes)
} else {
bytes, err := base64.StdEncoding.DecodeString(sessKey)
if err != nil {
log.Fatal("Error while initializing authentication -> decoding session key failed")
cclog.Fatal("Error while initializing authentication -> decoding session key failed")
}
authInstance.sessionStore = sessions.NewCookieStore(bytes)
}
@@ -114,41 +114,41 @@ func Init() {
if config.Keys.LdapConfig != nil {
ldapAuth := &LdapAuthenticator{}
if err := ldapAuth.Init(); err != nil {
log.Warn("Error while initializing authentication -> ldapAuth init failed")
cclog.Warn("Error while initializing authentication -> ldapAuth init failed")
} else {
authInstance.LdapAuth = ldapAuth
authInstance.authenticators = append(authInstance.authenticators, authInstance.LdapAuth)
}
} else {
log.Info("Missing LDAP configuration: No LDAP support!")
cclog.Info("Missing LDAP configuration: No LDAP support!")
}
if config.Keys.JwtConfig != nil {
authInstance.JwtAuth = &JWTAuthenticator{}
if err := authInstance.JwtAuth.Init(); err != nil {
log.Fatal("Error while initializing authentication -> jwtAuth init failed")
cclog.Fatal("Error while initializing authentication -> jwtAuth init failed")
}
jwtSessionAuth := &JWTSessionAuthenticator{}
if err := jwtSessionAuth.Init(); err != nil {
log.Info("jwtSessionAuth init failed: No JWT login support!")
cclog.Info("jwtSessionAuth init failed: No JWT login support!")
} else {
authInstance.authenticators = append(authInstance.authenticators, jwtSessionAuth)
}
jwtCookieSessionAuth := &JWTCookieSessionAuthenticator{}
if err := jwtCookieSessionAuth.Init(); err != nil {
log.Info("jwtCookieSessionAuth init failed: No JWT cookie login support!")
cclog.Info("jwtCookieSessionAuth init failed: No JWT cookie login support!")
} else {
authInstance.authenticators = append(authInstance.authenticators, jwtCookieSessionAuth)
}
} else {
log.Info("Missing JWT configuration: No JWT token support!")
cclog.Info("Missing JWT configuration: No JWT token support!")
}
authInstance.LocalAuth = &LocalAuthenticator{}
if err := authInstance.LocalAuth.Init(); err != nil {
log.Fatal("Error while initializing authentication -> localAuth init failed")
cclog.Fatal("Error while initializing authentication -> localAuth init failed")
}
authInstance.authenticators = append(authInstance.authenticators, authInstance.LocalAuth)
})
@@ -156,7 +156,7 @@ func Init() {
func GetAuthInstance() *Authentication {
if authInstance == nil {
log.Fatal("Authentication module not initialized!")
cclog.Fatal("Authentication module not initialized!")
}
return authInstance
@@ -167,14 +167,14 @@ func handleTokenUser(tokenUser *schema.User) {
dbUser, err := r.GetUser(tokenUser.Username)
if err != nil && err != sql.ErrNoRows {
log.Errorf("Error while loading user '%s': %v", tokenUser.Username, err)
cclog.Errorf("Error while loading user '%s': %v", tokenUser.Username, err)
} else if err == sql.ErrNoRows && config.Keys.JwtConfig.SyncUserOnLogin { // Adds New User
if err := r.AddUser(tokenUser); err != nil {
log.Errorf("Error while adding user '%s' to DB: %v", tokenUser.Username, err)
cclog.Errorf("Error while adding user '%s' to DB: %v", tokenUser.Username, err)
}
} else if err == nil && config.Keys.JwtConfig.UpdateUserOnLogin { // Update Existing User
if err := r.UpdateUser(dbUser, tokenUser); err != nil {
log.Errorf("Error while updating user '%s' to DB: %v", dbUser.Username, err)
cclog.Errorf("Error while updating user '%s' to DB: %v", dbUser.Username, err)
}
}
}
@@ -184,14 +184,14 @@ func handleOIDCUser(OIDCUser *schema.User) {
dbUser, err := r.GetUser(OIDCUser.Username)
if err != nil && err != sql.ErrNoRows {
log.Errorf("Error while loading user '%s': %v", OIDCUser.Username, err)
cclog.Errorf("Error while loading user '%s': %v", OIDCUser.Username, err)
} else if err == sql.ErrNoRows && config.Keys.OpenIDConfig.SyncUserOnLogin { // Adds New User
if err := r.AddUser(OIDCUser); err != nil {
log.Errorf("Error while adding user '%s' to DB: %v", OIDCUser.Username, err)
cclog.Errorf("Error while adding user '%s' to DB: %v", OIDCUser.Username, err)
}
} else if err == nil && config.Keys.OpenIDConfig.UpdateUserOnLogin { // Update Existing User
if err := r.UpdateUser(dbUser, OIDCUser); err != nil {
log.Errorf("Error while updating user '%s' to DB: %v", dbUser.Username, err)
cclog.Errorf("Error while updating user '%s' to DB: %v", dbUser.Username, err)
}
}
}
@@ -199,7 +199,7 @@ func handleOIDCUser(OIDCUser *schema.User) {
func (auth *Authentication) SaveSession(rw http.ResponseWriter, r *http.Request, user *schema.User) error {
session, err := auth.sessionStore.New(r, "session")
if err != nil {
log.Errorf("session creation failed: %s", err.Error())
cclog.Errorf("session creation failed: %s", err.Error())
http.Error(rw, err.Error(), http.StatusInternalServerError)
return err
}
@@ -215,7 +215,7 @@ func (auth *Authentication) SaveSession(rw http.ResponseWriter, r *http.Request,
session.Values["projects"] = user.Projects
session.Values["roles"] = user.Roles
if err := auth.sessionStore.Save(r, rw, session); err != nil {
log.Warnf("session save failed: %s", err.Error())
cclog.Warnf("session save failed: %s", err.Error())
http.Error(rw, err.Error(), http.StatusInternalServerError)
return err
}
@@ -236,8 +236,8 @@ func (auth *Authentication) Login(
limiter := getIPUserLimiter(ip, username)
if !limiter.Allow() {
log.Warnf("AUTH/RATE > Too many login attempts for combination IP: %s, Username: %s", ip, username)
onfailure(rw, r, errors.New("Too many login attempts, try again in a few minutes."))
cclog.Warnf("AUTH/RATE > Too many login attempts for combination IP: %s, Username: %s", ip, username)
onfailure(rw, r, errors.New("too many login attempts, try again in a few minutes"))
return
}
@@ -246,7 +246,7 @@ func (auth *Authentication) Login(
var err error
dbUser, err = repository.GetUserRepository().GetUser(username)
if err != nil && err != sql.ErrNoRows {
log.Errorf("Error while loading user '%v'", username)
cclog.Errorf("Error while loading user '%v'", username)
}
}
@@ -256,12 +256,12 @@ func (auth *Authentication) Login(
if user, ok = authenticator.CanLogin(dbUser, username, rw, r); !ok {
continue
} else {
log.Debugf("Can login with user %v", user)
cclog.Debugf("Can login with user %v", user)
}
user, err := authenticator.Login(user, rw, r)
if err != nil {
log.Warnf("user login failed: %s", err.Error())
cclog.Warnf("user login failed: %s", err.Error())
onfailure(rw, r, err)
return
}
@@ -270,7 +270,7 @@ func (auth *Authentication) Login(
return
}
log.Infof("login successfull: user: %#v (roles: %v, projects: %v)", user.Username, user.Roles, user.Projects)
cclog.Infof("login successfull: user: %#v (roles: %v, projects: %v)", user.Username, user.Roles, user.Projects)
ctx := context.WithValue(r.Context(), repository.ContextUserKey, user)
if r.FormValue("redirect") != "" {
@@ -282,7 +282,7 @@ func (auth *Authentication) Login(
return
}
log.Debugf("login failed: no authenticator applied")
cclog.Debugf("login failed: no authenticator applied")
onfailure(rw, r, errors.New("no authenticator applied"))
})
}
@@ -294,14 +294,14 @@ func (auth *Authentication) Auth(
return http.HandlerFunc(func(rw http.ResponseWriter, r *http.Request) {
user, err := auth.JwtAuth.AuthViaJWT(rw, r)
if err != nil {
log.Infof("auth -> authentication failed: %s", err.Error())
cclog.Infof("auth -> authentication failed: %s", err.Error())
http.Error(rw, err.Error(), http.StatusUnauthorized)
return
}
if user == nil {
user, err = auth.AuthViaSession(rw, r)
if err != nil {
log.Infof("auth -> authentication failed: %s", err.Error())
cclog.Infof("auth -> authentication failed: %s", err.Error())
http.Error(rw, err.Error(), http.StatusUnauthorized)
return
}
@@ -312,7 +312,7 @@ func (auth *Authentication) Auth(
return
}
log.Info("auth -> authentication failed")
cclog.Info("auth -> authentication failed")
onfailure(rw, r, errors.New("unauthorized (please login first)"))
})
}
@@ -324,14 +324,14 @@ func (auth *Authentication) AuthApi(
return http.HandlerFunc(func(rw http.ResponseWriter, r *http.Request) {
user, err := auth.JwtAuth.AuthViaJWT(rw, r)
if err != nil {
log.Infof("auth api -> authentication failed: %s", err.Error())
cclog.Infof("auth api -> authentication failed: %s", err.Error())
onfailure(rw, r, err)
return
}
ipErr := securedCheck(user, r)
if ipErr != nil {
log.Infof("auth api -> secured check failed: %s", ipErr.Error())
cclog.Infof("auth api -> secured check failed: %s", ipErr.Error())
onfailure(rw, r, ipErr)
return
}
@@ -351,11 +351,11 @@ func (auth *Authentication) AuthApi(
return
}
default:
log.Info("auth api -> authentication failed: missing role")
cclog.Info("auth api -> authentication failed: missing role")
onfailure(rw, r, errors.New("unauthorized"))
}
}
log.Info("auth api -> authentication failed: no auth")
cclog.Info("auth api -> authentication failed: no auth")
onfailure(rw, r, errors.New("unauthorized"))
})
}
@@ -367,7 +367,7 @@ func (auth *Authentication) AuthUserApi(
return http.HandlerFunc(func(rw http.ResponseWriter, r *http.Request) {
user, err := auth.JwtAuth.AuthViaJWT(rw, r)
if err != nil {
log.Infof("auth user api -> authentication failed: %s", err.Error())
cclog.Infof("auth user api -> authentication failed: %s", err.Error())
onfailure(rw, r, err)
return
}
@@ -387,11 +387,11 @@ func (auth *Authentication) AuthUserApi(
return
}
default:
log.Info("auth user api -> authentication failed: missing role")
cclog.Info("auth user api -> authentication failed: missing role")
onfailure(rw, r, errors.New("unauthorized"))
}
}
log.Info("auth user api -> authentication failed: no auth")
cclog.Info("auth user api -> authentication failed: no auth")
onfailure(rw, r, errors.New("unauthorized"))
})
}
@@ -403,7 +403,7 @@ func (auth *Authentication) AuthConfigApi(
return http.HandlerFunc(func(rw http.ResponseWriter, r *http.Request) {
user, err := auth.AuthViaSession(rw, r)
if err != nil {
log.Infof("auth config api -> authentication failed: %s", err.Error())
cclog.Infof("auth config api -> authentication failed: %s", err.Error())
onfailure(rw, r, err)
return
}
@@ -412,7 +412,7 @@ func (auth *Authentication) AuthConfigApi(
onsuccess.ServeHTTP(rw, r.WithContext(ctx))
return
}
log.Info("auth config api -> authentication failed: no auth")
cclog.Info("auth config api -> authentication failed: no auth")
onfailure(rw, r, errors.New("unauthorized"))
})
}
@@ -424,7 +424,7 @@ func (auth *Authentication) AuthFrontendApi(
return http.HandlerFunc(func(rw http.ResponseWriter, r *http.Request) {
user, err := auth.AuthViaSession(rw, r)
if err != nil {
log.Infof("auth frontend api -> authentication failed: %s", err.Error())
cclog.Infof("auth frontend api -> authentication failed: %s", err.Error())
onfailure(rw, r, err)
return
}
@@ -433,7 +433,7 @@ func (auth *Authentication) AuthFrontendApi(
onsuccess.ServeHTTP(rw, r.WithContext(ctx))
return
}
log.Info("auth frontend api -> authentication failed: no auth")
cclog.Info("auth frontend api -> authentication failed: no auth")
onfailure(rw, r, errors.New("unauthorized"))
})
}

18
internal/auth/jwt.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package auth
@@ -15,8 +15,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/golang-jwt/jwt/v5"
)
@@ -28,17 +28,17 @@ type JWTAuthenticator struct {
func (ja *JWTAuthenticator) Init() error {
pubKey, privKey := os.Getenv("JWT_PUBLIC_KEY"), os.Getenv("JWT_PRIVATE_KEY")
if pubKey == "" || privKey == "" {
log.Warn("environment variables 'JWT_PUBLIC_KEY' or 'JWT_PRIVATE_KEY' not set (token based authentication will not work)")
cclog.Warn("environment variables 'JWT_PUBLIC_KEY' or 'JWT_PRIVATE_KEY' not set (token based authentication will not work)")
} else {
bytes, err := base64.StdEncoding.DecodeString(pubKey)
if err != nil {
log.Warn("Could not decode JWT public key")
cclog.Warn("Could not decode JWT public key")
return err
}
ja.publicKey = ed25519.PublicKey(bytes)
bytes, err = base64.StdEncoding.DecodeString(privKey)
if err != nil {
log.Warn("Could not decode JWT private key")
cclog.Warn("Could not decode JWT private key")
return err
}
ja.privateKey = ed25519.PrivateKey(bytes)
@@ -70,11 +70,11 @@ func (ja *JWTAuthenticator) AuthViaJWT(
return ja.publicKey, nil
})
if err != nil {
log.Warn("Error while parsing JWT token")
cclog.Warn("Error while parsing JWT token")
return nil, err
}
if !token.Valid {
log.Warn("jwt token claims are not valid")
cclog.Warn("jwt token claims are not valid")
return nil, errors.New("jwt token claims are not valid")
}
@@ -90,7 +90,7 @@ func (ja *JWTAuthenticator) AuthViaJWT(
user, err := ur.GetUser(sub)
// Deny any logins for unknown usernames
if err != nil {
log.Warn("Could not find user from JWT in internal database.")
cclog.Warn("Could not find user from JWT in internal database.")
return nil, errors.New("unknown user")
}
// Take user roles from database instead of trusting the JWT

34
internal/auth/jwtCookieSession.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package auth
@@ -15,8 +15,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/golang-jwt/jwt/v5"
)
@@ -31,18 +31,18 @@ var _ Authenticator = (*JWTCookieSessionAuthenticator)(nil)
func (ja *JWTCookieSessionAuthenticator) Init() error {
pubKey, privKey := os.Getenv("JWT_PUBLIC_KEY"), os.Getenv("JWT_PRIVATE_KEY")
if pubKey == "" || privKey == "" {
log.Warn("environment variables 'JWT_PUBLIC_KEY' or 'JWT_PRIVATE_KEY' not set (token based authentication will not work)")
cclog.Warn("environment variables 'JWT_PUBLIC_KEY' or 'JWT_PRIVATE_KEY' not set (token based authentication will not work)")
return errors.New("environment variables 'JWT_PUBLIC_KEY' or 'JWT_PRIVATE_KEY' not set (token based authentication will not work)")
} else {
bytes, err := base64.StdEncoding.DecodeString(pubKey)
if err != nil {
log.Warn("Could not decode JWT public key")
cclog.Warn("Could not decode JWT public key")
return err
}
ja.publicKey = ed25519.PublicKey(bytes)
bytes, err = base64.StdEncoding.DecodeString(privKey)
if err != nil {
log.Warn("Could not decode JWT private key")
cclog.Warn("Could not decode JWT private key")
return err
}
ja.privateKey = ed25519.PrivateKey(bytes)
@@ -53,13 +53,13 @@ func (ja *JWTCookieSessionAuthenticator) Init() error {
if keyFound && pubKeyCrossLogin != "" {
bytes, err := base64.StdEncoding.DecodeString(pubKeyCrossLogin)
if err != nil {
log.Warn("Could not decode cross login JWT public key")
cclog.Warn("Could not decode cross login JWT public key")
return err
}
ja.publicKeyCrossLogin = ed25519.PublicKey(bytes)
} else {
ja.publicKeyCrossLogin = nil
log.Debug("environment variable 'CROSS_LOGIN_JWT_PUBLIC_KEY' not set (cross login token based authentication will not work)")
cclog.Debug("environment variable 'CROSS_LOGIN_JWT_PUBLIC_KEY' not set (cross login token based authentication will not work)")
return errors.New("environment variable 'CROSS_LOGIN_JWT_PUBLIC_KEY' not set (cross login token based authentication will not work)")
}
@@ -67,22 +67,22 @@ func (ja *JWTCookieSessionAuthenticator) Init() error {
// Warn if other necessary settings are not configured
if jc != nil {
if jc.CookieName == "" {
log.Info("cookieName for JWTs not configured (cross login via JWT cookie will fail)")
cclog.Info("cookieName for JWTs not configured (cross login via JWT cookie will fail)")
return errors.New("cookieName for JWTs not configured (cross login via JWT cookie will fail)")
}
if !jc.ValidateUser {
log.Info("forceJWTValidationViaDatabase not set to true: CC will accept users and roles defined in JWTs regardless of its own database!")
cclog.Info("forceJWTValidationViaDatabase not set to true: CC will accept users and roles defined in JWTs regardless of its own database!")
}
if jc.TrustedIssuer == "" {
log.Info("trustedExternalIssuer for JWTs not configured (cross login via JWT cookie will fail)")
cclog.Info("trustedExternalIssuer for JWTs not configured (cross login via JWT cookie will fail)")
return errors.New("trustedExternalIssuer for JWTs not configured (cross login via JWT cookie will fail)")
}
} else {
log.Warn("config for JWTs not configured (cross login via JWT cookie will fail)")
cclog.Warn("config for JWTs not configured (cross login via JWT cookie will fail)")
return errors.New("config for JWTs not configured (cross login via JWT cookie will fail)")
}
log.Info("JWT Cookie Session authenticator successfully registered")
cclog.Info("JWT Cookie Session authenticator successfully registered")
return nil
}
@@ -140,12 +140,12 @@ func (ja *JWTCookieSessionAuthenticator) Login(
return ja.publicKey, nil
})
if err != nil {
log.Warn("JWT cookie session: error while parsing token")
cclog.Warn("JWT cookie session: error while parsing token")
return nil, err
}
if !token.Valid {
log.Warn("jwt token claims are not valid")
cclog.Warn("jwt token claims are not valid")
return nil, errors.New("jwt token claims are not valid")
}
@@ -159,12 +159,12 @@ func (ja *JWTCookieSessionAuthenticator) Login(
var err error
user, err = repository.GetUserRepository().GetUser(sub)
if err != nil && err != sql.ErrNoRows {
log.Errorf("Error while loading user '%v'", sub)
cclog.Errorf("Error while loading user '%v'", sub)
}
// Deny any logins for unknown usernames
if user == nil {
log.Warn("Could not find user from JWT in internal database.")
cclog.Warn("Could not find user from JWT in internal database.")
return nil, errors.New("unknown user")
}
} else {

18
internal/auth/jwtSession.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package auth
@@ -15,8 +15,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/golang-jwt/jwt/v5"
)
@@ -30,13 +30,13 @@ func (ja *JWTSessionAuthenticator) Init() error {
if pubKey := os.Getenv("CROSS_LOGIN_JWT_HS512_KEY"); pubKey != "" {
bytes, err := base64.StdEncoding.DecodeString(pubKey)
if err != nil {
log.Warn("Could not decode cross login JWT HS512 key")
cclog.Warn("Could not decode cross login JWT HS512 key")
return err
}
ja.loginTokenKey = bytes
}
log.Info("JWT Session authenticator successfully registered")
cclog.Info("JWT Session authenticator successfully registered")
return nil
}
@@ -67,12 +67,12 @@ func (ja *JWTSessionAuthenticator) Login(
return nil, fmt.Errorf("unkown signing method for login token: %s (known: HS256, HS512, EdDSA)", t.Method.Alg())
})
if err != nil {
log.Warn("Error while parsing jwt token")
cclog.Warn("Error while parsing jwt token")
return nil, err
}
if !token.Valid {
log.Warn("jwt token claims are not valid")
cclog.Warn("jwt token claims are not valid")
return nil, errors.New("jwt token claims are not valid")
}
@@ -86,12 +86,12 @@ func (ja *JWTSessionAuthenticator) Login(
var err error
user, err = repository.GetUserRepository().GetUser(sub)
if err != nil && err != sql.ErrNoRows {
log.Errorf("Error while loading user '%v'", sub)
cclog.Errorf("Error while loading user '%v'", sub)
}
// Deny any logins for unknown usernames
if user == nil {
log.Warn("Could not find user from JWT in internal database.")
cclog.Warn("Could not find user from JWT in internal database.")
return nil, errors.New("unknown user")
}
} else {

34
internal/auth/ldap.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package auth
@@ -13,8 +13,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/go-ldap/ldap/v3"
)
@@ -28,7 +28,7 @@ var _ Authenticator = (*LdapAuthenticator)(nil)
func (la *LdapAuthenticator) Init() error {
la.syncPassword = os.Getenv("LDAP_ADMIN_PASSWORD")
if la.syncPassword == "" {
log.Warn("environment variable 'LDAP_ADMIN_PASSWORD' not set (ldap sync will not work)")
cclog.Warn("environment variable 'LDAP_ADMIN_PASSWORD' not set (ldap sync will not work)")
}
lc := config.Keys.LdapConfig
@@ -58,7 +58,7 @@ func (la *LdapAuthenticator) CanLogin(
if lc.SyncUserOnLogin {
l, err := la.getLdapConnection(true)
if err != nil {
log.Error("LDAP connection error")
cclog.Error("LDAP connection error")
}
defer l.Close()
@@ -71,12 +71,12 @@ func (la *LdapAuthenticator) CanLogin(
sr, err := l.Search(searchRequest)
if err != nil {
log.Warn(err)
cclog.Warn(err)
return nil, false
}
if len(sr.Entries) != 1 {
log.Warn("LDAP: User does not exist or too many entries returned")
cclog.Warn("LDAP: User does not exist or too many entries returned")
return nil, false
}
@@ -96,7 +96,7 @@ func (la *LdapAuthenticator) CanLogin(
}
if err := repository.GetUserRepository().AddUser(user); err != nil {
log.Errorf("User '%s' LDAP: Insert into DB failed", username)
cclog.Errorf("User '%s' LDAP: Insert into DB failed", username)
return nil, false
}
@@ -114,14 +114,14 @@ func (la *LdapAuthenticator) Login(
) (*schema.User, error) {
l, err := la.getLdapConnection(false)
if err != nil {
log.Warn("Error while getting ldap connection")
cclog.Warn("Error while getting ldap connection")
return nil, err
}
defer l.Close()
userDn := strings.Replace(config.Keys.LdapConfig.UserBind, "{username}", user.Username, -1)
if err := l.Bind(userDn, r.FormValue("password")); err != nil {
log.Errorf("AUTH/LDAP > Authentication for user %s failed: %v",
cclog.Errorf("AUTH/LDAP > Authentication for user %s failed: %v",
user.Username, err)
return nil, fmt.Errorf("Authentication failed")
}
@@ -148,7 +148,7 @@ func (la *LdapAuthenticator) Sync() error {
l, err := la.getLdapConnection(true)
if err != nil {
log.Error("LDAP connection error")
cclog.Error("LDAP connection error")
return err
}
defer l.Close()
@@ -159,7 +159,7 @@ func (la *LdapAuthenticator) Sync() error {
lc.UserFilter,
[]string{"dn", "uid", la.UserAttr}, nil))
if err != nil {
log.Warn("LDAP search error")
cclog.Warn("LDAP search error")
return err
}
@@ -182,7 +182,7 @@ func (la *LdapAuthenticator) Sync() error {
for username, where := range users {
if where == IN_DB && lc.SyncDelOldUsers {
ur.DelUser(username)
log.Debugf("sync: remove %v (does not show up in LDAP anymore)", username)
cclog.Debugf("sync: remove %v (does not show up in LDAP anymore)", username)
} else if where == IN_LDAP {
name := newnames[username]
@@ -198,9 +198,9 @@ func (la *LdapAuthenticator) Sync() error {
AuthSource: schema.AuthViaLDAP,
}
log.Debugf("sync: add %v (name: %v, roles: [user], ldap: true)", username, name)
cclog.Debugf("sync: add %v (name: %v, roles: [user], ldap: true)", username, name)
if err := ur.AddUser(user); err != nil {
log.Errorf("User '%s' LDAP: Insert into DB failed", username)
cclog.Errorf("User '%s' LDAP: Insert into DB failed", username)
return err
}
}
@@ -213,14 +213,14 @@ func (la *LdapAuthenticator) getLdapConnection(admin bool) (*ldap.Conn, error) {
lc := config.Keys.LdapConfig
conn, err := ldap.DialURL(lc.Url)
if err != nil {
log.Warn("LDAP URL dial failed")
cclog.Warn("LDAP URL dial failed")
return nil, err
}
if admin {
if err := conn.Bind(lc.SearchDN, la.syncPassword); err != nil {
conn.Close()
log.Warn("LDAP connection bind failed")
cclog.Warn("LDAP connection bind failed")
return nil, err
}
}

16
internal/auth/local.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package auth
@@ -8,8 +8,8 @@ import (
"fmt"
"net/http"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"golang.org/x/crypto/bcrypt"
)
@@ -27,19 +27,19 @@ func (la *LocalAuthenticator) CanLogin(
user *schema.User,
username string,
rw http.ResponseWriter,
r *http.Request) (*schema.User, bool) {
r *http.Request,
) (*schema.User, bool) {
return user, user != nil && user.AuthSource == schema.AuthViaLocalPassword
}
func (la *LocalAuthenticator) Login(
user *schema.User,
rw http.ResponseWriter,
r *http.Request) (*schema.User, error) {
r *http.Request,
) (*schema.User, error) {
if e := bcrypt.CompareHashAndPassword([]byte(user.Password),
[]byte(r.FormValue("password"))); e != nil {
log.Errorf("AUTH/LOCAL > Authentication for user %s failed!", user.Username)
cclog.Errorf("AUTH/LOCAL > Authentication for user %s failed!", user.Username)
return nil, fmt.Errorf("Authentication failed")
}

14
internal/auth/oidc.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package auth
@@ -15,8 +15,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/coreos/go-oidc/v3/oidc"
"github.com/gorilla/mux"
"golang.org/x/oauth2"
@@ -51,15 +51,15 @@ func setCallbackCookie(w http.ResponseWriter, r *http.Request, name, value strin
func NewOIDC(a *Authentication) *OIDC {
provider, err := oidc.NewProvider(context.Background(), config.Keys.OpenIDConfig.Provider)
if err != nil {
log.Fatal(err)
cclog.Fatal(err)
}
clientID := os.Getenv("OID_CLIENT_ID")
if clientID == "" {
log.Warn("environment variable 'OID_CLIENT_ID' not set (Open ID connect auth will not work)")
cclog.Warn("environment variable 'OID_CLIENT_ID' not set (Open ID connect auth will not work)")
}
clientSecret := os.Getenv("OID_CLIENT_SECRET")
if clientSecret == "" {
log.Warn("environment variable 'OID_CLIENT_SECRET' not set (Open ID connect auth will not work)")
cclog.Warn("environment variable 'OID_CLIENT_SECRET' not set (Open ID connect auth will not work)")
}
client := &oauth2.Config{
@@ -173,7 +173,7 @@ func (oa *OIDC) OAuth2Callback(rw http.ResponseWriter, r *http.Request) {
}
oa.authentication.SaveSession(rw, r, user)
log.Infof("login successfull: user: %#v (roles: %v, projects: %v)", user.Username, user.Roles, user.Projects)
cclog.Infof("login successfull: user: %#v (roles: %v, projects: %v)", user.Username, user.Roles, user.Projects)
ctx := context.WithValue(r.Context(), repository.ContextUserKey, user)
http.RedirectHandler("/", http.StatusTemporaryRedirect).ServeHTTP(rw, r.WithContext(ctx))
}

14
internal/config/config.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package config
@@ -9,8 +9,8 @@ import (
"encoding/json"
"os"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
var Keys schema.ProgramConfig = schema.ProgramConfig{
@@ -53,20 +53,20 @@ func Init(flagConfigFile string) {
raw, err := os.ReadFile(flagConfigFile)
if err != nil {
if !os.IsNotExist(err) {
log.Abortf("Config Init: Could not read config file '%s'.\nError: %s\n", flagConfigFile, err.Error())
cclog.Abortf("Config Init: Could not read config file '%s'.\nError: %s\n", flagConfigFile, err.Error())
}
} else {
if err := schema.Validate(schema.Config, bytes.NewReader(raw)); err != nil {
log.Abortf("Config Init: Could not validate config file '%s'.\nError: %s\n", flagConfigFile, err.Error())
cclog.Abortf("Config Init: Could not validate config file '%s'.\nError: %s\n", flagConfigFile, err.Error())
}
dec := json.NewDecoder(bytes.NewReader(raw))
dec.DisallowUnknownFields()
if err := dec.Decode(&Keys); err != nil {
log.Abortf("Config Init: Could not decode config file '%s'.\nError: %s\n", flagConfigFile, err.Error())
cclog.Abortf("Config Init: Could not decode config file '%s'.\nError: %s\n", flagConfigFile, err.Error())
}
if Keys.Clusters == nil || len(Keys.Clusters) < 1 {
log.Abort("Config Init: At least one cluster required in config. Exited with error.")
cclog.Abort("Config Init: At least one cluster required in config. Exited with error.")
}
}
}

2
internal/config/config_test.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package config

4
internal/config/default_metrics.go
View File

@@ -1,3 +1,7 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package config
import (

3938
internal/graph/generated/generated.go
View File

File diff suppressed because it is too large Load Diff

2
internal/graph/model/models.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package model

49
internal/graph/model/models_gen.go
View File

@@ -8,7 +8,7 @@ import (
"strconv"
"time"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
"github.com/ClusterCockpit/cc-lib/schema"
)
type Count struct {
@@ -50,6 +50,7 @@ type IntRangeOutput struct {
type JobFilter struct {
Tags []string `json:"tags,omitempty"`
DbID []string `json:"dbId,omitempty"`
JobID *StringInput `json:"jobId,omitempty"`
ArrayJobID *int `json:"arrayJobId,omitempty"`
User *StringInput `json:"user,omitempty"`
@@ -96,14 +97,16 @@ type JobResultList struct {
}
type JobStats struct {
Name string `json:"name"`
Stats *schema.MetricStatistics `json:"stats"`
}
type JobStatsWithScope struct {
Name string `json:"name"`
Scope schema.MetricScope `json:"scope"`
Stats []*ScopedStats `json:"stats"`
ID int `json:"id"`
JobID string `json:"jobId"`
StartTime int `json:"startTime"`
Duration int `json:"duration"`
Cluster string `json:"cluster"`
SubCluster string `json:"subCluster"`
NumNodes int `json:"numNodes"`
NumHWThreads *int `json:"numHWThreads,omitempty"`
NumAccelerators *int `json:"numAccelerators,omitempty"`
Stats []*NamedStats `json:"stats"`
}
type JobsStatistics struct {
@@ -153,12 +156,40 @@ type MetricStatItem struct {
type Mutation struct {
}
type NamedStats struct {
Name string `json:"name"`
Data *schema.MetricStatistics `json:"data"`
}
type NamedStatsWithScope struct {
Name string `json:"name"`
Scope schema.MetricScope `json:"scope"`
Stats []*ScopedStats `json:"stats"`
}
type NodeFilter struct {
Hostname *StringInput `json:"hostname,omitempty"`
Cluster *StringInput `json:"cluster,omitempty"`
NodeState *string `json:"nodeState,omitempty"`
HealthState *schema.NodeState `json:"healthState,omitempty"`
}
type NodeMetrics struct {
Host string `json:"host"`
SubCluster string `json:"subCluster"`
Metrics []*JobMetricWithName `json:"metrics"`
}
type NodeStateResultList struct {
Items []*schema.Node `json:"items"`
Count *int `json:"count,omitempty"`
}
type NodeStats struct {
State string `json:"state"`
Count int `json:"count"`
}
type NodesResultList struct {
Items []*NodeMetrics `json:"items"`
Offset *int `json:"offset,omitempty"`

4
internal/graph/resolver.go
View File

@@ -4,7 +4,7 @@ import (
"sync"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/jmoiron/sqlx"
)
@@ -32,7 +32,7 @@ func Init() {
func GetResolverInstance() *Resolver {
if resolverInstance == nil {
log.Fatal("Authentication module not initialized!")
cclog.Fatal("Authentication module not initialized!")
}
return resolverInstance

226
internal/graph/schema.resolvers.go
View File

@@ -20,8 +20,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/metricDataDispatcher"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
// Partitions is the resolver for the partitions field.
@@ -29,9 +29,15 @@ func (r *clusterResolver) Partitions(ctx context.Context, obj *schema.Cluster) (
return r.Repo.Partitions(obj.Name)
}
// StartTime is the resolver for the startTime field.
func (r *jobResolver) StartTime(ctx context.Context, obj *schema.Job) (*time.Time, error) {
timestamp := time.Unix(obj.StartTime, 0)
return &timestamp, nil
}
// Tags is the resolver for the tags field.
func (r *jobResolver) Tags(ctx context.Context, obj *schema.Job) ([]*schema.Tag, error) {
return r.Repo.GetTags(repository.GetUserFromContext(ctx), &obj.ID)
return r.Repo.GetTags(repository.GetUserFromContext(ctx), obj.ID)
}
// ConcurrentJobs is the resolver for the concurrentJobs field.
@@ -48,7 +54,7 @@ func (r *jobResolver) ConcurrentJobs(ctx context.Context, obj *schema.Job) (*mod
func (r *jobResolver) Footprint(ctx context.Context, obj *schema.Job) ([]*model.FootprintValue, error) {
rawFootprint, err := r.Repo.FetchFootprint(obj)
if err != nil {
log.Warn("Error while fetching job footprint data")
cclog.Warn("Error while fetching job footprint data")
return nil, err
}
@@ -73,7 +79,7 @@ func (r *jobResolver) Footprint(ctx context.Context, obj *schema.Job) ([]*model.
func (r *jobResolver) EnergyFootprint(ctx context.Context, obj *schema.Job) ([]*model.EnergyFootprintValue, error) {
rawEnergyFootprint, err := r.Repo.FetchEnergyFootprint(obj)
if err != nil {
log.Warn("Error while fetching job energy footprint data")
cclog.Warn("Error while fetching job energy footprint data")
return nil, err
}
@@ -137,13 +143,13 @@ func (r *mutationResolver) CreateTag(ctx context.Context, typeArg string, name s
// Create in DB
id, err := r.Repo.CreateTag(typeArg, name, scope)
if err != nil {
log.Warn("Error while creating tag")
cclog.Warn("Error while creating tag")
return nil, err
}
return &schema.Tag{ID: id, Type: typeArg, Name: name, Scope: scope}, nil
} else {
log.Warnf("Not authorized to create tag with scope: %s", scope)
return nil, fmt.Errorf("Not authorized to create tag with scope: %s", scope)
cclog.Warnf("Not authorized to create tag with scope: %s", scope)
return nil, fmt.Errorf("not authorized to create tag with scope: %s", scope)
}
}
@@ -162,7 +168,7 @@ func (r *mutationResolver) AddTagsToJob(ctx context.Context, job string, tagIds
jid, err := strconv.ParseInt(job, 10, 64)
if err != nil {
log.Warn("Error while adding tag to job")
cclog.Warn("Error while adding tag to job")
return nil, err
}
@@ -171,15 +177,15 @@ func (r *mutationResolver) AddTagsToJob(ctx context.Context, job string, tagIds
// Get ID
tid, err := strconv.ParseInt(tagId, 10, 64)
if err != nil {
log.Warn("Error while parsing tag id")
cclog.Warn("Error while parsing tag id")
return nil, err
}
// Test Exists
_, _, tscope, exists := r.Repo.TagInfo(tid)
if !exists {
log.Warnf("Tag does not exist (ID): %d", tid)
return nil, fmt.Errorf("Tag does not exist (ID): %d", tid)
cclog.Warnf("Tag does not exist (ID): %d", tid)
return nil, fmt.Errorf("tag does not exist (ID): %d", tid)
}
// Test Access: Admins && Admin Tag OR Support/Admin and Global Tag OR Everyone && Private Tag
@@ -188,12 +194,12 @@ func (r *mutationResolver) AddTagsToJob(ctx context.Context, job string, tagIds
user.Username == tscope {
// Add to Job
if tags, err = r.Repo.AddTag(user, jid, tid); err != nil {
log.Warn("Error while adding tag")
cclog.Warn("Error while adding tag")
return nil, err
}
} else {
log.Warnf("Not authorized to add tag: %d", tid)
return nil, fmt.Errorf("Not authorized to add tag: %d", tid)
cclog.Warnf("Not authorized to add tag: %d", tid)
return nil, fmt.Errorf("not authorized to add tag: %d", tid)
}
}
@@ -209,7 +215,7 @@ func (r *mutationResolver) RemoveTagsFromJob(ctx context.Context, job string, ta
jid, err := strconv.ParseInt(job, 10, 64)
if err != nil {
log.Warn("Error while parsing job id")
cclog.Warn("Error while parsing job id")
return nil, err
}
@@ -218,15 +224,15 @@ func (r *mutationResolver) RemoveTagsFromJob(ctx context.Context, job string, ta
// Get ID
tid, err := strconv.ParseInt(tagId, 10, 64)
if err != nil {
log.Warn("Error while parsing tag id")
cclog.Warn("Error while parsing tag id")
return nil, err
}
// Test Exists
_, _, tscope, exists := r.Repo.TagInfo(tid)
if !exists {
log.Warnf("Tag does not exist (ID): %d", tid)
return nil, fmt.Errorf("Tag does not exist (ID): %d", tid)
cclog.Warnf("Tag does not exist (ID): %d", tid)
return nil, fmt.Errorf("tag does not exist (ID): %d", tid)
}
// Test Access: Admins && Admin Tag OR Support/Admin and Global Tag OR Everyone && Private Tag
@@ -235,12 +241,12 @@ func (r *mutationResolver) RemoveTagsFromJob(ctx context.Context, job string, ta
user.Username == tscope {
// Remove from Job
if tags, err = r.Repo.RemoveTag(user, jid, tid); err != nil {
log.Warn("Error while removing tag")
cclog.Warn("Error while removing tag")
return nil, err
}
} else {
log.Warnf("Not authorized to remove tag: %d", tid)
return nil, fmt.Errorf("Not authorized to remove tag: %d", tid)
cclog.Warnf("Not authorized to remove tag: %d", tid)
return nil, fmt.Errorf("not authorized to remove tag: %d", tid)
}
}
@@ -261,29 +267,29 @@ func (r *mutationResolver) RemoveTagFromList(ctx context.Context, tagIds []strin
// Get ID
tid, err := strconv.ParseInt(tagId, 10, 64)
if err != nil {
log.Warn("Error while parsing tag id for removal")
cclog.Warn("Error while parsing tag id for removal")
return nil, err
}
// Test Exists
_, _, tscope, exists := r.Repo.TagInfo(tid)
if !exists {
log.Warnf("Tag does not exist (ID): %d", tid)
return nil, fmt.Errorf("Tag does not exist (ID): %d", tid)
cclog.Warnf("Tag does not exist (ID): %d", tid)
return nil, fmt.Errorf("tag does not exist (ID): %d", tid)
}
// Test Access: Admins && Admin Tag OR Everyone && Private Tag
if user.HasRole(schema.RoleAdmin) && (tscope == "global" || tscope == "admin") || user.Username == tscope {
// Remove from DB
if err = r.Repo.RemoveTagById(tid); err != nil {
log.Warn("Error while removing tag")
cclog.Warn("Error while removing tag")
return nil, err
} else {
tags = append(tags, int(tid))
}
} else {
log.Warnf("Not authorized to remove tag: %d", tid)
return nil, fmt.Errorf("Not authorized to remove tag: %d", tid)
cclog.Warnf("Not authorized to remove tag: %d", tid)
return nil, fmt.Errorf("not authorized to remove tag: %d", tid)
}
}
return tags, nil
@@ -292,13 +298,28 @@ func (r *mutationResolver) RemoveTagFromList(ctx context.Context, tagIds []strin
// UpdateConfiguration is the resolver for the updateConfiguration field.
func (r *mutationResolver) UpdateConfiguration(ctx context.Context, name string, value string) (*string, error) {
if err := repository.GetUserCfgRepo().UpdateConfig(name, value, repository.GetUserFromContext(ctx)); err != nil {
log.Warn("Error while updating user config")
cclog.Warn("Error while updating user config")
return nil, err
}
return nil, nil
}
// NodeState is the resolver for the nodeState field.
func (r *nodeResolver) NodeState(ctx context.Context, obj *schema.Node) (string, error) {
panic(fmt.Errorf("not implemented: NodeState - nodeState"))
}
// HealthState is the resolver for the HealthState field.
func (r *nodeResolver) HealthState(ctx context.Context, obj *schema.Node) (schema.NodeState, error) {
panic(fmt.Errorf("not implemented: HealthState - HealthState"))
}
// MetaData is the resolver for the metaData field.
func (r *nodeResolver) MetaData(ctx context.Context, obj *schema.Node) (any, error) {
panic(fmt.Errorf("not implemented: MetaData - metaData"))
}
// Clusters is the resolver for the clusters field.
func (r *queryResolver) Clusters(ctx context.Context) ([]*schema.Cluster, error) {
return archive.Clusters, nil
@@ -323,7 +344,7 @@ func (r *queryResolver) User(ctx context.Context, username string) (*model.User,
func (r *queryResolver) AllocatedNodes(ctx context.Context, cluster string) ([]*model.Count, error) {
data, err := r.Repo.AllocatedNodes(cluster)
if err != nil {
log.Warn("Error while fetching allocated nodes")
cclog.Warn("Error while fetching allocated nodes")
return nil, err
}
@@ -338,17 +359,41 @@ func (r *queryResolver) AllocatedNodes(ctx context.Context, cluster string) ([]*
return counts, nil
}
// Node is the resolver for the node field.
func (r *queryResolver) Node(ctx context.Context, id string) (*schema.Node, error) {
repo := repository.GetNodeRepository()
numericId, err := strconv.ParseInt(id, 10, 64)
if err != nil {
cclog.Warn("Error while parsing job id")
return nil, err
}
return repo.GetNode(numericId, false)
}
// Nodes is the resolver for the nodes field.
func (r *queryResolver) Nodes(ctx context.Context, filter []*model.NodeFilter, order *model.OrderByInput) (*model.NodeStateResultList, error) {
repo := repository.GetNodeRepository()
nodes, err := repo.QueryNodes(ctx, filter, order)
count := len(nodes)
return &model.NodeStateResultList{Items: nodes, Count: &count}, err
}
// NodeStats is the resolver for the nodeStats field.
func (r *queryResolver) NodeStats(ctx context.Context, filter []*model.NodeFilter) ([]*model.NodeStats, error) {
panic(fmt.Errorf("not implemented: NodeStats - nodeStats"))
}
// Job is the resolver for the job field.
func (r *queryResolver) Job(ctx context.Context, id string) (*schema.Job, error) {
numericId, err := strconv.ParseInt(id, 10, 64)
if err != nil {
log.Warn("Error while parsing job id")
cclog.Warn("Error while parsing job id")
return nil, err
}
job, err := r.Repo.FindById(ctx, numericId)
if err != nil {
log.Warn("Error while finding job by id")
cclog.Warn("Error while finding job by id")
return nil, err
}
@@ -375,13 +420,13 @@ func (r *queryResolver) JobMetrics(ctx context.Context, id string, metrics []str
job, err := r.Query().Job(ctx, id)
if err != nil {
log.Warn("Error while querying job for metrics")
cclog.Warn("Error while querying job for metrics")
return nil, err
}
data, err := metricDataDispatcher.LoadData(job, metrics, scopes, ctx, *resolution)
if err != nil {
log.Warn("Error while loading job data")
cclog.Warn("Error while loading job data")
return nil, err
}
@@ -400,24 +445,24 @@ func (r *queryResolver) JobMetrics(ctx context.Context, id string, metrics []str
}
// JobStats is the resolver for the jobStats field.
func (r *queryResolver) JobStats(ctx context.Context, id string, metrics []string) ([]*model.JobStats, error) {
func (r *queryResolver) JobStats(ctx context.Context, id string, metrics []string) ([]*model.NamedStats, error) {
job, err := r.Query().Job(ctx, id)
if err != nil {
log.Warnf("Error while querying job %s for metadata", id)
cclog.Warnf("Error while querying job %s for metadata", id)
return nil, err
}
data, err := metricDataDispatcher.LoadJobStats(job, metrics, ctx)
if err != nil {
log.Warnf("Error while loading jobStats data for job id %s", id)
cclog.Warnf("Error while loading jobStats data for job id %s", id)
return nil, err
}
res := []*model.JobStats{}
res := []*model.NamedStats{}
for name, md := range data {
res = append(res, &model.JobStats{
Name: name,
Stats: &md,
res = append(res, &model.NamedStats{
Name: name,
Data: &md,
})
}
@@ -425,20 +470,20 @@ func (r *queryResolver) JobStats(ctx context.Context, id string, metrics []strin
}
// ScopedJobStats is the resolver for the scopedJobStats field.
func (r *queryResolver) ScopedJobStats(ctx context.Context, id string, metrics []string, scopes []schema.MetricScope) ([]*model.JobStatsWithScope, error) {
func (r *queryResolver) ScopedJobStats(ctx context.Context, id string, metrics []string, scopes []schema.MetricScope) ([]*model.NamedStatsWithScope, error) {
job, err := r.Query().Job(ctx, id)
if err != nil {
log.Warnf("Error while querying job %s for metadata", id)
cclog.Warnf("Error while querying job %s for metadata", id)
return nil, err
}
data, err := metricDataDispatcher.LoadScopedJobStats(job, metrics, scopes, ctx)
if err != nil {
log.Warnf("Error while loading scopedJobStats data for job id %s", id)
cclog.Warnf("Error while loading scopedJobStats data for job id %s", id)
return nil, err
}
res := make([]*model.JobStatsWithScope, 0)
res := make([]*model.NamedStatsWithScope, 0)
for name, scoped := range data {
for scope, stats := range scoped {
@@ -451,7 +496,7 @@ func (r *queryResolver) ScopedJobStats(ctx context.Context, id string, metrics [
})
}
res = append(res, &model.JobStatsWithScope{
res = append(res, &model.NamedStatsWithScope{
Name: name,
Scope: scope,
Stats: mdlStats,
@@ -462,12 +507,6 @@ func (r *queryResolver) ScopedJobStats(ctx context.Context, id string, metrics [
return res, nil
}
// JobsFootprints is the resolver for the jobsFootprints field.
func (r *queryResolver) JobsFootprints(ctx context.Context, filter []*model.JobFilter, metrics []string) (*model.Footprints, error) {
// NOTE: Legacy Naming! This resolver is for normalized histograms in analysis view only - *Not* related to DB "footprint" column!
return r.jobsFootprints(ctx, filter, metrics)
}
// Jobs is the resolver for the jobs field.
func (r *queryResolver) Jobs(ctx context.Context, filter []*model.JobFilter, page *model.PageRequest, order *model.OrderByInput) (*model.JobResultList, error) {
if page == nil {
@@ -479,13 +518,13 @@ func (r *queryResolver) Jobs(ctx context.Context, filter []*model.JobFilter, pag
jobs, err := r.Repo.QueryJobs(ctx, filter, page, order)
if err != nil {
log.Warn("Error while querying jobs")
cclog.Warn("Error while querying jobs")
return nil, err
}
count, err := r.Repo.CountJobs(ctx, filter)
if err != nil {
log.Warn("Error while counting jobs")
cclog.Warn("Error while counting jobs")
return nil, err
}
@@ -501,14 +540,11 @@ func (r *queryResolver) Jobs(ctx context.Context, filter []*model.JobFilter, pag
}
nextJobs, err := r.Repo.QueryJobs(ctx, filter, nextPage, order)
if err != nil {
log.Warn("Error while querying next jobs")
cclog.Warn("Error while querying next jobs")
return nil, err
}
hasNextPage := false
if len(nextJobs) == 1 {
hasNextPage = true
}
hasNextPage := len(nextJobs) == 1
return &model.JobResultList{Items: jobs, Count: &count, HasNextPage: &hasNextPage}, nil
}
@@ -519,8 +555,8 @@ func (r *queryResolver) JobsStatistics(ctx context.Context, filter []*model.JobF
var stats []*model.JobsStatistics
// Top Level Defaults
var defaultDurationBins string = "1h"
var defaultMetricBins int = 10
defaultDurationBins := "1h"
defaultMetricBins := 10
if requireField(ctx, "totalJobs") || requireField(ctx, "totalWalltime") || requireField(ctx, "totalNodes") || requireField(ctx, "totalCores") ||
requireField(ctx, "totalAccs") || requireField(ctx, "totalNodeHours") || requireField(ctx, "totalCoreHours") || requireField(ctx, "totalAccHours") {
@@ -589,6 +625,62 @@ func (r *queryResolver) JobsStatistics(ctx context.Context, filter []*model.JobF
return stats, nil
}
// JobsMetricStats is the resolver for the jobsMetricStats field.
func (r *queryResolver) JobsMetricStats(ctx context.Context, filter []*model.JobFilter, metrics []string) ([]*model.JobStats, error) {
// No Paging, Fixed Order by StartTime ASC
order := &model.OrderByInput{
Field: "startTime",
Type: "col",
Order: "ASC",
}
jobs, err := r.Repo.QueryJobs(ctx, filter, nil, order)
if err != nil {
cclog.Warn("Error while querying jobs for comparison")
return nil, err
}
res := []*model.JobStats{}
for _, job := range jobs {
data, err := metricDataDispatcher.LoadJobStats(job, metrics, ctx)
if err != nil {
cclog.Warnf("Error while loading comparison jobStats data for job id %d", job.JobID)
continue
// return nil, err
}
sres := []*model.NamedStats{}
for name, md := range data {
sres = append(sres, &model.NamedStats{
Name: name,
Data: &md,
})
}
numThreadsInt := int(job.NumHWThreads)
numAccsInt := int(job.NumAcc)
res = append(res, &model.JobStats{
ID: int(*job.ID),
JobID: strconv.Itoa(int(job.JobID)),
StartTime: int(job.StartTime),
Duration: int(job.Duration),
Cluster: job.Cluster,
SubCluster: job.SubCluster,
NumNodes: int(job.NumNodes),
NumHWThreads: &numThreadsInt,
NumAccelerators: &numAccsInt,
Stats: sres,
})
}
return res, err
}
// JobsFootprints is the resolver for the jobsFootprints field.
func (r *queryResolver) JobsFootprints(ctx context.Context, filter []*model.JobFilter, metrics []string) (*model.Footprints, error) {
// NOTE: Legacy Naming! This resolver is for normalized histograms in analysis view only - *Not* related to DB "footprint" column!
return r.jobsFootprints(ctx, filter, metrics)
}
// RooflineHeatmap is the resolver for the rooflineHeatmap field.
func (r *queryResolver) RooflineHeatmap(ctx context.Context, filter []*model.JobFilter, rows int, cols int, minX float64, minY float64, maxX float64, maxY float64) ([][]float64, error) {
return r.rooflineHeatmap(ctx, filter, rows, cols, minX, minY, maxX, maxY)
@@ -609,7 +701,7 @@ func (r *queryResolver) NodeMetrics(ctx context.Context, cluster string, nodes [
data, err := metricDataDispatcher.LoadNodeData(cluster, metrics, nodes, scopes, from, to, ctx)
if err != nil {
log.Warn("error while loading node data")
cclog.Warn("error while loading node data")
return nil, err
}
@@ -621,7 +713,7 @@ func (r *queryResolver) NodeMetrics(ctx context.Context, cluster string, nodes [
}
host.SubCluster, err = archive.GetSubClusterByNode(cluster, hostname)
if err != nil {
log.Warnf("error in nodeMetrics resolver: %s", err)
cclog.Warnf("error in nodeMetrics resolver: %s", err)
}
for metric, scopedMetrics := range metrics {
@@ -665,7 +757,7 @@ func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, sub
data, totalNodes, hasNextPage, err := metricDataDispatcher.LoadNodeListData(cluster, subCluster, nodeFilter, metrics, scopes, *resolution, from, to, page, ctx)
if err != nil {
log.Warn("error while loading node data")
cclog.Warn("error while loading node data")
return nil, err
}
@@ -677,7 +769,7 @@ func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, sub
}
host.SubCluster, err = archive.GetSubClusterByNode(cluster, hostname)
if err != nil {
log.Warnf("error in nodeMetrics resolver: %s", err)
cclog.Warnf("error in nodeMetrics resolver: %s", err)
}
for metric, scopedMetrics := range metrics {
@@ -723,6 +815,9 @@ func (r *Resolver) MetricValue() generated.MetricValueResolver { return &metricV
// Mutation returns generated.MutationResolver implementation.
func (r *Resolver) Mutation() generated.MutationResolver { return &mutationResolver{r} }
// Node returns generated.NodeResolver implementation.
func (r *Resolver) Node() generated.NodeResolver { return &nodeResolver{r} }
// Query returns generated.QueryResolver implementation.
func (r *Resolver) Query() generated.QueryResolver { return &queryResolver{r} }
@@ -733,5 +828,6 @@ type clusterResolver struct{ *Resolver }
type jobResolver struct{ *Resolver }
type metricValueResolver struct{ *Resolver }
type mutationResolver struct{ *Resolver }
type nodeResolver struct{ *Resolver }
type queryResolver struct{ *Resolver }
type subClusterResolver struct{ *Resolver }

19
internal/graph/util.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package graph
@@ -12,9 +12,8 @@ import (
"github.com/99designs/gqlgen/graphql"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/internal/metricDataDispatcher"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
// "github.com/ClusterCockpit/cc-backend/pkg/archive"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
const MAX_JOBS_FOR_ANALYSIS = 500
@@ -28,7 +27,7 @@ func (r *queryResolver) rooflineHeatmap(
) ([][]float64, error) {
jobs, err := r.Repo.QueryJobs(ctx, filter, &model.PageRequest{Page: 1, ItemsPerPage: MAX_JOBS_FOR_ANALYSIS + 1}, nil)
if err != nil {
log.Error("Error while querying jobs for roofline")
cclog.Error("Error while querying jobs for roofline")
return nil, err
}
if len(jobs) > MAX_JOBS_FOR_ANALYSIS {
@@ -56,13 +55,13 @@ func (r *queryResolver) rooflineHeatmap(
jobdata, err := metricDataDispatcher.LoadData(job, []string{"flops_any", "mem_bw"}, []schema.MetricScope{schema.MetricScopeNode}, ctx, 0)
if err != nil {
log.Errorf("Error while loading roofline metrics for job %d", job.ID)
cclog.Errorf("Error while loading roofline metrics for job %d", job.ID)
return nil, err
}
flops_, membw_ := jobdata["flops_any"], jobdata["mem_bw"]
if flops_ == nil && membw_ == nil {
log.Infof("rooflineHeatmap(): 'flops_any' or 'mem_bw' missing for job %d", job.ID)
cclog.Infof("rooflineHeatmap(): 'flops_any' or 'mem_bw' missing for job %d", job.ID)
continue
// return nil, fmt.Errorf("GRAPH/UTIL > 'flops_any' or 'mem_bw' missing for job %d", job.ID)
}
@@ -70,7 +69,7 @@ func (r *queryResolver) rooflineHeatmap(
flops, ok1 := flops_["node"]
membw, ok2 := membw_["node"]
if !ok1 || !ok2 {
log.Info("rooflineHeatmap() query not implemented for where flops_any or mem_bw not available at 'node' level")
cclog.Info("rooflineHeatmap() query not implemented for where flops_any or mem_bw not available at 'node' level")
continue
// TODO/FIXME:
// return nil, errors.New("GRAPH/UTIL > todo: rooflineHeatmap() query not implemented for where flops_any or mem_bw not available at 'node' level")
@@ -105,7 +104,7 @@ func (r *queryResolver) rooflineHeatmap(
func (r *queryResolver) jobsFootprints(ctx context.Context, filter []*model.JobFilter, metrics []string) (*model.Footprints, error) {
jobs, err := r.Repo.QueryJobs(ctx, filter, &model.PageRequest{Page: 1, ItemsPerPage: MAX_JOBS_FOR_ANALYSIS + 1}, nil)
if err != nil {
log.Error("Error while querying jobs for footprint")
cclog.Error("Error while querying jobs for footprint")
return nil, err
}
if len(jobs) > MAX_JOBS_FOR_ANALYSIS {
@@ -128,7 +127,7 @@ func (r *queryResolver) jobsFootprints(ctx context.Context, filter []*model.JobF
}
if err := metricDataDispatcher.LoadAverages(job, metrics, avgs, ctx); err != nil {
log.Error("Error while loading averages for footprint")
cclog.Error("Error while loading averages for footprint")
return nil, err
}

45
internal/importer/handleImport.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package importer
@@ -15,8 +15,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
// Import all jobs specified as `<path-to-meta.json>:<path-to-data.json>,...`
@@ -31,7 +31,7 @@ func HandleImportFlag(flag string) error {
raw, err := os.ReadFile(files[0])
if err != nil {
log.Warn("Error while reading metadata file for import")
cclog.Warn("Error while reading metadata file for import")
return err
}
@@ -42,15 +42,18 @@ func HandleImportFlag(flag string) error {
}
dec := json.NewDecoder(bytes.NewReader(raw))
dec.DisallowUnknownFields()
job := schema.JobMeta{BaseJob: schema.JobDefaults}
job := schema.Job{
Exclusive: 1,
MonitoringStatus: schema.MonitoringStatusRunningOrArchiving,
}
if err = dec.Decode(&job); err != nil {
log.Warn("Error while decoding raw json metadata for import")
cclog.Warn("Error while decoding raw json metadata for import")
return err
}
raw, err = os.ReadFile(files[1])
if err != nil {
log.Warn("Error while reading jobdata file for import")
cclog.Warn("Error while reading jobdata file for import")
return err
}
@@ -63,7 +66,7 @@ func HandleImportFlag(flag string) error {
dec.DisallowUnknownFields()
jobData := schema.JobData{}
if err = dec.Decode(&jobData); err != nil {
log.Warn("Error while decoding raw json jobdata for import")
cclog.Warn("Error while decoding raw json jobdata for import")
return err
}
@@ -71,7 +74,7 @@ func HandleImportFlag(flag string) error {
sc, err := archive.GetSubCluster(job.Cluster, job.SubCluster)
if err != nil {
log.Errorf("cannot get subcluster: %s", err.Error())
cclog.Errorf("cannot get subcluster: %s", err.Error())
return err
}
@@ -91,7 +94,7 @@ func HandleImportFlag(flag string) error {
job.RawFootprint, err = json.Marshal(job.Footprint)
if err != nil {
log.Warn("Error while marshaling job footprint")
cclog.Warn("Error while marshaling job footprint")
return err
}
@@ -105,7 +108,7 @@ func HandleImportFlag(flag string) error {
if i, err := archive.MetricIndex(sc.MetricConfig, fp); err == nil {
// Note: For DB data, calculate and save as kWh
if sc.MetricConfig[i].Energy == "energy" { // this metric has energy as unit (Joules)
log.Warnf("Update EnergyFootprint for Job %d and Metric %s on cluster %s: Set to 'energy' in cluster.json: Not implemented, will return 0.0", job.JobID, job.Cluster, fp)
cclog.Warnf("Update EnergyFootprint for Job %d and Metric %s on cluster %s: Set to 'energy' in cluster.json: Not implemented, will return 0.0", job.JobID, job.Cluster, fp)
// FIXME: Needs sum as stats type
} else if sc.MetricConfig[i].Energy == "power" { // this metric has power as unit (Watt)
// Energy: Power (in Watts) * Time (in Seconds)
@@ -117,7 +120,7 @@ func HandleImportFlag(flag string) error {
metricEnergy = math.Round(rawEnergy*100.0) / 100.0
}
} else {
log.Warnf("Error while collecting energy metric %s for job, DB ID '%v', return '0.0'", fp, job.ID)
cclog.Warnf("Error while collecting energy metric %s for job, DB ID '%v', return '0.0'", fp, job.ID)
}
job.EnergyFootprint[fp] = metricEnergy
@@ -126,45 +129,45 @@ func HandleImportFlag(flag string) error {
job.Energy = (math.Round(totalEnergy*100.0) / 100.0)
if job.RawEnergyFootprint, err = json.Marshal(job.EnergyFootprint); err != nil {
log.Warnf("Error while marshaling energy footprint for job INTO BYTES, DB ID '%v'", job.ID)
cclog.Warnf("Error while marshaling energy footprint for job INTO BYTES, DB ID '%v'", job.ID)
return err
}
job.RawResources, err = json.Marshal(job.Resources)
if err != nil {
log.Warn("Error while marshaling job resources")
cclog.Warn("Error while marshaling job resources")
return err
}
job.RawMetaData, err = json.Marshal(job.MetaData)
if err != nil {
log.Warn("Error while marshaling job metadata")
cclog.Warn("Error while marshaling job metadata")
return err
}
if err = SanityChecks(&job.BaseJob); err != nil {
log.Warn("BaseJob SanityChecks failed")
if err = SanityChecks(&job); err != nil {
cclog.Warn("BaseJob SanityChecks failed")
return err
}
if err = archive.GetHandle().ImportJob(&job, &jobData); err != nil {
log.Error("Error while importing job")
cclog.Error("Error while importing job")
return err
}
id, err := r.InsertJob(&job)
if err != nil {
log.Warn("Error while job db insert")
cclog.Warn("Error while job db insert")
return err
}
for _, tag := range job.Tags {
if err := r.ImportTag(id, tag.Type, tag.Name, tag.Scope); err != nil {
log.Error("Error while adding or creating tag on import")
cclog.Error("Error while adding or creating tag on import")
return err
}
}
log.Infof("successfully imported a new job (jobId: %d, cluster: %s, dbid: %d)", job.JobID, job.Cluster, id)
cclog.Infof("successfully imported a new job (jobId: %d, cluster: %s, dbid: %d)", job.JobID, job.Cluster, id)
}
return nil
}

8
internal/importer/importer_test.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package importer_test
@@ -16,7 +16,7 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/importer"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
)
func copyFile(s string, d string) error {
@@ -78,7 +78,7 @@ func setup(t *testing.T) *repository.JobRepository {
}
]}`
log.Init("info", true)
cclog.Init("info", true)
tmpdir := t.TempDir()
jobarchive := filepath.Join(tmpdir, "job-archive")
@@ -166,7 +166,7 @@ func TestHandleImportFlag(t *testing.T) {
}
result := readResult(t, testname)
job, err := r.Find(&result.JobId, &result.Cluster, &result.StartTime)
job, err := r.FindCached(&result.JobId, &result.Cluster, &result.StartTime)
if err != nil {
t.Fatal(err)
}

69
internal/importer/initDB.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package importer
@@ -13,8 +13,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
const (
@@ -27,15 +27,15 @@ const (
func InitDB() error {
r := repository.GetJobRepository()
if err := r.Flush(); err != nil {
log.Errorf("repository initDB(): %v", err)
cclog.Errorf("repository initDB(): %v", err)
return err
}
starttime := time.Now()
log.Print("Building job table...")
cclog.Print("Building job table...")
t, err := r.TransactionInit()
if err != nil {
log.Warn("Error while initializing SQL transactions")
cclog.Warn("Error while initializing SQL transactions")
return err
}
tags := make(map[string]int64)
@@ -60,19 +60,14 @@ func InitDB() error {
}
jobMeta.MonitoringStatus = schema.MonitoringStatusArchivingSuccessful
job := schema.Job{
BaseJob: jobMeta.BaseJob,
StartTime: time.Unix(jobMeta.StartTime, 0),
StartTimeUnix: jobMeta.StartTime,
}
sc, err := archive.GetSubCluster(jobMeta.Cluster, jobMeta.SubCluster)
if err != nil {
log.Errorf("cannot get subcluster: %s", err.Error())
cclog.Errorf("cannot get subcluster: %s", err.Error())
return err
}
job.Footprint = make(map[string]float64)
jobMeta.Footprint = make(map[string]float64)
for _, fp := range sc.Footprint {
statType := "avg"
@@ -83,16 +78,16 @@ func InitDB() error {
name := fmt.Sprintf("%s_%s", fp, statType)
job.Footprint[name] = repository.LoadJobStat(jobMeta, fp, statType)
jobMeta.Footprint[name] = repository.LoadJobStat(jobMeta, fp, statType)
}
job.RawFootprint, err = json.Marshal(job.Footprint)
jobMeta.RawFootprint, err = json.Marshal(jobMeta.Footprint)
if err != nil {
log.Warn("Error while marshaling job footprint")
cclog.Warn("Error while marshaling job footprint")
return err
}
job.EnergyFootprint = make(map[string]float64)
jobMeta.EnergyFootprint = make(map[string]float64)
// Total Job Energy Outside Loop
totalEnergy := 0.0
@@ -102,7 +97,7 @@ func InitDB() error {
if i, err := archive.MetricIndex(sc.MetricConfig, fp); err == nil {
// Note: For DB data, calculate and save as kWh
if sc.MetricConfig[i].Energy == "energy" { // this metric has energy as unit (Joules)
log.Warnf("Update EnergyFootprint for Job %d and Metric %s on cluster %s: Set to 'energy' in cluster.json: Not implemented, will return 0.0", jobMeta.JobID, jobMeta.Cluster, fp)
cclog.Warnf("Update EnergyFootprint for Job %d and Metric %s on cluster %s: Set to 'energy' in cluster.json: Not implemented, will return 0.0", jobMeta.JobID, jobMeta.Cluster, fp)
// FIXME: Needs sum as stats type
} else if sc.MetricConfig[i].Energy == "power" { // this metric has power as unit (Watt)
// Energy: Power (in Watts) * Time (in Seconds)
@@ -114,48 +109,48 @@ func InitDB() error {
metricEnergy = math.Round(rawEnergy*100.0) / 100.0
}
} else {
log.Warnf("Error while collecting energy metric %s for job, DB ID '%v', return '0.0'", fp, jobMeta.ID)
cclog.Warnf("Error while collecting energy metric %s for job, DB ID '%v', return '0.0'", fp, jobMeta.ID)
}
job.EnergyFootprint[fp] = metricEnergy
jobMeta.EnergyFootprint[fp] = metricEnergy
totalEnergy += metricEnergy
}
job.Energy = (math.Round(totalEnergy*100.0) / 100.0)
if job.RawEnergyFootprint, err = json.Marshal(job.EnergyFootprint); err != nil {
log.Warnf("Error while marshaling energy footprint for job INTO BYTES, DB ID '%v'", jobMeta.ID)
jobMeta.Energy = (math.Round(totalEnergy*100.0) / 100.0)
if jobMeta.RawEnergyFootprint, err = json.Marshal(jobMeta.EnergyFootprint); err != nil {
cclog.Warnf("Error while marshaling energy footprint for job INTO BYTES, DB ID '%v'", jobMeta.ID)
return err
}
job.RawResources, err = json.Marshal(job.Resources)
jobMeta.RawResources, err = json.Marshal(jobMeta.Resources)
if err != nil {
log.Errorf("repository initDB(): %v", err)
cclog.Errorf("repository initDB(): %v", err)
errorOccured++
continue
}
job.RawMetaData, err = json.Marshal(job.MetaData)
jobMeta.RawMetaData, err = json.Marshal(jobMeta.MetaData)
if err != nil {
log.Errorf("repository initDB(): %v", err)
cclog.Errorf("repository initDB(): %v", err)
errorOccured++
continue
}
if err := SanityChecks(&job.BaseJob); err != nil {
log.Errorf("repository initDB(): %v", err)
if err := SanityChecks(jobMeta); err != nil {
cclog.Errorf("repository initDB(): %v", err)
errorOccured++
continue
}
id, err := r.TransactionAddNamed(t,
repository.NamedJobInsert, job)
repository.NamedJobInsert, jobMeta)
if err != nil {
log.Errorf("repository initDB(): %v", err)
cclog.Errorf("repository initDB(): %v", err)
errorOccured++
continue
}
for _, tag := range job.Tags {
for _, tag := range jobMeta.Tags {
tagstr := tag.Name + ":" + tag.Type
tagId, ok := tags[tagstr]
if !ok {
@@ -163,7 +158,7 @@ func InitDB() error {
addTagQuery,
tag.Name, tag.Type)
if err != nil {
log.Errorf("Error adding tag: %v", err)
cclog.Errorf("Error adding tag: %v", err)
errorOccured++
continue
}
@@ -181,21 +176,21 @@ func InitDB() error {
}
if errorOccured > 0 {
log.Warnf("Error in import of %d jobs!", errorOccured)
cclog.Warnf("Error in import of %d jobs!", errorOccured)
}
r.TransactionEnd(t)
log.Printf("A total of %d jobs have been registered in %.3f seconds.\n", i, time.Since(starttime).Seconds())
cclog.Printf("A total of %d jobs have been registered in %.3f seconds.\n", i, time.Since(starttime).Seconds())
return nil
}
// This function also sets the subcluster if necessary!
func SanityChecks(job *schema.BaseJob) error {
func SanityChecks(job *schema.Job) error {
if c := archive.GetCluster(job.Cluster); c == nil {
return fmt.Errorf("no such cluster: %v", job.Cluster)
}
if err := archive.AssignSubCluster(job); err != nil {
log.Warn("Error while assigning subcluster to job")
cclog.Warn("Error while assigning subcluster to job")
return err
}
if !job.State.Valid() {

4
internal/importer/normalize.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package importer
@@ -7,7 +7,7 @@ package importer
import (
"math"
ccunits "github.com/ClusterCockpit/cc-units"
ccunits "github.com/ClusterCockpit/cc-lib/ccUnits"
)
func getNormalizationFactor(v float64) (float64, int) {

4
internal/importer/normalize_test.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package importer
@@ -8,7 +8,7 @@ import (
"fmt"
"testing"
ccunits "github.com/ClusterCockpit/cc-units"
ccunits "github.com/ClusterCockpit/cc-lib/ccUnits"
)
func TestNormalizeFactor(t *testing.T) {

39
internal/metricDataDispatcher/dataLoader.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package metricDataDispatcher
@@ -14,10 +14,10 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/internal/metricdata"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/lrucache"
"github.com/ClusterCockpit/cc-backend/pkg/resampler"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/lrucache"
"github.com/ClusterCockpit/cc-lib/resampler"
"github.com/ClusterCockpit/cc-lib/schema"
)
var cache *lrucache.Cache = lrucache.New(128 * 1024 * 1024)
@@ -68,10 +68,10 @@ func LoadData(job *schema.Job,
jd, err = repo.LoadData(job, metrics, scopes, ctx, resolution)
if err != nil {
if len(jd) != 0 {
log.Warnf("partial error: %s", err.Error())
cclog.Warnf("partial error: %s", err.Error())
// return err, 0, 0 // Reactivating will block archiving on one partial error
} else {
log.Error("Error while loading job data from metric repository")
cclog.Error("Error while loading job data from metric repository")
return err, 0, 0
}
}
@@ -80,15 +80,15 @@ func LoadData(job *schema.Job,
var jd_temp schema.JobData
jd_temp, err = archive.GetHandle().LoadJobData(job)
if err != nil {
log.Error("Error while loading job data from archive")
cclog.Error("Error while loading job data from archive")
return err, 0, 0
}
//Deep copy the cached archive hashmap
// Deep copy the cached archive hashmap
jd = metricdata.DeepCopy(jd_temp)
//Resampling for archived data.
//Pass the resolution from frontend here.
// Resampling for archived data.
// Pass the resolution from frontend here.
for _, v := range jd {
for _, v_ := range v {
timestep := 0
@@ -178,7 +178,7 @@ func LoadData(job *schema.Job,
})
if err, ok := data.(error); ok {
log.Error("Error in returned dataset")
cclog.Error("Error in returned dataset")
return nil, err
}
@@ -203,7 +203,7 @@ func LoadAverages(
stats, err := repo.LoadStats(job, metrics, ctx) // #166 how to handle stats for acc normalizazion?
if err != nil {
log.Errorf("Error while loading statistics for job %v (User %v, Project %v)", job.JobID, job.User, job.Project)
cclog.Errorf("Error while loading statistics for job %v (User %v, Project %v)", job.JobID, job.User, job.Project)
return err
}
@@ -231,7 +231,6 @@ func LoadScopedJobStats(
scopes []schema.MetricScope,
ctx context.Context,
) (schema.ScopedJobStats, error) {
if job.State != schema.JobStateRunning && !config.Keys.DisableArchive {
return archive.LoadScopedStatsFromArchive(job, metrics, scopes)
}
@@ -243,7 +242,7 @@ func LoadScopedJobStats(
scopedStats, err := repo.LoadScopedStats(job, metrics, scopes, ctx)
if err != nil {
log.Errorf("error while loading scoped statistics for job %d (User %s, Project %s)", job.JobID, job.User, job.Project)
cclog.Errorf("error while loading scoped statistics for job %d (User %s, Project %s)", job.JobID, job.User, job.Project)
return nil, err
}
@@ -268,7 +267,7 @@ func LoadJobStats(
stats, err := repo.LoadStats(job, metrics, ctx)
if err != nil {
log.Errorf("error while loading statistics for job %d (User %s, Project %s)", job.JobID, job.User, job.Project)
cclog.Errorf("error while loading statistics for job %d (User %s, Project %s)", job.JobID, job.User, job.Project)
return data, err
}
@@ -318,9 +317,9 @@ func LoadNodeData(
data, err := repo.LoadNodeData(cluster, metrics, nodes, scopes, from, to, ctx)
if err != nil {
if len(data) != 0 {
log.Warnf("partial error: %s", err.Error())
cclog.Warnf("partial error: %s", err.Error())
} else {
log.Error("Error while loading node data from metric repository")
cclog.Error("Error while loading node data from metric repository")
return nil, err
}
}
@@ -355,9 +354,9 @@ func LoadNodeListData(
data, totalNodes, hasNextPage, err := repo.LoadNodeListData(cluster, subCluster, nodeFilter, metrics, scopes, resolution, from, to, page, ctx)
if err != nil {
if len(data) != 0 {
log.Warnf("partial error: %s", err.Error())
cclog.Warnf("partial error: %s", err.Error())
} else {
log.Error("Error while loading node data from metric repository")
cclog.Error("Error while loading node data from metric repository")
return nil, totalNodes, hasNextPage, err
}
}

61
internal/metricdata/cc-metric-store.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package metricdata
@@ -18,8 +18,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
type CCMetricStoreConfig struct {
@@ -82,7 +82,7 @@ type ApiMetricData struct {
func (ccms *CCMetricStore) Init(rawConfig json.RawMessage) error {
var config CCMetricStoreConfig
if err := json.Unmarshal(rawConfig, &config); err != nil {
log.Warn("Error while unmarshaling raw json config")
cclog.Warn("Error while unmarshaling raw json config")
return err
}
@@ -129,13 +129,13 @@ func (ccms *CCMetricStore) doRequest(
) (*ApiQueryResponse, error) {
buf := &bytes.Buffer{}
if err := json.NewEncoder(buf).Encode(body); err != nil {
log.Errorf("Error while encoding request body: %s", err.Error())
cclog.Errorf("Error while encoding request body: %s", err.Error())
return nil, err
}
req, err := http.NewRequestWithContext(ctx, http.MethodGet, ccms.queryEndpoint, buf)
if err != nil {
log.Errorf("Error while building request body: %s", err.Error())
cclog.Errorf("Error while building request body: %s", err.Error())
return nil, err
}
if ccms.jwt != "" {
@@ -151,7 +151,7 @@ func (ccms *CCMetricStore) doRequest(
res, err := ccms.client.Do(req)
if err != nil {
log.Errorf("Error while performing request: %s", err.Error())
cclog.Errorf("Error while performing request: %s", err.Error())
return nil, err
}
@@ -161,7 +161,7 @@ func (ccms *CCMetricStore) doRequest(
var resBody ApiQueryResponse
if err := json.NewDecoder(bufio.NewReader(res.Body)).Decode(&resBody); err != nil {
log.Errorf("Error while decoding result body: %s", err.Error())
cclog.Errorf("Error while decoding result body: %s", err.Error())
return nil, err
}
@@ -177,14 +177,14 @@ func (ccms *CCMetricStore) LoadData(
) (schema.JobData, error) {
queries, assignedScope, err := ccms.buildQueries(job, metrics, scopes, resolution)
if err != nil {
log.Errorf("Error while building queries for jobId %d, Metrics %v, Scopes %v: %s", job.JobID, metrics, scopes, err.Error())
cclog.Errorf("Error while building queries for jobId %d, Metrics %v, Scopes %v: %s", job.JobID, metrics, scopes, err.Error())
return nil, err
}
req := ApiQueryRequest{
Cluster: job.Cluster,
From: job.StartTime.Unix(),
To: job.StartTime.Add(time.Duration(job.Duration) * time.Second).Unix(),
From: job.StartTime,
To: job.StartTime + int64(job.Duration),
Queries: queries,
WithStats: true,
WithData: true,
@@ -192,7 +192,7 @@ func (ccms *CCMetricStore) LoadData(
resBody, err := ccms.doRequest(ctx, &req)
if err != nil {
log.Errorf("Error while performing request: %s", err.Error())
cclog.Errorf("Error while performing request: %s", err.Error())
return nil, err
}
@@ -298,7 +298,7 @@ func (ccms *CCMetricStore) buildQueries(
mc := archive.GetMetricConfig(job.Cluster, metric)
if mc == nil {
// return nil, fmt.Errorf("METRICDATA/CCMS > metric '%s' is not specified for cluster '%s'", metric, job.Cluster)
log.Infof("metric '%s' is not specified for cluster '%s'", metric, job.Cluster)
cclog.Infof("metric '%s' is not specified for cluster '%s'", metric, job.Cluster)
continue
}
@@ -570,17 +570,16 @@ func (ccms *CCMetricStore) LoadStats(
metrics []string,
ctx context.Context,
) (map[string]map[string]schema.MetricStatistics, error) {
queries, _, err := ccms.buildQueries(job, metrics, []schema.MetricScope{schema.MetricScopeNode}, 0) // #166 Add scope shere for analysis view accelerator normalization?
if err != nil {
log.Errorf("Error while building queries for jobId %d, Metrics %v: %s", job.JobID, metrics, err.Error())
cclog.Errorf("Error while building queries for jobId %d, Metrics %v: %s", job.JobID, metrics, err.Error())
return nil, err
}
req := ApiQueryRequest{
Cluster: job.Cluster,
From: job.StartTime.Unix(),
To: job.StartTime.Add(time.Duration(job.Duration) * time.Second).Unix(),
From: job.StartTime,
To: job.StartTime + int64(job.Duration),
Queries: queries,
WithStats: true,
WithData: false,
@@ -588,7 +587,7 @@ func (ccms *CCMetricStore) LoadStats(
resBody, err := ccms.doRequest(ctx, &req)
if err != nil {
log.Errorf("Error while performing request: %s", err.Error())
cclog.Errorf("Error while performing request: %s", err.Error())
return nil, err
}
@@ -598,7 +597,7 @@ func (ccms *CCMetricStore) LoadStats(
metric := ccms.toLocalName(query.Metric)
data := res[0]
if data.Error != nil {
log.Errorf("fetching %s for node %s failed: %s", metric, query.Hostname, *data.Error)
cclog.Errorf("fetching %s for node %s failed: %s", metric, query.Hostname, *data.Error)
continue
}
@@ -609,7 +608,7 @@ func (ccms *CCMetricStore) LoadStats(
}
if data.Avg.IsNaN() || data.Min.IsNaN() || data.Max.IsNaN() {
log.Warnf("fetching %s for node %s failed: one of avg/min/max is NaN", metric, query.Hostname)
cclog.Warnf("fetching %s for node %s failed: one of avg/min/max is NaN", metric, query.Hostname)
continue
}
@@ -632,14 +631,14 @@ func (ccms *CCMetricStore) LoadScopedStats(
) (schema.ScopedJobStats, error) {
queries, assignedScope, err := ccms.buildQueries(job, metrics, scopes, 0)
if err != nil {
log.Errorf("Error while building queries for jobId %d, Metrics %v, Scopes %v: %s", job.JobID, metrics, scopes, err.Error())
cclog.Errorf("Error while building queries for jobId %d, Metrics %v, Scopes %v: %s", job.JobID, metrics, scopes, err.Error())
return nil, err
}
req := ApiQueryRequest{
Cluster: job.Cluster,
From: job.StartTime.Unix(),
To: job.StartTime.Add(time.Duration(job.Duration) * time.Second).Unix(),
From: job.StartTime,
To: job.StartTime + int64(job.Duration),
Queries: queries,
WithStats: true,
WithData: false,
@@ -647,7 +646,7 @@ func (ccms *CCMetricStore) LoadScopedStats(
resBody, err := ccms.doRequest(ctx, &req)
if err != nil {
log.Errorf("Error while performing request: %s", err.Error())
cclog.Errorf("Error while performing request: %s", err.Error())
return nil, err
}
@@ -748,7 +747,7 @@ func (ccms *CCMetricStore) LoadNodeData(
resBody, err := ccms.doRequest(ctx, &req)
if err != nil {
log.Errorf("Error while performing request: %s", err.Error())
cclog.Errorf("Error while performing request: %s", err.Error())
return nil, err
}
@@ -816,7 +815,6 @@ func (ccms *CCMetricStore) LoadNodeListData(
page *model.PageRequest,
ctx context.Context,
) (map[string]schema.JobData, int, bool, error) {
// 0) Init additional vars
var totalNodes int = 0
var hasNextPage bool = false
@@ -852,7 +850,7 @@ func (ccms *CCMetricStore) LoadNodeListData(
if len(nodes) > page.ItemsPerPage {
start := (page.Page - 1) * page.ItemsPerPage
end := start + page.ItemsPerPage
if end > len(nodes) {
if end >= len(nodes) {
end = len(nodes)
hasNextPage = false
} else {
@@ -865,7 +863,7 @@ func (ccms *CCMetricStore) LoadNodeListData(
queries, assignedScope, err := ccms.buildNodeQueries(cluster, subCluster, nodes, metrics, scopes, resolution)
if err != nil {
log.Errorf("Error while building node queries for Cluster %s, SubCLuster %s, Metrics %v, Scopes %v: %s", cluster, subCluster, metrics, scopes, err.Error())
cclog.Errorf("Error while building node queries for Cluster %s, SubCLuster %s, Metrics %v, Scopes %v: %s", cluster, subCluster, metrics, scopes, err.Error())
return nil, totalNodes, hasNextPage, err
}
@@ -880,7 +878,7 @@ func (ccms *CCMetricStore) LoadNodeListData(
resBody, err := ccms.doRequest(ctx, &req)
if err != nil {
log.Errorf("Error while performing request: %s", err.Error())
cclog.Errorf("Error while performing request: %s", err.Error())
return nil, totalNodes, hasNextPage, err
}
@@ -975,7 +973,6 @@ func (ccms *CCMetricStore) buildNodeQueries(
scopes []schema.MetricScope,
resolution int,
) ([]ApiQuery, []schema.MetricScope, error) {
queries := make([]ApiQuery, 0, len(metrics)*len(scopes)*len(nodes))
assignedScope := []schema.MetricScope{}
@@ -985,7 +982,7 @@ func (ccms *CCMetricStore) buildNodeQueries(
if subCluster != "" {
subClusterTopol, scterr = archive.GetSubCluster(cluster, subCluster)
if scterr != nil {
log.Errorf("could not load cluster %s subCluster %s topology: %s", cluster, subCluster, scterr.Error())
cclog.Errorf("could not load cluster %s subCluster %s topology: %s", cluster, subCluster, scterr.Error())
return nil, nil, scterr
}
}
@@ -995,7 +992,7 @@ func (ccms *CCMetricStore) buildNodeQueries(
mc := archive.GetMetricConfig(cluster, metric)
if mc == nil {
// return nil, fmt.Errorf("METRICDATA/CCMS > metric '%s' is not specified for cluster '%s'", metric, cluster)
log.Warnf("metric '%s' is not specified for cluster '%s'", metric, cluster)
cclog.Warnf("metric '%s' is not specified for cluster '%s'", metric, cluster)
continue
}

575
internal/metricdata/influxdb-v2.go
View File

@@ -1,575 +0,0 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package metricdata
import (
"context"
"crypto/tls"
"encoding/json"
"errors"
"fmt"
"math"
"sort"
"strings"
"time"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
influxdb2 "github.com/influxdata/influxdb-client-go/v2"
influxdb2Api "github.com/influxdata/influxdb-client-go/v2/api"
)
type InfluxDBv2DataRepositoryConfig struct {
Url string `json:"url"`
Token string `json:"token"`
Bucket string `json:"bucket"`
Org string `json:"org"`
SkipTls bool `json:"skiptls"`
}
type InfluxDBv2DataRepository struct {
client influxdb2.Client
queryClient influxdb2Api.QueryAPI
bucket, measurement string
}
func (idb *InfluxDBv2DataRepository) Init(rawConfig json.RawMessage) error {
var config InfluxDBv2DataRepositoryConfig
if err := json.Unmarshal(rawConfig, &config); err != nil {
log.Warn("Error while unmarshaling raw json config")
return err
}
idb.client = influxdb2.NewClientWithOptions(config.Url, config.Token, influxdb2.DefaultOptions().SetTLSConfig(&tls.Config{InsecureSkipVerify: config.SkipTls}))
idb.queryClient = idb.client.QueryAPI(config.Org)
idb.bucket = config.Bucket
return nil
}
func (idb *InfluxDBv2DataRepository) formatTime(t time.Time) string {
return t.Format(time.RFC3339) // Like “2006-01-02T15:04:05Z07:00”
}
func (idb *InfluxDBv2DataRepository) epochToTime(epoch int64) time.Time {
return time.Unix(epoch, 0)
}
func (idb *InfluxDBv2DataRepository) LoadData(
job *schema.Job,
metrics []string,
scopes []schema.MetricScope,
ctx context.Context,
resolution int) (schema.JobData, error) {
log.Infof("InfluxDB 2 Backend: Resolution Scaling not Implemented, will return default timestep. Requested Resolution %d", resolution)
measurementsConds := make([]string, 0, len(metrics))
for _, m := range metrics {
measurementsConds = append(measurementsConds, fmt.Sprintf(`r["_measurement"] == "%s"`, m))
}
measurementsCond := strings.Join(measurementsConds, " or ")
hostsConds := make([]string, 0, len(job.Resources))
for _, h := range job.Resources {
if h.HWThreads != nil || h.Accelerators != nil {
// TODO
return nil, errors.New("METRICDATA/INFLUXV2 > the InfluxDB metric data repository does not yet support HWThreads or Accelerators")
}
hostsConds = append(hostsConds, fmt.Sprintf(`r["hostname"] == "%s"`, h.Hostname))
}
hostsCond := strings.Join(hostsConds, " or ")
jobData := make(schema.JobData) // Empty Schema: map[<string>FIELD]map[<MetricScope>SCOPE]<*JobMetric>METRIC
// Requested Scopes
for _, scope := range scopes {
query := ""
switch scope {
case "node":
// Get Finest Granularity, Groupy By Measurement and Hostname (== Metric / Node), Calculate Mean for 60s windows <-- Resolution could be added here?
// log.Info("Scope 'node' requested. ")
query = fmt.Sprintf(`
from(bucket: "%s")
|> range(start: %s, stop: %s)
|> filter(fn: (r) => (%s) and (%s) )
|> drop(columns: ["_start", "_stop"])
|> group(columns: ["hostname", "_measurement"])
|> aggregateWindow(every: 60s, fn: mean)
|> drop(columns: ["_time"])`,
idb.bucket,
idb.formatTime(job.StartTime), idb.formatTime(idb.epochToTime(job.StartTimeUnix+int64(job.Duration)+int64(1))),
measurementsCond, hostsCond)
case "socket":
log.Info("Scope 'socket' requested, but not yet supported: Will return 'node' scope only. ")
continue
case "core":
log.Info(" Scope 'core' requested, but not yet supported: Will return 'node' scope only. ")
continue
// Get Finest Granularity only, Set NULL to 0.0
// query = fmt.Sprintf(`
// from(bucket: "%s")
// |> range(start: %s, stop: %s)
// |> filter(fn: (r) => %s )
// |> filter(fn: (r) => %s )
// |> drop(columns: ["_start", "_stop", "cluster"])
// |> map(fn: (r) => (if exists r._value then {r with _value: r._value} else {r with _value: 0.0}))`,
// idb.bucket,
// idb.formatTime(job.StartTime), idb.formatTime(idb.epochToTime(job.StartTimeUnix + int64(job.Duration) + int64(1) )),
// measurementsCond, hostsCond)
case "hwthread":
log.Info(" Scope 'hwthread' requested, but not yet supported: Will return 'node' scope only. ")
continue
case "accelerator":
log.Info(" Scope 'accelerator' requested, but not yet supported: Will return 'node' scope only. ")
continue
default:
log.Infof("Unknown scope '%s' requested: Will return 'node' scope.", scope)
continue
// return nil, errors.New("METRICDATA/INFLUXV2 > the InfluxDB metric data repository does not yet support other scopes than 'node'")
}
rows, err := idb.queryClient.Query(ctx, query)
if err != nil {
log.Error("Error while performing query")
return nil, err
}
// Init Metrics: Only Node level now -> TODO: Matching /check on scope level ...
for _, metric := range metrics {
jobMetric, ok := jobData[metric]
if !ok {
mc := archive.GetMetricConfig(job.Cluster, metric)
jobMetric = map[schema.MetricScope]*schema.JobMetric{
scope: { // uses scope var from above!
Unit: mc.Unit,
Timestep: mc.Timestep,
Series: make([]schema.Series, 0, len(job.Resources)),
StatisticsSeries: nil, // Should be: &schema.StatsSeries{},
},
}
}
jobData[metric] = jobMetric
}
// Process Result: Time-Data
field, host, hostSeries := "", "", schema.Series{}
// typeId := 0
switch scope {
case "node":
for rows.Next() {
row := rows.Record()
if host == "" || host != row.ValueByKey("hostname").(string) || rows.TableChanged() {
if host != "" {
// Append Series before reset
jobData[field][scope].Series = append(jobData[field][scope].Series, hostSeries)
}
field, host = row.Measurement(), row.ValueByKey("hostname").(string)
hostSeries = schema.Series{
Hostname: host,
Statistics: schema.MetricStatistics{}, //TODO Add Statistics
Data: make([]schema.Float, 0),
}
}
val, ok := row.Value().(float64)
if ok {
hostSeries.Data = append(hostSeries.Data, schema.Float(val))
} else {
hostSeries.Data = append(hostSeries.Data, schema.Float(0))
}
}
case "socket":
continue
case "accelerator":
continue
case "hwthread":
// See below @ core
continue
case "core":
continue
// Include Series.Id in hostSeries
// for rows.Next() {
// row := rows.Record()
// if ( host == "" || host != row.ValueByKey("hostname").(string) || typeId != row.ValueByKey("type-id").(int) || rows.TableChanged() ) {
// if ( host != "" ) {
// // Append Series before reset
// jobData[field][scope].Series = append(jobData[field][scope].Series, hostSeries)
// }
// field, host, typeId = row.Measurement(), row.ValueByKey("hostname").(string), row.ValueByKey("type-id").(int)
// hostSeries = schema.Series{
// Hostname: host,
// Id: &typeId,
// Statistics: nil,
// Data: make([]schema.Float, 0),
// }
// }
// val := row.Value().(float64)
// hostSeries.Data = append(hostSeries.Data, schema.Float(val))
// }
default:
log.Infof("Unknown scope '%s' requested: Will return 'node' scope.", scope)
continue
// return nil, errors.New("the InfluxDB metric data repository does not yet support other scopes than 'node, core'")
}
// Append last Series
jobData[field][scope].Series = append(jobData[field][scope].Series, hostSeries)
}
// Get Stats
stats, err := idb.LoadStats(job, metrics, ctx)
if err != nil {
log.Warn("Error while loading statistics")
return nil, err
}
for _, scope := range scopes {
if scope == "node" { // No 'socket/core' support yet
for metric, nodes := range stats {
for node, stats := range nodes {
for index, _ := range jobData[metric][scope].Series {
if jobData[metric][scope].Series[index].Hostname == node {
jobData[metric][scope].Series[index].Statistics = schema.MetricStatistics{Avg: stats.Avg, Min: stats.Min, Max: stats.Max}
}
}
}
}
}
}
return jobData, nil
}
func (idb *InfluxDBv2DataRepository) LoadStats(
job *schema.Job,
metrics []string,
ctx context.Context) (map[string]map[string]schema.MetricStatistics, error) {
stats := map[string]map[string]schema.MetricStatistics{}
hostsConds := make([]string, 0, len(job.Resources))
for _, h := range job.Resources {
if h.HWThreads != nil || h.Accelerators != nil {
// TODO
return nil, errors.New("METRICDATA/INFLUXV2 > the InfluxDB metric data repository does not yet support HWThreads or Accelerators")
}
hostsConds = append(hostsConds, fmt.Sprintf(`r["hostname"] == "%s"`, h.Hostname))
}
hostsCond := strings.Join(hostsConds, " or ")
// lenMet := len(metrics)
for _, metric := range metrics {
// log.Debugf("<< You are here: %s (Index %d of %d metrics)", metric, index, lenMet)
query := fmt.Sprintf(`
data = from(bucket: "%s")
|> range(start: %s, stop: %s)
|> filter(fn: (r) => r._measurement == "%s" and r._field == "value" and (%s))
union(tables: [data |> mean(column: "_value") |> set(key: "_field", value: "avg"),
data |> min(column: "_value") |> set(key: "_field", value: "min"),
data |> max(column: "_value") |> set(key: "_field", value: "max")])
|> pivot(rowKey: ["hostname"], columnKey: ["_field"], valueColumn: "_value")
|> group()`,
idb.bucket,
idb.formatTime(job.StartTime), idb.formatTime(idb.epochToTime(job.StartTimeUnix+int64(job.Duration)+int64(1))),
metric, hostsCond)
rows, err := idb.queryClient.Query(ctx, query)
if err != nil {
log.Error("Error while performing query")
return nil, err
}
nodes := map[string]schema.MetricStatistics{}
for rows.Next() {
row := rows.Record()
host := row.ValueByKey("hostname").(string)
avg, avgok := row.ValueByKey("avg").(float64)
if !avgok {
// log.Debugf(">> Assertion error for metric %s, statistic AVG. Expected 'float64', got %v", metric, avg)
avg = 0.0
}
min, minok := row.ValueByKey("min").(float64)
if !minok {
// log.Debugf(">> Assertion error for metric %s, statistic MIN. Expected 'float64', got %v", metric, min)
min = 0.0
}
max, maxok := row.ValueByKey("max").(float64)
if !maxok {
// log.Debugf(">> Assertion error for metric %s, statistic MAX. Expected 'float64', got %v", metric, max)
max = 0.0
}
nodes[host] = schema.MetricStatistics{
Avg: avg,
Min: min,
Max: max,
}
}
stats[metric] = nodes
}
return stats, nil
}
// Used in Job-View StatsTable
// UNTESTED
func (idb *InfluxDBv2DataRepository) LoadScopedStats(
job *schema.Job,
metrics []string,
scopes []schema.MetricScope,
ctx context.Context) (schema.ScopedJobStats, error) {
// Assumption: idb.loadData() only returns series node-scope - use node scope for statsTable
scopedJobStats := make(schema.ScopedJobStats)
data, err := idb.LoadData(job, metrics, []schema.MetricScope{schema.MetricScopeNode}, ctx, 0 /*resolution here*/)
if err != nil {
log.Warn("Error while loading job for scopedJobStats")
return nil, err
}
for metric, metricData := range data {
for _, scope := range scopes {
if scope != schema.MetricScopeNode {
logOnce.Do(func() {
log.Infof("Note: Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
})
continue
}
if _, ok := scopedJobStats[metric]; !ok {
scopedJobStats[metric] = make(map[schema.MetricScope][]*schema.ScopedStats)
}
if _, ok := scopedJobStats[metric][scope]; !ok {
scopedJobStats[metric][scope] = make([]*schema.ScopedStats, 0)
}
for _, series := range metricData[scope].Series {
scopedJobStats[metric][scope] = append(scopedJobStats[metric][scope], &schema.ScopedStats{
Hostname: series.Hostname,
Data: &series.Statistics,
})
}
}
}
return scopedJobStats, nil
}
// Used in Systems-View @ Node-Overview
// UNTESTED
func (idb *InfluxDBv2DataRepository) LoadNodeData(
cluster string,
metrics, nodes []string,
scopes []schema.MetricScope,
from, to time.Time,
ctx context.Context) (map[string]map[string][]*schema.JobMetric, error) {
// Note: scopes[] Array will be ignored, only return node scope
// CONVERT ARGS TO INFLUX
measurementsConds := make([]string, 0)
for _, m := range metrics {
measurementsConds = append(measurementsConds, fmt.Sprintf(`r["_measurement"] == "%s"`, m))
}
measurementsCond := strings.Join(measurementsConds, " or ")
hostsConds := make([]string, 0)
if nodes == nil {
var allNodes []string
subClusterNodeLists := archive.NodeLists[cluster]
for _, nodeList := range subClusterNodeLists {
allNodes = append(nodes, nodeList.PrintList()...)
}
for _, node := range allNodes {
nodes = append(nodes, node)
hostsConds = append(hostsConds, fmt.Sprintf(`r["hostname"] == "%s"`, node))
}
} else {
for _, node := range nodes {
hostsConds = append(hostsConds, fmt.Sprintf(`r["hostname"] == "%s"`, node))
}
}
hostsCond := strings.Join(hostsConds, " or ")
// BUILD AND PERFORM QUERY
query := fmt.Sprintf(`
from(bucket: "%s")
|> range(start: %s, stop: %s)
|> filter(fn: (r) => (%s) and (%s) )
|> drop(columns: ["_start", "_stop"])
|> group(columns: ["hostname", "_measurement"])
|> aggregateWindow(every: 60s, fn: mean)
|> drop(columns: ["_time"])`,
idb.bucket,
idb.formatTime(from), idb.formatTime(to),
measurementsCond, hostsCond)
rows, err := idb.queryClient.Query(ctx, query)
if err != nil {
log.Error("Error while performing query")
return nil, err
}
// HANDLE QUERY RETURN
// Collect Float Arrays for Node@Metric -> No Scope Handling!
influxData := make(map[string]map[string][]schema.Float)
for rows.Next() {
row := rows.Record()
host, field := row.ValueByKey("hostname").(string), row.Measurement()
influxHostData, ok := influxData[host]
if !ok {
influxHostData = make(map[string][]schema.Float)
influxData[host] = influxHostData
}
influxFieldData, ok := influxData[host][field]
if !ok {
influxFieldData = make([]schema.Float, 0)
influxData[host][field] = influxFieldData
}
val, ok := row.Value().(float64)
if ok {
influxData[host][field] = append(influxData[host][field], schema.Float(val))
} else {
influxData[host][field] = append(influxData[host][field], schema.Float(0))
}
}
// BUILD FUNCTION RETURN
data := make(map[string]map[string][]*schema.JobMetric)
for node, metricData := range influxData {
nodeData, ok := data[node]
if !ok {
nodeData = make(map[string][]*schema.JobMetric)
data[node] = nodeData
}
for metric, floatArray := range metricData {
avg, min, max := 0.0, 0.0, 0.0
for _, val := range floatArray {
avg += float64(val)
min = math.Min(min, float64(val))
max = math.Max(max, float64(val))
}
stats := schema.MetricStatistics{
Avg: (math.Round((avg/float64(len(floatArray)))*100) / 100),
Min: (math.Round(min*100) / 100),
Max: (math.Round(max*100) / 100),
}
mc := archive.GetMetricConfig(cluster, metric)
nodeData[metric] = append(nodeData[metric], &schema.JobMetric{
Unit: mc.Unit,
Timestep: mc.Timestep,
Series: []schema.Series{
{
Hostname: node,
Statistics: stats,
Data: floatArray,
},
},
})
}
}
return data, nil
}
// Used in Systems-View @ Node-List
// UNTESTED
func (idb *InfluxDBv2DataRepository) LoadNodeListData(
cluster, subCluster, nodeFilter string,
metrics []string,
scopes []schema.MetricScope,
resolution int,
from, to time.Time,
page *model.PageRequest,
ctx context.Context,
) (map[string]schema.JobData, int, bool, error) {
// Assumption: idb.loadData() only returns series node-scope - use node scope for NodeList
// 0) Init additional vars
var totalNodes int = 0
var hasNextPage bool = false
// 1) Get list of all nodes
var nodes []string
if subCluster != "" {
scNodes := archive.NodeLists[cluster][subCluster]
nodes = scNodes.PrintList()
} else {
subClusterNodeLists := archive.NodeLists[cluster]
for _, nodeList := range subClusterNodeLists {
nodes = append(nodes, nodeList.PrintList()...)
}
}
// 2) Filter nodes
if nodeFilter != "" {
filteredNodes := []string{}
for _, node := range nodes {
if strings.Contains(node, nodeFilter) {
filteredNodes = append(filteredNodes, node)
}
}
nodes = filteredNodes
}
// 2.1) Count total nodes && Sort nodes -> Sorting invalidated after return ...
totalNodes = len(nodes)
sort.Strings(nodes)
// 3) Apply paging
if len(nodes) > page.ItemsPerPage {
start := (page.Page - 1) * page.ItemsPerPage
end := start + page.ItemsPerPage
if end > len(nodes) {
end = len(nodes)
hasNextPage = false
} else {
hasNextPage = true
}
nodes = nodes[start:end]
}
// 4) Fetch And Convert Data, use idb.LoadNodeData() for query
rawNodeData, err := idb.LoadNodeData(cluster, metrics, nodes, scopes, from, to, ctx)
if err != nil {
log.Error(fmt.Sprintf("Error while loading influx nodeData for nodeListData %#v\n", err))
return nil, totalNodes, hasNextPage, err
}
data := make(map[string]schema.JobData)
for node, nodeData := range rawNodeData {
// Init Nested Map Data Structures If Not Found
hostData, ok := data[node]
if !ok {
hostData = make(schema.JobData)
data[node] = hostData
}
for metric, nodeMetricData := range nodeData {
metricData, ok := hostData[metric]
if !ok {
metricData = make(map[schema.MetricScope]*schema.JobMetric)
data[node][metric] = metricData
}
data[node][metric][schema.MetricScopeNode] = nodeMetricData[0] // Only Node Scope Returned from loadNodeData
}
}
return data, totalNodes, hasNextPage, nil
}

12
internal/metricdata/metricdata.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package metricdata
@@ -12,8 +12,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
type MetricDataRepository interface {
@@ -46,7 +46,7 @@ func Init() error {
Kind string `json:"kind"`
}
if err := json.Unmarshal(cluster.MetricDataRepository, &kind); err != nil {
log.Warn("Error while unmarshaling raw json MetricDataRepository")
cclog.Warn("Error while unmarshaling raw json MetricDataRepository")
return err
}
@@ -54,8 +54,6 @@ func Init() error {
switch kind.Kind {
case "cc-metric-store":
mdr = &CCMetricStore{}
case "influxdb":
mdr = &InfluxDBv2DataRepository{}
case "prometheus":
mdr = &PrometheusDataRepository{}
case "test":
@@ -65,7 +63,7 @@ func Init() error {
}
if err := mdr.Init(cluster.MetricDataRepository); err != nil {
log.Errorf("Error initializing MetricDataRepository %v for cluster %v", kind.Kind, cluster.Name)
cclog.Errorf("Error initializing MetricDataRepository %v for cluster %v", kind.Kind, cluster.Name)
return err
}
metricDataRepos[cluster.Name] = mdr

67
internal/metricdata/prometheus.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) 2022 DKRZ
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package metricdata
@@ -22,8 +22,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
promapi "github.com/prometheus/client_golang/api"
promv1 "github.com/prometheus/client_golang/api/prometheus/v1"
promcfg "github.com/prometheus/common/config"
@@ -160,7 +160,7 @@ func (pdb *PrometheusDataRepository) Init(rawConfig json.RawMessage) error {
var config PrometheusDataRepositoryConfig
// parse config
if err := json.Unmarshal(rawConfig, &config); err != nil {
log.Warn("Error while unmarshaling raw json config")
cclog.Warn("Error while unmarshaling raw json config")
return err
}
// support basic authentication
@@ -179,7 +179,7 @@ func (pdb *PrometheusDataRepository) Init(rawConfig json.RawMessage) error {
RoundTripper: rt,
})
if err != nil {
log.Error("Error while initializing new prometheus client")
cclog.Error("Error while initializing new prometheus client")
return err
}
// init query client
@@ -192,9 +192,9 @@ func (pdb *PrometheusDataRepository) Init(rawConfig json.RawMessage) error {
for metric, templ := range config.Templates {
pdb.templates[metric], err = template.New(metric).Parse(templ)
if err == nil {
log.Debugf("Added PromQL template for %s: %s", metric, templ)
cclog.Debugf("Added PromQL template for %s: %s", metric, templ)
} else {
log.Warnf("Failed to parse PromQL template %s for metric %s", templ, metric)
cclog.Warnf("Failed to parse PromQL template %s for metric %s", templ, metric)
}
}
return nil
@@ -221,7 +221,7 @@ func (pdb *PrometheusDataRepository) FormatQuery(
return "", errors.New(fmt.Sprintf("METRICDATA/PROMETHEUS > Error compiling template %v", templ))
} else {
query := buf.String()
log.Debugf("PromQL: %s", query)
cclog.Debugf("PromQL: %s", query)
return query, nil
}
} else {
@@ -279,13 +279,13 @@ func (pdb *PrometheusDataRepository) LoadData(
for i, resource := range job.Resources {
nodes[i] = resource.Hostname
}
from := job.StartTime
to := job.StartTime.Add(time.Duration(job.Duration) * time.Second)
from := time.Unix(job.StartTime, 0)
to := time.Unix(job.StartTime+int64(job.Duration), 0)
for _, scope := range scopes {
if scope != schema.MetricScopeNode {
logOnce.Do(func() {
log.Infof("Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
cclog.Infof("Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
})
continue
}
@@ -293,12 +293,12 @@ func (pdb *PrometheusDataRepository) LoadData(
for _, metric := range metrics {
metricConfig := archive.GetMetricConfig(job.Cluster, metric)
if metricConfig == nil {
log.Warnf("Error in LoadData: Metric %s for cluster %s not configured", metric, job.Cluster)
cclog.Warnf("Error in LoadData: Metric %s for cluster %s not configured", metric, job.Cluster)
return nil, errors.New("Prometheus config error")
}
query, err := pdb.FormatQuery(metric, scope, nodes, job.Cluster)
if err != nil {
log.Warn("Error while formatting prometheus query")
cclog.Warn("Error while formatting prometheus query")
return nil, err
}
@@ -310,11 +310,11 @@ func (pdb *PrometheusDataRepository) LoadData(
}
result, warnings, err := pdb.queryClient.QueryRange(ctx, query, r)
if err != nil {
log.Errorf("Prometheus query error in LoadData: %v\nQuery: %s", err, query)
cclog.Errorf("Prometheus query error in LoadData: %v\nQuery: %s", err, query)
return nil, errors.New("Prometheus query error")
}
if len(warnings) > 0 {
log.Warnf("Warnings: %v\n", warnings)
cclog.Warnf("Warnings: %v\n", warnings)
}
// init data structures
@@ -360,7 +360,7 @@ func (pdb *PrometheusDataRepository) LoadStats(
data, err := pdb.LoadData(job, metrics, []schema.MetricScope{schema.MetricScopeNode}, ctx, 0 /*resolution here*/)
if err != nil {
log.Warn("Error while loading job for stats")
cclog.Warn("Error while loading job for stats")
return nil, err
}
for metric, metricData := range data {
@@ -391,19 +391,19 @@ func (pdb *PrometheusDataRepository) LoadNodeData(
for _, scope := range scopes {
if scope != schema.MetricScopeNode {
logOnce.Do(func() {
log.Infof("Note: Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
cclog.Infof("Note: Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
})
continue
}
for _, metric := range metrics {
metricConfig := archive.GetMetricConfig(cluster, metric)
if metricConfig == nil {
log.Warnf("Error in LoadNodeData: Metric %s for cluster %s not configured", metric, cluster)
cclog.Warnf("Error in LoadNodeData: Metric %s for cluster %s not configured", metric, cluster)
return nil, errors.New("Prometheus config error")
}
query, err := pdb.FormatQuery(metric, scope, nodes, cluster)
if err != nil {
log.Warn("Error while formatting prometheus query")
cclog.Warn("Error while formatting prometheus query")
return nil, err
}
@@ -415,11 +415,11 @@ func (pdb *PrometheusDataRepository) LoadNodeData(
}
result, warnings, err := pdb.queryClient.QueryRange(ctx, query, r)
if err != nil {
log.Errorf("Prometheus query error in LoadNodeData: %v\n", err)
cclog.Errorf("Prometheus query error in LoadNodeData: %v\n", err)
return nil, errors.New("Prometheus query error")
}
if len(warnings) > 0 {
log.Warnf("Warnings: %v\n", warnings)
cclog.Warnf("Warnings: %v\n", warnings)
}
step := int64(metricConfig.Timestep)
@@ -444,7 +444,7 @@ func (pdb *PrometheusDataRepository) LoadNodeData(
}
}
t1 := time.Since(t0)
log.Debugf("LoadNodeData of %v nodes took %s", len(data), t1)
cclog.Debugf("LoadNodeData of %v nodes took %s", len(data), t1)
return data, nil
}
@@ -453,13 +453,13 @@ func (pdb *PrometheusDataRepository) LoadScopedStats(
job *schema.Job,
metrics []string,
scopes []schema.MetricScope,
ctx context.Context) (schema.ScopedJobStats, error) {
ctx context.Context,
) (schema.ScopedJobStats, error) {
// Assumption: pdb.loadData() only returns series node-scope - use node scope for statsTable
scopedJobStats := make(schema.ScopedJobStats)
data, err := pdb.LoadData(job, metrics, []schema.MetricScope{schema.MetricScopeNode}, ctx, 0 /*resolution here*/)
if err != nil {
log.Warn("Error while loading job for scopedJobStats")
cclog.Warn("Error while loading job for scopedJobStats")
return nil, err
}
@@ -467,7 +467,7 @@ func (pdb *PrometheusDataRepository) LoadScopedStats(
for _, scope := range scopes {
if scope != schema.MetricScopeNode {
logOnce.Do(func() {
log.Infof("Note: Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
cclog.Infof("Note: Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
})
continue
}
@@ -502,7 +502,6 @@ func (pdb *PrometheusDataRepository) LoadNodeListData(
page *model.PageRequest,
ctx context.Context,
) (map[string]schema.JobData, int, bool, error) {
// Assumption: pdb.loadData() only returns series node-scope - use node scope for NodeList
// 0) Init additional vars
@@ -540,7 +539,7 @@ func (pdb *PrometheusDataRepository) LoadNodeListData(
if len(nodes) > page.ItemsPerPage {
start := (page.Page - 1) * page.ItemsPerPage
end := start + page.ItemsPerPage
if end > len(nodes) {
if end >= len(nodes) {
end = len(nodes)
hasNextPage = false
} else {
@@ -564,7 +563,7 @@ func (pdb *PrometheusDataRepository) LoadNodeListData(
for _, scope := range scopes {
if scope != schema.MetricScopeNode {
logOnce.Do(func() {
log.Infof("Note: Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
cclog.Infof("Note: Scope '%s' requested, but not yet supported: Will return 'node' scope only.", scope)
})
continue
}
@@ -572,12 +571,12 @@ func (pdb *PrometheusDataRepository) LoadNodeListData(
for _, metric := range metrics {
metricConfig := archive.GetMetricConfig(cluster, metric)
if metricConfig == nil {
log.Warnf("Error in LoadNodeListData: Metric %s for cluster %s not configured", metric, cluster)
cclog.Warnf("Error in LoadNodeListData: Metric %s for cluster %s not configured", metric, cluster)
return nil, totalNodes, hasNextPage, errors.New("Prometheus config error")
}
query, err := pdb.FormatQuery(metric, scope, nodes, cluster)
if err != nil {
log.Warn("Error while formatting prometheus query")
cclog.Warn("Error while formatting prometheus query")
return nil, totalNodes, hasNextPage, err
}
@@ -589,11 +588,11 @@ func (pdb *PrometheusDataRepository) LoadNodeListData(
}
result, warnings, err := pdb.queryClient.QueryRange(ctx, query, r)
if err != nil {
log.Errorf("Prometheus query error in LoadNodeData: %v\n", err)
cclog.Errorf("Prometheus query error in LoadNodeData: %v\n", err)
return nil, totalNodes, hasNextPage, errors.New("Prometheus query error")
}
if len(warnings) > 0 {
log.Warnf("Warnings: %v\n", warnings)
cclog.Warnf("Warnings: %v\n", warnings)
}
step := int64(metricConfig.Timestep)
@@ -629,6 +628,6 @@ func (pdb *PrometheusDataRepository) LoadNodeListData(
}
}
t1 := time.Since(t0)
log.Debugf("LoadNodeListData of %v nodes took %s", len(data), t1)
cclog.Debugf("LoadNodeListData of %v nodes took %s", len(data), t1)
return data, totalNodes, hasNextPage, nil
}

21
internal/metricdata/utils.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package metricdata
@@ -10,7 +10,7 @@ import (
"time"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
"github.com/ClusterCockpit/cc-lib/schema"
)
var TestLoadDataCallback func(job *schema.Job, metrics []string, scopes []schema.MetricScope, ctx context.Context, resolution int) (schema.JobData, error) = func(job *schema.Job, metrics []string, scopes []schema.MetricScope, ctx context.Context, resolution int) (schema.JobData, error) {
@@ -29,16 +29,16 @@ func (tmdr *TestMetricDataRepository) LoadData(
metrics []string,
scopes []schema.MetricScope,
ctx context.Context,
resolution int) (schema.JobData, error) {
resolution int,
) (schema.JobData, error) {
return TestLoadDataCallback(job, metrics, scopes, ctx, resolution)
}
func (tmdr *TestMetricDataRepository) LoadStats(
job *schema.Job,
metrics []string,
ctx context.Context) (map[string]map[string]schema.MetricStatistics, error) {
ctx context.Context,
) (map[string]map[string]schema.MetricStatistics, error) {
panic("TODO")
}
@@ -46,8 +46,8 @@ func (tmdr *TestMetricDataRepository) LoadScopedStats(
job *schema.Job,
metrics []string,
scopes []schema.MetricScope,
ctx context.Context) (schema.ScopedJobStats, error) {
ctx context.Context,
) (schema.ScopedJobStats, error) {
panic("TODO")
}
@@ -56,8 +56,8 @@ func (tmdr *TestMetricDataRepository) LoadNodeData(
metrics, nodes []string,
scopes []schema.MetricScope,
from, to time.Time,
ctx context.Context) (map[string]map[string][]*schema.JobMetric, error) {
ctx context.Context,
) (map[string]map[string][]*schema.JobMetric, error) {
panic("TODO")
}
@@ -70,7 +70,6 @@ func (tmdr *TestMetricDataRepository) LoadNodeListData(
page *model.PageRequest,
ctx context.Context,
) (map[string]schema.JobData, int, bool, error) {
panic("TODO")
}

53
internal/repository/dbConnection.go
View File

@@ -1,15 +1,17 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
import (
"database/sql"
"fmt"
"net/url"
"sync"
"time"
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/jmoiron/sqlx"
"github.com/mattn/go-sqlite3"
"github.com/qustavo/sqlhooks/v2"
@@ -33,6 +35,27 @@ type DatabaseOptions struct {
ConnectionMaxIdleTime time.Duration
}
func setupSqlite(db *sql.DB) (err error) {
pragmas := []string{
// "journal_mode = WAL",
// "busy_timeout = 5000",
// "synchronous = NORMAL",
// "cache_size = 1000000000", // 1GB
// "foreign_keys = true",
"temp_store = memory",
// "mmap_size = 3000000000",
}
for _, pragma := range pragmas {
_, err = db.Exec("PRAGMA " + pragma)
if err != nil {
return
}
}
return nil
}
func Connect(driver string, db string) {
var err error
var dbHandle *sqlx.DB
@@ -48,26 +71,34 @@ func Connect(driver string, db string) {
switch driver {
case "sqlite3":
// - Set WAL mode (not strictly necessary each time because it's persisted in the database, but good for first run)
// - Set busy timeout, so concurrent writers wait on each other instead of erroring immediately
// - Enable foreign key checks
opts.URL += "?_journal=WAL&_timeout=5000&_fk=true"
// TODO: Have separate DB handles for Writes and Reads
// Optimize SQLite connection: https://kerkour.com/sqlite-for-servers
connectionUrlParams := make(url.Values)
connectionUrlParams.Add("_txlock", "immediate")
connectionUrlParams.Add("_journal_mode", "WAL")
connectionUrlParams.Add("_busy_timeout", "5000")
connectionUrlParams.Add("_synchronous", "NORMAL")
connectionUrlParams.Add("_cache_size", "1000000000")
connectionUrlParams.Add("_foreign_keys", "true")
opts.URL = fmt.Sprintf("file:%s?%s", opts.URL, connectionUrlParams.Encode())
if log.Loglevel() == "debug" {
if cclog.Loglevel() == "debug" {
sql.Register("sqlite3WithHooks", sqlhooks.Wrap(&sqlite3.SQLiteDriver{}, &Hooks{}))
dbHandle, err = sqlx.Open("sqlite3WithHooks", opts.URL)
} else {
dbHandle, err = sqlx.Open("sqlite3", opts.URL)
}
setupSqlite(dbHandle.DB)
case "mysql":
opts.URL += "?multiStatements=true"
dbHandle, err = sqlx.Open("mysql", opts.URL)
default:
log.Abortf("DB Connection: Unsupported database driver '%s'.\n", driver)
cclog.Abortf("DB Connection: Unsupported database driver '%s'.\n", driver)
}
if err != nil {
log.Abortf("DB Connection: Could not connect to '%s' database with sqlx.Open().\nError: %s\n", driver, err.Error())
cclog.Abortf("DB Connection: Could not connect to '%s' database with sqlx.Open().\nError: %s\n", driver, err.Error())
}
dbHandle.SetMaxOpenConns(opts.MaxOpenConnections)
@@ -78,14 +109,14 @@ func Connect(driver string, db string) {
dbConnInstance = &DBConnection{DB: dbHandle, Driver: driver}
err = checkDBVersion(driver, dbHandle.DB)
if err != nil {
log.Abortf("DB Connection: Failed DB version check.\nError: %s\n", err.Error())
cclog.Abortf("DB Connection: Failed DB version check.\nError: %s\n", err.Error())
}
})
}
func GetConnection() *DBConnection {
if dbConnInstance == nil {
log.Fatalf("Database connection not initialized!")
cclog.Fatalf("Database connection not initialized!")
}
return dbConnInstance

12
internal/repository/hooks.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -8,21 +8,21 @@ import (
"context"
"time"
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
)
// Hooks satisfies the sqlhook.Hooks interface
type Hooks struct{}
// Before hook will print the query with it's args and return the context with the timestamp
func (h *Hooks) Before(ctx context.Context, query string, args ...interface{}) (context.Context, error) {
log.Debugf("SQL query %s %q", query, args)
func (h *Hooks) Before(ctx context.Context, query string, args ...any) (context.Context, error) {
cclog.Debugf("SQL query %s %q", query, args)
return context.WithValue(ctx, "begin", time.Now()), nil
}
// After hook will get the timestamp registered on the Before hook and print the elapsed time
func (h *Hooks) After(ctx context.Context, query string, args ...interface{}) (context.Context, error) {
func (h *Hooks) After(ctx context.Context, query string, args ...any) (context.Context, error) {
begin := ctx.Value("begin").(time.Time)
log.Debugf("Took: %s\n", time.Since(begin))
cclog.Debugf("Took: %s\n", time.Since(begin))
return ctx, nil
}

182
internal/repository/job.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -9,16 +9,16 @@ import (
"encoding/json"
"errors"
"fmt"
"maps"
"math"
"strconv"
"sync"
"time"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/lrucache"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/lrucache"
"github.com/ClusterCockpit/cc-lib/schema"
sq "github.com/Masterminds/squirrel"
"github.com/jmoiron/sqlx"
)
@@ -33,6 +33,7 @@ type JobRepository struct {
stmtCache *sq.StmtCache
cache *lrucache.Cache
driver string
Mutex sync.Mutex
}
func GetJobRepository() *JobRepository {
@@ -51,38 +52,49 @@ func GetJobRepository() *JobRepository {
}
var jobColumns []string = []string{
"job.id", "job.job_id", "job.hpc_user", "job.project", "job.cluster", "job.subcluster", "job.start_time", "job.cluster_partition", "job.array_job_id",
"job.num_nodes", "job.num_hwthreads", "job.num_acc", "job.exclusive", "job.monitoring_status", "job.smt", "job.job_state",
"job.duration", "job.walltime", "job.resources", "job.footprint", "job.energy",
"job.id", "job.job_id", "job.hpc_user", "job.project", "job.cluster", "job.subcluster",
"job.start_time", "job.cluster_partition", "job.array_job_id", "job.num_nodes",
"job.num_hwthreads", "job.num_acc", "job.exclusive", "job.monitoring_status",
"job.smt", "job.job_state", "job.duration", "job.walltime", "job.resources",
"job.footprint", "job.energy",
}
func scanJob(row interface{ Scan(...interface{}) error }) (*schema.Job, error) {
var jobCacheColumns []string = []string{
"job_cache.id", "job_cache.job_id", "job_cache.hpc_user", "job_cache.project", "job_cache.cluster",
"job_cache.subcluster", "job_cache.start_time", "job_cache.cluster_partition",
"job_cache.array_job_id", "job_cache.num_nodes", "job_cache.num_hwthreads",
"job_cache.num_acc", "job_cache.exclusive", "job_cache.monitoring_status", "job_cache.smt",
"job_cache.job_state", "job_cache.duration", "job_cache.walltime", "job_cache.resources",
"job_cache.footprint", "job_cache.energy",
}
func scanJob(row interface{ Scan(...any) error }) (*schema.Job, error) {
job := &schema.Job{}
if err := row.Scan(
&job.ID, &job.JobID, &job.User, &job.Project, &job.Cluster, &job.SubCluster, &job.StartTimeUnix, &job.Partition, &job.ArrayJobId,
&job.NumNodes, &job.NumHWThreads, &job.NumAcc, &job.Exclusive, &job.MonitoringStatus, &job.SMT, &job.State,
&job.ID, &job.JobID, &job.User, &job.Project, &job.Cluster, &job.SubCluster,
&job.StartTime, &job.Partition, &job.ArrayJobId, &job.NumNodes, &job.NumHWThreads,
&job.NumAcc, &job.Exclusive, &job.MonitoringStatus, &job.SMT, &job.State,
&job.Duration, &job.Walltime, &job.RawResources, &job.RawFootprint, &job.Energy); err != nil {
log.Warnf("Error while scanning rows (Job): %v", err)
cclog.Warnf("Error while scanning rows (Job): %v", err)
return nil, err
}
if err := json.Unmarshal(job.RawResources, &job.Resources); err != nil {
log.Warn("Error while unmarshaling raw resources json")
cclog.Warn("Error while unmarshaling raw resources json")
return nil, err
}
job.RawResources = nil
if err := json.Unmarshal(job.RawFootprint, &job.Footprint); err != nil {
log.Warnf("Error while unmarshaling raw footprint json: %v", err)
cclog.Warnf("Error while unmarshaling raw footprint json: %v", err)
return nil, err
}
job.RawFootprint = nil
job.StartTime = time.Unix(job.StartTimeUnix, 0)
// Always ensure accurate duration for running jobs
if job.State == schema.JobStateRunning {
job.Duration = int32(time.Since(job.StartTime).Seconds())
job.Duration = int32(time.Now().Unix() - job.StartTime)
}
return job, nil
@@ -97,7 +109,7 @@ func (r *JobRepository) Optimize() error {
return err
}
case "mysql":
log.Info("Optimize currently not supported for mysql driver")
cclog.Info("Optimize currently not supported for mysql driver")
}
return nil
@@ -138,17 +150,6 @@ func (r *JobRepository) Flush() error {
return nil
}
func scanJobLink(row interface{ Scan(...interface{}) error }) (*model.JobLink, error) {
jobLink := &model.JobLink{}
if err := row.Scan(
&jobLink.ID, &jobLink.JobID); err != nil {
log.Warn("Error while scanning rows (jobLink)")
return nil, err
}
return jobLink, nil
}
func (r *JobRepository) FetchMetadata(job *schema.Job) (map[string]string, error) {
start := time.Now()
cachekey := fmt.Sprintf("metadata:%d", job.ID)
@@ -159,7 +160,7 @@ func (r *JobRepository) FetchMetadata(job *schema.Job) (map[string]string, error
if err := sq.Select("job.meta_data").From("job").Where("job.id = ?", job.ID).
RunWith(r.stmtCache).QueryRow().Scan(&job.RawMetaData); err != nil {
log.Warn("Error while scanning for job metadata")
cclog.Warn("Error while scanning for job metadata")
return nil, err
}
@@ -168,12 +169,12 @@ func (r *JobRepository) FetchMetadata(job *schema.Job) (map[string]string, error
}
if err := json.Unmarshal(job.RawMetaData, &job.MetaData); err != nil {
log.Warn("Error while unmarshaling raw metadata json")
cclog.Warn("Error while unmarshaling raw metadata json")
return nil, err
}
r.cache.Put(cachekey, job.MetaData, len(job.RawMetaData), 24*time.Hour)
log.Debugf("Timer FetchMetadata %s", time.Since(start))
cclog.Debugf("Timer FetchMetadata %s", time.Since(start))
return job.MetaData, nil
}
@@ -182,16 +183,14 @@ func (r *JobRepository) UpdateMetadata(job *schema.Job, key, val string) (err er
r.cache.Del(cachekey)
if job.MetaData == nil {
if _, err = r.FetchMetadata(job); err != nil {
log.Warnf("Error while fetching metadata for job, DB ID '%v'", job.ID)
cclog.Warnf("Error while fetching metadata for job, DB ID '%v'", job.ID)
return err
}
}
if job.MetaData != nil {
cpy := make(map[string]string, len(job.MetaData)+1)
for k, v := range job.MetaData {
cpy[k] = v
}
maps.Copy(cpy, job.MetaData)
cpy[key] = val
job.MetaData = cpy
} else {
@@ -199,7 +198,7 @@ func (r *JobRepository) UpdateMetadata(job *schema.Job, key, val string) (err er
}
if job.RawMetaData, err = json.Marshal(job.MetaData); err != nil {
log.Warnf("Error while marshaling metadata for job, DB ID '%v'", job.ID)
cclog.Warnf("Error while marshaling metadata for job, DB ID '%v'", job.ID)
return err
}
@@ -207,7 +206,7 @@ func (r *JobRepository) UpdateMetadata(job *schema.Job, key, val string) (err er
Set("meta_data", job.RawMetaData).
Where("job.id = ?", job.ID).
RunWith(r.stmtCache).Exec(); err != nil {
log.Warnf("Error while updating metadata for job, DB ID '%v'", job.ID)
cclog.Warnf("Error while updating metadata for job, DB ID '%v'", job.ID)
return err
}
@@ -220,7 +219,7 @@ func (r *JobRepository) FetchFootprint(job *schema.Job) (map[string]float64, err
if err := sq.Select("job.footprint").From("job").Where("job.id = ?", job.ID).
RunWith(r.stmtCache).QueryRow().Scan(&job.RawFootprint); err != nil {
log.Warn("Error while scanning for job footprint")
cclog.Warn("Error while scanning for job footprint")
return nil, err
}
@@ -229,11 +228,11 @@ func (r *JobRepository) FetchFootprint(job *schema.Job) (map[string]float64, err
}
if err := json.Unmarshal(job.RawFootprint, &job.Footprint); err != nil {
log.Warn("Error while unmarshaling raw footprint json")
cclog.Warn("Error while unmarshaling raw footprint json")
return nil, err
}
log.Debugf("Timer FetchFootprint %s", time.Since(start))
cclog.Debugf("Timer FetchFootprint %s", time.Since(start))
return job.Footprint, nil
}
@@ -247,7 +246,7 @@ func (r *JobRepository) FetchEnergyFootprint(job *schema.Job) (map[string]float6
if err := sq.Select("job.energy_footprint").From("job").Where("job.id = ?", job.ID).
RunWith(r.stmtCache).QueryRow().Scan(&job.RawEnergyFootprint); err != nil {
log.Warn("Error while scanning for job energy_footprint")
cclog.Warn("Error while scanning for job energy_footprint")
return nil, err
}
@@ -256,12 +255,12 @@ func (r *JobRepository) FetchEnergyFootprint(job *schema.Job) (map[string]float6
}
if err := json.Unmarshal(job.RawEnergyFootprint, &job.EnergyFootprint); err != nil {
log.Warn("Error while unmarshaling raw energy footprint json")
cclog.Warn("Error while unmarshaling raw energy footprint json")
return nil, err
}
r.cache.Put(cachekey, job.EnergyFootprint, len(job.EnergyFootprint), 24*time.Hour)
log.Debugf("Timer FetchEnergyFootprint %s", time.Since(start))
cclog.Debugf("Timer FetchEnergyFootprint %s", time.Since(start))
return job.EnergyFootprint, nil
}
@@ -274,9 +273,9 @@ func (r *JobRepository) DeleteJobsBefore(startTime int64) (int, error) {
if err != nil {
s, _, _ := qd.ToSql()
log.Errorf(" DeleteJobsBefore(%d) with %s: error %#v", startTime, s, err)
cclog.Errorf(" DeleteJobsBefore(%d) with %s: error %#v", startTime, s, err)
} else {
log.Debugf("DeleteJobsBefore(%d): Deleted %d jobs", startTime, cnt)
cclog.Debugf("DeleteJobsBefore(%d): Deleted %d jobs", startTime, cnt)
}
return cnt, err
}
@@ -287,9 +286,9 @@ func (r *JobRepository) DeleteJobById(id int64) error {
if err != nil {
s, _, _ := qd.ToSql()
log.Errorf("DeleteJobById(%d) with %s : error %#v", id, s, err)
cclog.Errorf("DeleteJobById(%d) with %s : error %#v", id, s, err)
} else {
log.Debugf("DeleteJobById(%d): Success", id)
cclog.Debugf("DeleteJobById(%d): Success", id)
}
return err
}
@@ -352,7 +351,7 @@ func (r *JobRepository) FindColumnValue(user *schema.User, searchterm string, ta
}
return "", ErrNotFound
} else {
log.Infof("Non-Admin User %s : Requested Query '%s' on table '%s' : Forbidden", user.Name, query, table)
cclog.Infof("Non-Admin User %s : Requested Query '%s' on table '%s' : Forbidden", user.Name, query, table)
return "", ErrForbidden
}
}
@@ -371,7 +370,7 @@ func (r *JobRepository) FindColumnValues(user *schema.User, query string, table
err := rows.Scan(&result)
if err != nil {
rows.Close()
log.Warnf("Error while scanning rows: %v", err)
cclog.Warnf("Error while scanning rows: %v", err)
return emptyResult, err
}
results = append(results, result)
@@ -381,7 +380,7 @@ func (r *JobRepository) FindColumnValues(user *schema.User, query string, table
return emptyResult, ErrNotFound
} else {
log.Infof("Non-Admin User %s : Requested Query '%s' on table '%s' : Forbidden", user.Name, query, table)
cclog.Infof("Non-Admin User %s : Requested Query '%s' on table '%s' : Forbidden", user.Name, query, table)
return emptyResult, ErrForbidden
}
}
@@ -389,7 +388,7 @@ func (r *JobRepository) FindColumnValues(user *schema.User, query string, table
func (r *JobRepository) Partitions(cluster string) ([]string, error) {
var err error
start := time.Now()
partitions := r.cache.Get("partitions:"+cluster, func() (interface{}, time.Duration, int) {
partitions := r.cache.Get("partitions:"+cluster, func() (any, time.Duration, int) {
parts := []string{}
if err = r.DB.Select(&parts, `SELECT DISTINCT job.cluster_partition FROM job WHERE job.cluster = ?;`, cluster); err != nil {
return nil, 0, 1000
@@ -400,7 +399,7 @@ func (r *JobRepository) Partitions(cluster string) ([]string, error) {
if err != nil {
return nil, err
}
log.Debugf("Timer Partitions %s", time.Since(start))
cclog.Debugf("Timer Partitions %s", time.Since(start))
return partitions.([]string), nil
}
@@ -414,7 +413,7 @@ func (r *JobRepository) AllocatedNodes(cluster string) (map[string]map[string]in
Where("job.cluster = ?", cluster).
RunWith(r.stmtCache).Query()
if err != nil {
log.Error("Error while running query")
cclog.Error("Error while running query")
return nil, err
}
@@ -425,11 +424,11 @@ func (r *JobRepository) AllocatedNodes(cluster string) (map[string]map[string]in
var resources []*schema.Resource
var subcluster string
if err := rows.Scan(&raw, &subcluster); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
if err := json.Unmarshal(raw, &resources); err != nil {
log.Warn("Error while unmarshaling raw resources json")
cclog.Warn("Error while unmarshaling raw resources json")
return nil, err
}
@@ -444,7 +443,7 @@ func (r *JobRepository) AllocatedNodes(cluster string) (map[string]map[string]in
}
}
log.Debugf("Timer AllocatedNodes %s", time.Since(start))
cclog.Debugf("Timer AllocatedNodes %s", time.Since(start))
return subclusters, nil
}
@@ -460,23 +459,50 @@ func (r *JobRepository) StopJobsExceedingWalltimeBy(seconds int) error {
Where(fmt.Sprintf("(%d - job.start_time) > (job.walltime + %d)", time.Now().Unix(), seconds)).
RunWith(r.DB).Exec()
if err != nil {
log.Warn("Error while stopping jobs exceeding walltime")
cclog.Warn("Error while stopping jobs exceeding walltime")
return err
}
rowsAffected, err := res.RowsAffected()
if err != nil {
log.Warn("Error while fetching affected rows after stopping due to exceeded walltime")
cclog.Warn("Error while fetching affected rows after stopping due to exceeded walltime")
return err
}
if rowsAffected > 0 {
log.Infof("%d jobs have been marked as failed due to running too long", rowsAffected)
cclog.Infof("%d jobs have been marked as failed due to running too long", rowsAffected)
}
log.Debugf("Timer StopJobsExceedingWalltimeBy %s", time.Since(start))
cclog.Debugf("Timer StopJobsExceedingWalltimeBy %s", time.Since(start))
return nil
}
func (r *JobRepository) FindJobIdsByTag(tagId int64) ([]int64, error) {
query := sq.Select("job.id").From("job").
Join("jobtag ON jobtag.job_id = job.id").
Where(sq.Eq{"jobtag.tag_id": tagId}).Distinct()
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
cclog.Error("Error while running query")
return nil, err
}
jobIds := make([]int64, 0, 100)
for rows.Next() {
var jobId int64
if err := rows.Scan(&jobId); err != nil {
rows.Close()
cclog.Warn("Error while scanning rows")
return nil, err
}
jobIds = append(jobIds, jobId)
}
return jobIds, nil
}
// FIXME: Reconsider filtering short jobs with harcoded threshold
func (r *JobRepository) FindRunningJobs(cluster string) ([]*schema.Job, error) {
query := sq.Select(jobColumns...).From("job").
Where(fmt.Sprintf("job.cluster = '%s'", cluster)).
@@ -485,7 +511,7 @@ func (r *JobRepository) FindRunningJobs(cluster string) ([]*schema.Job, error) {
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
log.Error("Error while running query")
cclog.Error("Error while running query")
return nil, err
}
@@ -494,13 +520,13 @@ func (r *JobRepository) FindRunningJobs(cluster string) ([]*schema.Job, error) {
job, err := scanJob(rows)
if err != nil {
rows.Close()
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
jobs = append(jobs, job)
}
log.Infof("Return job count %d", len(jobs))
cclog.Infof("Return job count %d", len(jobs))
return jobs, nil
}
@@ -525,18 +551,18 @@ func (r *JobRepository) FindJobsBetween(startTimeBegin int64, startTimeEnd int64
}
if startTimeBegin == 0 {
log.Infof("Find jobs before %d", startTimeEnd)
cclog.Infof("Find jobs before %d", startTimeEnd)
query = sq.Select(jobColumns...).From("job").Where(fmt.Sprintf(
"job.start_time < %d", startTimeEnd))
} else {
log.Infof("Find jobs between %d and %d", startTimeBegin, startTimeEnd)
cclog.Infof("Find jobs between %d and %d", startTimeBegin, startTimeEnd)
query = sq.Select(jobColumns...).From("job").Where(fmt.Sprintf(
"job.start_time BETWEEN %d AND %d", startTimeBegin, startTimeEnd))
}
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
log.Error("Error while running query")
cclog.Error("Error while running query")
return nil, err
}
@@ -545,13 +571,13 @@ func (r *JobRepository) FindJobsBetween(startTimeBegin int64, startTimeEnd int64
job, err := scanJob(rows)
if err != nil {
rows.Close()
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
jobs = append(jobs, job)
}
log.Infof("Return job count %d", len(jobs))
cclog.Infof("Return job count %d", len(jobs))
return jobs, nil
}
@@ -581,12 +607,12 @@ func (r *JobRepository) MarkArchived(
func (r *JobRepository) UpdateEnergy(
stmt sq.UpdateBuilder,
jobMeta *schema.JobMeta,
jobMeta *schema.Job,
) (sq.UpdateBuilder, error) {
/* Note: Only Called for Running Jobs during Intermediate Update or on Archiving */
sc, err := archive.GetSubCluster(jobMeta.Cluster, jobMeta.SubCluster)
if err != nil {
log.Errorf("cannot get subcluster: %s", err.Error())
cclog.Errorf("cannot get subcluster: %s", err.Error())
return stmt, err
}
energyFootprint := make(map[string]float64)
@@ -599,7 +625,7 @@ func (r *JobRepository) UpdateEnergy(
if i, err := archive.MetricIndex(sc.MetricConfig, fp); err == nil {
// Note: For DB data, calculate and save as kWh
if sc.MetricConfig[i].Energy == "energy" { // this metric has energy as unit (Joules or Wh)
log.Warnf("Update EnergyFootprint for Job %d and Metric %s on cluster %s: Set to 'energy' in cluster.json: Not implemented, will return 0.0", jobMeta.JobID, jobMeta.Cluster, fp)
cclog.Warnf("Update EnergyFootprint for Job %d and Metric %s on cluster %s: Set to 'energy' in cluster.json: Not implemented, will return 0.0", jobMeta.JobID, jobMeta.Cluster, fp)
// FIXME: Needs sum as stats type
} else if sc.MetricConfig[i].Energy == "power" { // this metric has power as unit (Watt)
// Energy: Power (in Watts) * Time (in Seconds)
@@ -611,18 +637,18 @@ func (r *JobRepository) UpdateEnergy(
metricEnergy = math.Round(rawEnergy*100.0) / 100.0
}
} else {
log.Warnf("Error while collecting energy metric %s for job, DB ID '%v', return '0.0'", fp, jobMeta.ID)
cclog.Warnf("Error while collecting energy metric %s for job, DB ID '%v', return '0.0'", fp, jobMeta.ID)
}
energyFootprint[fp] = metricEnergy
totalEnergy += metricEnergy
// log.Infof("Metric %s Average %f -> %f kWh | Job %d Total -> %f kWh", fp, LoadJobStat(jobMeta, fp, "avg"), energy, jobMeta.JobID, totalEnergy)
// cclog.Infof("Metric %s Average %f -> %f kWh | Job %d Total -> %f kWh", fp, LoadJobStat(jobMeta, fp, "avg"), energy, jobMeta.JobID, totalEnergy)
}
var rawFootprint []byte
if rawFootprint, err = json.Marshal(energyFootprint); err != nil {
log.Warnf("Error while marshaling energy footprint for job INTO BYTES, DB ID '%v'", jobMeta.ID)
cclog.Warnf("Error while marshaling energy footprint for job INTO BYTES, DB ID '%v'", jobMeta.ID)
return stmt, err
}
@@ -631,12 +657,12 @@ func (r *JobRepository) UpdateEnergy(
func (r *JobRepository) UpdateFootprint(
stmt sq.UpdateBuilder,
jobMeta *schema.JobMeta,
jobMeta *schema.Job,
) (sq.UpdateBuilder, error) {
/* Note: Only Called for Running Jobs during Intermediate Update or on Archiving */
sc, err := archive.GetSubCluster(jobMeta.Cluster, jobMeta.SubCluster)
if err != nil {
log.Errorf("cannot get subcluster: %s", err.Error())
cclog.Errorf("cannot get subcluster: %s", err.Error())
return stmt, err
}
footprint := make(map[string]float64)
@@ -650,7 +676,7 @@ func (r *JobRepository) UpdateFootprint(
}
if statType != "avg" && statType != "min" && statType != "max" {
log.Warnf("unknown statType for footprint update: %s", statType)
cclog.Warnf("unknown statType for footprint update: %s", statType)
return stmt, fmt.Errorf("unknown statType for footprint update: %s", statType)
}
@@ -664,7 +690,7 @@ func (r *JobRepository) UpdateFootprint(
var rawFootprint []byte
if rawFootprint, err = json.Marshal(footprint); err != nil {
log.Warnf("Error while marshaling footprint for job INTO BYTES, DB ID '%v'", jobMeta.ID)
cclog.Warnf("Error while marshaling footprint for job INTO BYTES, DB ID '%v'", jobMeta.ID)
return stmt, err
}

81
internal/repository/jobCreate.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -8,11 +8,19 @@ import (
"encoding/json"
"fmt"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
sq "github.com/Masterminds/squirrel"
)
const NamedJobCacheInsert string = `INSERT INTO job_cache (
job_id, hpc_user, project, cluster, subcluster, cluster_partition, array_job_id, num_nodes, num_hwthreads, num_acc,
exclusive, monitoring_status, smt, job_state, start_time, duration, walltime, footprint, energy, energy_footprint, resources, meta_data
) VALUES (
:job_id, :hpc_user, :project, :cluster, :subcluster, :cluster_partition, :array_job_id, :num_nodes, :num_hwthreads, :num_acc,
:exclusive, :monitoring_status, :smt, :job_state, :start_time, :duration, :walltime, :footprint, :energy, :energy_footprint, :resources, :meta_data
);`
const NamedJobInsert string = `INSERT INTO job (
job_id, hpc_user, project, cluster, subcluster, cluster_partition, array_job_id, num_nodes, num_hwthreads, num_acc,
exclusive, monitoring_status, smt, job_state, start_time, duration, walltime, footprint, energy, energy_footprint, resources, meta_data
@@ -21,24 +29,65 @@ const NamedJobInsert string = `INSERT INTO job (
:exclusive, :monitoring_status, :smt, :job_state, :start_time, :duration, :walltime, :footprint, :energy, :energy_footprint, :resources, :meta_data
);`
func (r *JobRepository) InsertJob(job *schema.JobMeta) (int64, error) {
res, err := r.DB.NamedExec(NamedJobInsert, job)
func (r *JobRepository) InsertJob(job *schema.Job) (int64, error) {
r.Mutex.Lock()
res, err := r.DB.NamedExec(NamedJobCacheInsert, job)
r.Mutex.Unlock()
if err != nil {
log.Warn("Error while NamedJobInsert")
cclog.Warn("Error while NamedJobInsert")
return 0, err
}
id, err := res.LastInsertId()
if err != nil {
log.Warn("Error while getting last insert ID")
cclog.Warn("Error while getting last insert ID")
return 0, err
}
return id, nil
}
func (r *JobRepository) SyncJobs() ([]*schema.Job, error) {
r.Mutex.Lock()
defer r.Mutex.Unlock()
query := sq.Select(jobCacheColumns...).From("job_cache")
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
cclog.Errorf("Error while running query %v", err)
return nil, err
}
jobs := make([]*schema.Job, 0, 50)
for rows.Next() {
job, err := scanJob(rows)
if err != nil {
rows.Close()
cclog.Warn("Error while scanning rows")
return nil, err
}
jobs = append(jobs, job)
}
_, err = r.DB.Exec(
"INSERT INTO job (job_id, cluster, subcluster, start_time, hpc_user, project, cluster_partition, array_job_id, num_nodes, num_hwthreads, num_acc, exclusive, monitoring_status, smt, job_state, duration, walltime, footprint, energy, energy_footprint, resources, meta_data) SELECT job_id, cluster, subcluster, start_time, hpc_user, project, cluster_partition, array_job_id, num_nodes, num_hwthreads, num_acc, exclusive, monitoring_status, smt, job_state, duration, walltime, footprint, energy, energy_footprint, resources, meta_data FROM job_cache")
if err != nil {
cclog.Warnf("Error while Job sync: %v", err)
return nil, err
}
_, err = r.DB.Exec("DELETE FROM job_cache")
if err != nil {
cclog.Warnf("Error while Job cache clean: %v", err)
return nil, err
}
return jobs, nil
}
// Start inserts a new job in the table, returning the unique job ID.
// Statistics are not transfered!
func (r *JobRepository) Start(job *schema.JobMeta) (id int64, err error) {
func (r *JobRepository) Start(job *schema.Job) (id int64, err error) {
job.RawFootprint, err = json.Marshal(job.Footprint)
if err != nil {
return -1, fmt.Errorf("REPOSITORY/JOB > encoding footprint field failed: %w", err)
@@ -73,3 +122,19 @@ func (r *JobRepository) Stop(
_, err = stmt.RunWith(r.stmtCache).Exec()
return
}
func (r *JobRepository) StopCached(
jobId int64,
duration int32,
state schema.JobState,
monitoringStatus int32,
) (err error) {
stmt := sq.Update("job_cache").
Set("job_state", state).
Set("duration", duration).
Set("monitoring_status", monitoringStatus).
Where("job.id = ?", jobId)
_, err = stmt.RunWith(r.stmtCache).Exec()
return
}

73
internal/repository/jobFind.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -11,8 +11,8 @@ import (
"time"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
sq "github.com/Masterminds/squirrel"
)
@@ -39,7 +39,27 @@ func (r *JobRepository) Find(
q = q.OrderBy("job.id DESC") // always use newest matching job by db id if more than one match
log.Debugf("Timer Find %s", time.Since(start))
cclog.Debugf("Timer Find %s", time.Since(start))
return scanJob(q.RunWith(r.stmtCache).QueryRow())
}
func (r *JobRepository) FindCached(
jobId *int64,
cluster *string,
startTime *int64,
) (*schema.Job, error) {
q := sq.Select(jobCacheColumns...).From("job_cache").
Where("job_cache.job_id = ?", *jobId)
if cluster != nil {
q = q.Where("job_cache.cluster = ?", *cluster)
}
if startTime != nil {
q = q.Where("job_cache.start_time = ?", *startTime)
}
q = q.OrderBy("job_cache.id DESC") // always use newest matching job by db id if more than one match
return scanJob(q.RunWith(r.stmtCache).QueryRow())
}
@@ -66,7 +86,7 @@ func (r *JobRepository) FindAll(
rows, err := q.RunWith(r.stmtCache).Query()
if err != nil {
log.Error("Error while running query")
cclog.Error("Error while running query")
return nil, err
}
@@ -74,15 +94,44 @@ func (r *JobRepository) FindAll(
for rows.Next() {
job, err := scanJob(rows)
if err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
jobs = append(jobs, job)
}
log.Debugf("Timer FindAll %s", time.Since(start))
cclog.Debugf("Timer FindAll %s", time.Since(start))
return jobs, nil
}
// Get complete joblist only consisting of db ids.
// This is useful to process large job counts and intended to be used
// together with FindById to process jobs one by one
func (r *JobRepository) GetJobList() ([]int64, error) {
query := sq.Select("id").From("job").
Where("job.job_state != 'running'")
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
cclog.Error("Error while running query")
return nil, err
}
jl := make([]int64, 0, 1000)
for rows.Next() {
var id int64
err := rows.Scan(&id)
if err != nil {
rows.Close()
cclog.Warn("Error while scanning rows")
return nil, err
}
jl = append(jl, id)
}
cclog.Infof("Return job count %d", len(jl))
return jl, nil
}
// FindById executes a SQL query to find a specific batch job.
// The job is queried using the database id.
// It returns a pointer to a schema.Job data structure and an error variable.
@@ -178,7 +227,7 @@ func (r *JobRepository) FindConcurrentJobs(
var startTime int64
var stopTime int64
startTime = job.StartTimeUnix
startTime = job.StartTime
hostname := job.Resources[0].Hostname
if job.State == schema.JobStateRunning {
@@ -204,7 +253,7 @@ func (r *JobRepository) FindConcurrentJobs(
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
log.Errorf("Error while running query: %v", err)
cclog.Errorf("Error while running query: %v", err)
return nil, err
}
@@ -215,7 +264,7 @@ func (r *JobRepository) FindConcurrentJobs(
var id, jobId, startTime sql.NullInt64
if err = rows.Scan(&id, &jobId, &startTime); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
@@ -231,7 +280,7 @@ func (r *JobRepository) FindConcurrentJobs(
rows, err = queryRunning.RunWith(r.stmtCache).Query()
if err != nil {
log.Errorf("Error while running query: %v", err)
cclog.Errorf("Error while running query: %v", err)
return nil, err
}
@@ -239,7 +288,7 @@ func (r *JobRepository) FindConcurrentJobs(
var id, jobId, startTime sql.NullInt64
if err := rows.Scan(&id, &jobId, &startTime); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}

57
internal/repository/jobHooks.go Normal file
View File

@@ -0,0 +1,57 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
import (
"sync"
"github.com/ClusterCockpit/cc-lib/schema"
)
type JobHook interface {
JobStartCallback(job *schema.Job)
JobStopCallback(job *schema.Job)
}
var (
initOnce sync.Once
hooks []JobHook
)
func RegisterJobJook(hook JobHook) {
initOnce.Do(func() {
hooks = make([]JobHook, 0)
})
if hook != nil {
hooks = append(hooks, hook)
}
}
func CallJobStartHooks(jobs []*schema.Job) {
if hooks == nil {
return
}
for _, hook := range hooks {
if hook != nil {
for _, job := range jobs {
hook.JobStartCallback(job)
}
}
}
}
func CallJobStopHooks(job *schema.Job) {
if hooks == nil {
return
}
for _, hook := range hooks {
if hook != nil {
hook.JobStopCallback(job)
}
}
}

21
internal/repository/jobQuery.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -13,8 +13,8 @@ import (
"time"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
sq "github.com/Masterminds/squirrel"
)
@@ -68,7 +68,7 @@ func (r *JobRepository) QueryJobs(
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
queryString, queryVars, _ := query.ToSql()
log.Errorf("Error while running query '%s' %v: %v", queryString, queryVars, err)
cclog.Errorf("Error while running query '%s' %v: %v", queryString, queryVars, err)
return nil, err
}
@@ -77,7 +77,7 @@ func (r *JobRepository) QueryJobs(
job, err := scanJob(rows)
if err != nil {
rows.Close()
log.Warn("Error while scanning rows (Jobs)")
cclog.Warn("Error while scanning rows (Jobs)")
return nil, err
}
jobs = append(jobs, job)
@@ -123,7 +123,7 @@ func SecurityCheckWithUser(user *schema.User, query sq.SelectBuilder) (sq.Select
if len(user.Projects) != 0 {
return query.Where(sq.Or{sq.Eq{"job.project": user.Projects}, sq.Eq{"job.hpc_user": user.Username}}), nil
} else {
log.Debugf("Manager-User '%s' has no defined projects to lookup! Query only personal jobs ...", user.Username)
cclog.Debugf("Manager-User '%s' has no defined projects to lookup! Query only personal jobs ...", user.Username)
return query.Where("job.hpc_user = ?", user.Username), nil
}
case user.HasRole(schema.RoleUser): // User : Only personal jobs
@@ -146,6 +146,11 @@ func BuildWhereClause(filter *model.JobFilter, query sq.SelectBuilder) sq.Select
// This is an OR-Logic query: Returns all distinct jobs with at least one of the requested tags; TODO: AND-Logic query?
query = query.Join("jobtag ON jobtag.job_id = job.id").Where(sq.Eq{"jobtag.tag_id": filter.Tags}).Distinct()
}
if filter.DbID != nil {
dbIDs := make([]string, len(filter.DbID))
copy(dbIDs, filter.DbID)
query = query.Where(sq.Eq{"job.id": dbIDs})
}
if filter.JobID != nil {
query = buildStringCondition("job.job_id", filter.JobID, query)
}
@@ -239,7 +244,7 @@ func buildTimeCondition(field string, cond *schema.TimeRange, query sq.SelectBui
case "last30d":
then = now - (60 * 60 * 24 * 30)
default:
log.Debugf("No known named timeRange: startTime.range = %s", cond.Range)
cclog.Debugf("No known named timeRange: startTime.range = %s", cond.Range)
return query
}
return query.Where(field+" BETWEEN ? AND ?", then, now)
@@ -330,7 +335,7 @@ var (
func toSnakeCase(str string) string {
for _, c := range str {
if c == '\'' || c == '\\' {
log.Panic("toSnakeCase() attack vector!")
cclog.Panic("toSnakeCase() attack vector!")
}
}

6
internal/repository/job_test.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -9,7 +9,7 @@ import (
"fmt"
"testing"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
"github.com/ClusterCockpit/cc-lib/schema"
_ "github.com/mattn/go-sqlite3"
)
@@ -24,7 +24,7 @@ func TestFind(t *testing.T) {
// fmt.Printf("%+v", job)
if job.ID != 5 {
if *job.ID != 5 {
t.Errorf("wrong summary for diagnostic 3\ngot: %d \nwant: 1366", job.JobID)
}
}

29
internal/repository/migration.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -9,14 +9,14 @@ import (
"embed"
"fmt"
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/golang-migrate/migrate/v4"
"github.com/golang-migrate/migrate/v4/database/mysql"
"github.com/golang-migrate/migrate/v4/database/sqlite3"
"github.com/golang-migrate/migrate/v4/source/iofs"
)
const Version uint = 8
const Version uint = 10
//go:embed migrations/*
var migrationFiles embed.FS
@@ -54,13 +54,13 @@ func checkDBVersion(backend string, db *sql.DB) error {
return err
}
default:
log.Abortf("Migration: Unsupported database backend '%s'.\n", backend)
cclog.Abortf("Migration: Unsupported database backend '%s'.\n", backend)
}
v, dirty, err := m.Version()
if err != nil {
if err == migrate.ErrNilVersion {
log.Warn("Legacy database without version or missing database file!")
cclog.Warn("Legacy database without version or missing database file!")
} else {
return err
}
@@ -84,7 +84,7 @@ func getMigrateInstance(backend string, db string) (m *migrate.Migrate, err erro
case "sqlite3":
d, err := iofs.New(migrationFiles, "migrations/sqlite3")
if err != nil {
log.Fatal(err)
cclog.Fatal(err)
}
m, err = migrate.NewWithSourceInstance("iofs", d, fmt.Sprintf("sqlite3://%s?_foreign_keys=on", db))
@@ -102,7 +102,7 @@ func getMigrateInstance(backend string, db string) (m *migrate.Migrate, err erro
return m, err
}
default:
log.Abortf("Migration: Unsupported database backend '%s'.\n", backend)
cclog.Abortf("Migration: Unsupported database backend '%s'.\n", backend)
}
return m, nil
@@ -115,8 +115,17 @@ func MigrateDB(backend string, db string) error {
}
v, dirty, err := m.Version()
if err != nil {
if err == migrate.ErrNilVersion {
cclog.Warn("Legacy database without version or missing database file!")
} else {
return err
}
}
log.Infof("unsupported database version %d, need %d.\nPlease backup your database file and run cc-backend -migrate-db", v, Version)
if v < Version {
cclog.Infof("unsupported database version %d, need %d.\nPlease backup your database file and run cc-backend -migrate-db", v, Version)
}
if dirty {
return fmt.Errorf("last migration to version %d has failed, please fix the db manually and force version with -force-db flag", Version)
@@ -124,7 +133,7 @@ func MigrateDB(backend string, db string) error {
if err := m.Up(); err != nil {
if err == migrate.ErrNoChange {
log.Info("DB already up to date!")
cclog.Info("DB already up to date!")
} else {
return err
}
@@ -142,7 +151,7 @@ func RevertDB(backend string, db string) error {
if err := m.Migrate(Version - 1); err != nil {
if err == migrate.ErrNoChange {
log.Info("DB already up to date!")
cclog.Info("DB already up to date!")
} else {
return err
}

1
internal/repository/migrations/sqlite3/09_add-job-cache.down.sql Normal file
View File

@@ -0,0 +1 @@
DROP TABLE IF EXISTS job_cache;

31
internal/repository/migrations/sqlite3/09_add-job-cache.up.sql Normal file
View File

@@ -0,0 +1,31 @@
CREATE TABLE "job_cache" (
id INTEGER PRIMARY KEY,
job_id BIGINT NOT NULL,
cluster VARCHAR(255) NOT NULL,
subcluster VARCHAR(255) NOT NULL,
start_time BIGINT NOT NULL, -- Unix timestamp
hpc_user VARCHAR(255) NOT NULL,
project VARCHAR(255) NOT NULL,
cluster_partition VARCHAR(255),
array_job_id BIGINT,
duration INT NOT NULL,
walltime INT NOT NULL,
job_state VARCHAR(255) NOT NULL
CHECK (job_state IN (
'running', 'completed', 'failed', 'cancelled',
'stopped', 'timeout', 'preempted', 'out_of_memory'
)),
meta_data TEXT, -- JSON
resources TEXT NOT NULL, -- JSON
num_nodes INT NOT NULL,
num_hwthreads INT,
num_acc INT,
smt TINYINT NOT NULL DEFAULT 1 CHECK (smt IN (0, 1)),
exclusive TINYINT NOT NULL DEFAULT 1 CHECK (exclusive IN (0, 1, 2)),
monitoring_status TINYINT NOT NULL DEFAULT 1
CHECK (monitoring_status IN (0, 1, 2, 3)),
energy REAL NOT NULL DEFAULT 0.0,
energy_footprint TEXT DEFAULT NULL,
footprint TEXT DEFAULT NULL,
UNIQUE (job_id, cluster, start_time)
);

1
internal/repository/migrations/sqlite3/10_node-table.down.sql Normal file
View File

@@ -0,0 +1 @@
DROP TABLE IF EXISTS node;

23
internal/repository/migrations/sqlite3/10_node-table.up.sql Normal file
View File

@@ -0,0 +1,23 @@
CREATE TABLE "node" (
id INTEGER PRIMARY KEY,
hostname VARCHAR(255) NOT NULL,
cluster VARCHAR(255) NOT NULL,
subcluster VARCHAR(255) NOT NULL,
cpus_allocated INTEGER NOT NULL,
cpus_total INTEGER NOT NULL,
memory_allocated INTEGER NOT NULL,
memory_total INTEGER NOT NULL,
gpus_allocated INTEGER NOT NULL,
gpus_total INTEGER NOT NULL,
node_state VARCHAR(255) NOT NULL
CHECK (node_state IN (
'allocated', 'reserved', 'idle', 'mixed',
'down', 'unknown'
)),
health_state VARCHAR(255) NOT NULL
CHECK (health_state IN (
'full', 'partial', 'failed'
)),
meta_data TEXT, -- JSON
UNIQUE (hostname, cluster)
);

289
internal/repository/node.go Normal file
View File

@@ -0,0 +1,289 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
import (
"context"
"database/sql"
"encoding/json"
"fmt"
"maps"
"sync"
"time"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/lrucache"
"github.com/ClusterCockpit/cc-lib/schema"
sq "github.com/Masterminds/squirrel"
"github.com/jmoiron/sqlx"
)
var (
nodeRepoOnce sync.Once
nodeRepoInstance *NodeRepository
)
type NodeRepository struct {
DB *sqlx.DB
stmtCache *sq.StmtCache
cache *lrucache.Cache
driver string
}
func GetNodeRepository() *NodeRepository {
nodeRepoOnce.Do(func() {
db := GetConnection()
nodeRepoInstance = &NodeRepository{
DB: db.DB,
driver: db.Driver,
stmtCache: sq.NewStmtCache(db.DB),
cache: lrucache.New(1024 * 1024),
}
})
return nodeRepoInstance
}
func (r *NodeRepository) FetchMetadata(node *schema.Node) (map[string]string, error) {
start := time.Now()
cachekey := fmt.Sprintf("metadata:%d", node.ID)
if cached := r.cache.Get(cachekey, nil); cached != nil {
node.MetaData = cached.(map[string]string)
return node.MetaData, nil
}
if err := sq.Select("node.meta_data").From("node").Where("node.id = ?", node.ID).
RunWith(r.stmtCache).QueryRow().Scan(&node.RawMetaData); err != nil {
cclog.Warn("Error while scanning for node metadata")
return nil, err
}
if len(node.RawMetaData) == 0 {
return nil, nil
}
if err := json.Unmarshal(node.RawMetaData, &node.MetaData); err != nil {
cclog.Warn("Error while unmarshaling raw metadata json")
return nil, err
}
r.cache.Put(cachekey, node.MetaData, len(node.RawMetaData), 24*time.Hour)
cclog.Debugf("Timer FetchMetadata %s", time.Since(start))
return node.MetaData, nil
}
func (r *NodeRepository) UpdateMetadata(node *schema.Node, key, val string) (err error) {
cachekey := fmt.Sprintf("metadata:%d", node.ID)
r.cache.Del(cachekey)
if node.MetaData == nil {
if _, err = r.FetchMetadata(node); err != nil {
cclog.Warnf("Error while fetching metadata for node, DB ID '%v'", node.ID)
return err
}
}
if node.MetaData != nil {
cpy := make(map[string]string, len(node.MetaData)+1)
maps.Copy(cpy, node.MetaData)
cpy[key] = val
node.MetaData = cpy
} else {
node.MetaData = map[string]string{key: val}
}
if node.RawMetaData, err = json.Marshal(node.MetaData); err != nil {
cclog.Warnf("Error while marshaling metadata for node, DB ID '%v'", node.ID)
return err
}
if _, err = sq.Update("node").
Set("meta_data", node.RawMetaData).
Where("node.id = ?", node.ID).
RunWith(r.stmtCache).Exec(); err != nil {
cclog.Warnf("Error while updating metadata for node, DB ID '%v'", node.ID)
return err
}
r.cache.Put(cachekey, node.MetaData, len(node.RawMetaData), 24*time.Hour)
return nil
}
func (r *NodeRepository) GetNode(id int64, withMeta bool) (*schema.Node, error) {
node := &schema.Node{}
if err := sq.Select("id", "hostname", "cluster", "subcluster", "node_state",
"health_state").From("node").
Where("node.id = ?", id).RunWith(r.DB).
QueryRow().Scan(&node.ID, &node.Hostname, &node.Cluster, &node.SubCluster, &node.NodeState,
&node.HealthState); err != nil {
cclog.Warnf("Error while querying node '%v' from database", id)
return nil, err
}
if withMeta {
var err error
var meta map[string]string
if meta, err = r.FetchMetadata(node); err != nil {
cclog.Warnf("Error while fetching metadata for node '%v'", id)
return nil, err
}
node.MetaData = meta
}
return node, nil
}
const NamedNodeInsert string = `
INSERT INTO node (hostname, cluster, subcluster, node_state, health_state)
VALUES (:hostname, :cluster, :subcluster, :node_state, :health_state);`
func (r *NodeRepository) AddNode(node *schema.Node) (int64, error) {
var err error
res, err := r.DB.NamedExec(NamedNodeInsert, node)
if err != nil {
cclog.Errorf("Error while adding node '%v' to database", node.Hostname)
return 0, err
}
node.ID, err = res.LastInsertId()
if err != nil {
cclog.Errorf("Error while getting last insert id for node '%v' from database", node.Hostname)
return 0, err
}
return node.ID, nil
}
func (r *NodeRepository) UpdateNodeState(hostname string, cluster string, nodeState *schema.NodeState) error {
var id int64
if err := sq.Select("id").From("node").
Where("node.hostname = ?", hostname).Where("node.cluster = ?", cluster).RunWith(r.DB).
QueryRow().Scan(&id); err != nil {
if err == sql.ErrNoRows {
subcluster, err := archive.GetSubClusterByNode(cluster, hostname)
if err != nil {
cclog.Errorf("Error while getting subcluster for node '%s' in cluster '%s': %v", hostname, cluster, err)
return err
}
node := schema.Node{
Hostname: hostname, Cluster: cluster, SubCluster: subcluster, NodeState: *nodeState,
HealthState: schema.MonitoringStateFull,
}
_, err = r.AddNode(&node)
if err != nil {
cclog.Errorf("Error while adding node '%s' to database: %v", hostname, err)
return err
}
cclog.Infof("Added node '%s' to database", hostname)
return nil
} else {
cclog.Warnf("Error while querying node '%v' from database", id)
return err
}
}
if _, err := sq.Update("node").Set("node_state", nodeState).Where("node.id = ?", id).RunWith(r.DB).Exec(); err != nil {
cclog.Errorf("error while updating node '%s'", hostname)
return err
}
cclog.Infof("Updated node '%s' in database", hostname)
return nil
}
// func (r *NodeRepository) UpdateHealthState(hostname string, healthState *schema.MonitoringState) error {
// if _, err := sq.Update("node").Set("health_state", healthState).Where("node.id = ?", id).RunWith(r.DB).Exec(); err != nil {
// cclog.Errorf("error while updating node '%d'", id)
// return err
// }
//
// return nil
// }
func (r *NodeRepository) DeleteNode(id int64) error {
_, err := r.DB.Exec(`DELETE FROM node WHERE node.id = ?`, id)
if err != nil {
cclog.Errorf("Error while deleting node '%d' from DB", id)
return err
}
cclog.Infof("deleted node '%d' from DB", id)
return nil
}
// TODO: Implement order by
func (r *NodeRepository) QueryNodes(
ctx context.Context,
filters []*model.NodeFilter,
order *model.OrderByInput,
) ([]*schema.Node, error) {
query, qerr := SecurityCheck(ctx, sq.Select(jobColumns...).From("node"))
if qerr != nil {
return nil, qerr
}
for _, f := range filters {
if f.Hostname != nil {
query = buildStringCondition("node.hostname", f.Hostname, query)
}
if f.Cluster != nil {
query = buildStringCondition("node.cluster", f.Cluster, query)
}
if f.NodeState != nil {
query = query.Where("node.node_state = ?", f.NodeState)
}
if f.HealthState != nil {
query = query.Where("node.health_state = ?", f.HealthState)
}
}
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
queryString, queryVars, _ := query.ToSql()
cclog.Errorf("Error while running query '%s' %v: %v", queryString, queryVars, err)
return nil, err
}
nodes := make([]*schema.Node, 0, 50)
for rows.Next() {
node := schema.Node{}
if err := rows.Scan(&node.Hostname, &node.Cluster, &node.SubCluster,
&node.NodeState, &node.HealthState); err != nil {
rows.Close()
cclog.Warn("Error while scanning rows (Nodes)")
return nil, err
}
nodes = append(nodes, &node)
}
return nodes, nil
}
func (r *NodeRepository) ListNodes(cluster string) ([]*schema.Node, error) {
q := sq.Select("hostname", "cluster", "subcluster", "node_state",
"health_state").From("node").Where("node.cluster = ?", cluster).OrderBy("node.hostname ASC")
rows, err := q.RunWith(r.DB).Query()
if err != nil {
cclog.Warn("Error while querying user list")
return nil, err
}
nodeList := make([]*schema.Node, 0, 100)
defer rows.Close()
for rows.Next() {
node := &schema.Node{}
if err := rows.Scan(&node.Hostname, &node.Cluster,
&node.SubCluster, &node.NodeState, &node.HealthState); err != nil {
cclog.Warn("Error while scanning node list")
return nil, err
}
nodeList = append(nodeList, node)
}
return nodeList, nil
}

10
internal/repository/repository_test.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -9,8 +9,8 @@ import (
"testing"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
_ "github.com/mattn/go-sqlite3"
)
@@ -65,7 +65,7 @@ func BenchmarkDB_FindJobById(b *testing.B) {
func BenchmarkDB_FindJob(b *testing.B) {
var jobId int64 = 107266
var startTime int64 = 1657557241
var cluster = "fritz"
cluster := "fritz"
b.Run("FindJob", func(b *testing.B) {
db := setup(b)
@@ -147,7 +147,7 @@ func getContext(tb testing.TB) context.Context {
func setup(tb testing.TB) *JobRepository {
tb.Helper()
log.Init("warn", true)
cclog.Init("warn", true)
dbfile := "testdata/job.db"
err := MigrateDB("sqlite3", dbfile)
noErr(tb, err)

89
internal/repository/stats.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -14,8 +14,8 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/internal/metricDataDispatcher"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
sq "github.com/Masterminds/squirrel"
)
@@ -158,7 +158,7 @@ func (r *JobRepository) JobsStatsGrouped(
rows, err := query.RunWith(r.DB).Query()
if err != nil {
log.Warn("Error while querying DB for job statistics")
cclog.Warn("Error while querying DB for job statistics")
return nil, err
}
@@ -169,7 +169,7 @@ func (r *JobRepository) JobsStatsGrouped(
var name sql.NullString
var jobs, walltime, nodes, nodeHours, cores, coreHours, accs, accHours sql.NullInt64
if err := rows.Scan(&id, &jobs, &name, &walltime, &nodes, &nodeHours, &cores, &coreHours, &accs, &accHours); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
@@ -241,7 +241,7 @@ func (r *JobRepository) JobsStatsGrouped(
}
}
log.Debugf("Timer JobsStatsGrouped %s", time.Since(start))
cclog.Debugf("Timer JobsStatsGrouped %s", time.Since(start))
return stats, nil
}
@@ -261,7 +261,7 @@ func (r *JobRepository) JobsStats(
var jobs, walltime, nodes, nodeHours, cores, coreHours, accs, accHours sql.NullInt64
if err := row.Scan(&jobs, &walltime, &nodes, &nodeHours, &cores, &coreHours, &accs, &accHours); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
@@ -287,11 +287,11 @@ func (r *JobRepository) JobsStats(
})
}
log.Debugf("Timer JobStats %s", time.Since(start))
cclog.Debugf("Timer JobStats %s", time.Since(start))
return stats, nil
}
func LoadJobStat(job *schema.JobMeta, metric string, statType string) float64 {
func LoadJobStat(job *schema.Job, metric string, statType string) float64 {
if stats, ok := job.Statistics[metric]; ok {
switch statType {
case "avg":
@@ -301,7 +301,7 @@ func LoadJobStat(job *schema.JobMeta, metric string, statType string) float64 {
case "min":
return stats.Min
default:
log.Errorf("Unknown stat type %s", statType)
cclog.Errorf("Unknown stat type %s", statType)
}
}
@@ -322,7 +322,7 @@ func (r *JobRepository) JobCountGrouped(
}
rows, err := query.RunWith(r.DB).Query()
if err != nil {
log.Warn("Error while querying DB for job statistics")
cclog.Warn("Error while querying DB for job statistics")
return nil, err
}
@@ -332,7 +332,7 @@ func (r *JobRepository) JobCountGrouped(
var id sql.NullString
var cnt sql.NullInt64
if err := rows.Scan(&id, &cnt); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
if id.Valid {
@@ -344,7 +344,7 @@ func (r *JobRepository) JobCountGrouped(
}
}
log.Debugf("Timer JobCountGrouped %s", time.Since(start))
cclog.Debugf("Timer JobCountGrouped %s", time.Since(start))
return stats, nil
}
@@ -364,7 +364,7 @@ func (r *JobRepository) AddJobCountGrouped(
}
rows, err := query.RunWith(r.DB).Query()
if err != nil {
log.Warn("Error while querying DB for job statistics")
cclog.Warn("Error while querying DB for job statistics")
return nil, err
}
@@ -374,7 +374,7 @@ func (r *JobRepository) AddJobCountGrouped(
var id sql.NullString
var cnt sql.NullInt64
if err := rows.Scan(&id, &cnt); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
if id.Valid {
@@ -393,7 +393,7 @@ func (r *JobRepository) AddJobCountGrouped(
}
}
log.Debugf("Timer AddJobCountGrouped %s", time.Since(start))
cclog.Debugf("Timer AddJobCountGrouped %s", time.Since(start))
return stats, nil
}
@@ -411,7 +411,7 @@ func (r *JobRepository) AddJobCount(
}
rows, err := query.RunWith(r.DB).Query()
if err != nil {
log.Warn("Error while querying DB for job statistics")
cclog.Warn("Error while querying DB for job statistics")
return nil, err
}
@@ -420,7 +420,7 @@ func (r *JobRepository) AddJobCount(
for rows.Next() {
var cnt sql.NullInt64
if err := rows.Scan(&cnt); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
@@ -438,7 +438,7 @@ func (r *JobRepository) AddJobCount(
}
}
log.Debugf("Timer AddJobCount %s", time.Since(start))
cclog.Debugf("Timer AddJobCount %s", time.Since(start))
return stats, nil
}
@@ -479,29 +479,29 @@ func (r *JobRepository) AddHistograms(
value := fmt.Sprintf(`CAST(ROUND(((CASE WHEN job.job_state = "running" THEN %d - job.start_time ELSE job.duration END) / %d) + 1) as %s) as value`, time.Now().Unix(), targetBinSize, castType)
stat.HistDuration, err = r.jobsDurationStatisticsHistogram(ctx, value, filter, targetBinSize, &targetBinCount)
if err != nil {
log.Warn("Error while loading job statistics histogram: job duration")
cclog.Warn("Error while loading job statistics histogram: job duration")
return nil, err
}
stat.HistNumNodes, err = r.jobsStatisticsHistogram(ctx, "job.num_nodes as value", filter)
if err != nil {
log.Warn("Error while loading job statistics histogram: num nodes")
cclog.Warn("Error while loading job statistics histogram: num nodes")
return nil, err
}
stat.HistNumCores, err = r.jobsStatisticsHistogram(ctx, "job.num_hwthreads as value", filter)
if err != nil {
log.Warn("Error while loading job statistics histogram: num hwthreads")
cclog.Warn("Error while loading job statistics histogram: num hwthreads")
return nil, err
}
stat.HistNumAccs, err = r.jobsStatisticsHistogram(ctx, "job.num_acc as value", filter)
if err != nil {
log.Warn("Error while loading job statistics histogram: num acc")
cclog.Warn("Error while loading job statistics histogram: num acc")
return nil, err
}
log.Debugf("Timer AddHistograms %s", time.Since(start))
cclog.Debugf("Timer AddHistograms %s", time.Since(start))
return stat, nil
}
@@ -520,7 +520,7 @@ func (r *JobRepository) AddMetricHistograms(
if f.State != nil {
if len(f.State) == 1 && f.State[0] == "running" {
stat.HistMetrics = r.runningJobsMetricStatisticsHistogram(ctx, metrics, filter, targetBinCount)
log.Debugf("Timer AddMetricHistograms %s", time.Since(start))
cclog.Debugf("Timer AddMetricHistograms %s", time.Since(start))
return stat, nil
}
}
@@ -530,13 +530,13 @@ func (r *JobRepository) AddMetricHistograms(
for _, m := range metrics {
metricHisto, err := r.jobsMetricStatisticsHistogram(ctx, m, filter, targetBinCount)
if err != nil {
log.Warnf("Error while loading job metric statistics histogram: %s", m)
cclog.Warnf("Error while loading job metric statistics histogram: %s", m)
continue
}
stat.HistMetrics = append(stat.HistMetrics, metricHisto)
}
log.Debugf("Timer AddMetricHistograms %s", time.Since(start))
cclog.Debugf("Timer AddMetricHistograms %s", time.Since(start))
return stat, nil
}
@@ -560,7 +560,7 @@ func (r *JobRepository) jobsStatisticsHistogram(
rows, err := query.GroupBy("value").RunWith(r.DB).Query()
if err != nil {
log.Error("Error while running query")
cclog.Error("Error while running query")
return nil, err
}
@@ -569,13 +569,13 @@ func (r *JobRepository) jobsStatisticsHistogram(
for rows.Next() {
point := model.HistoPoint{}
if err := rows.Scan(&point.Value, &point.Count); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
points = append(points, &point)
}
log.Debugf("Timer jobsStatisticsHistogram %s", time.Since(start))
cclog.Debugf("Timer jobsStatisticsHistogram %s", time.Since(start))
return points, nil
}
@@ -607,7 +607,7 @@ func (r *JobRepository) jobsDurationStatisticsHistogram(
rows, err := query.GroupBy("value").RunWith(r.DB).Query()
if err != nil {
log.Error("Error while running query")
cclog.Error("Error while running query")
return nil, err
}
@@ -615,7 +615,7 @@ func (r *JobRepository) jobsDurationStatisticsHistogram(
for rows.Next() {
point := model.HistoPoint{}
if err := rows.Scan(&point.Value, &point.Count); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
@@ -630,7 +630,7 @@ func (r *JobRepository) jobsDurationStatisticsHistogram(
}
}
log.Debugf("Timer jobsStatisticsHistogram %s", time.Since(start))
cclog.Debugf("Timer jobsStatisticsHistogram %s", time.Since(start))
return points, nil
}
@@ -652,7 +652,7 @@ func (r *JobRepository) jobsMetricStatisticsHistogram(
peak = metricConfig.Peak
unit = metricConfig.Unit.Prefix + metricConfig.Unit.Base
footprintStat = metricConfig.Footprint
log.Debugf("Cluster %s filter found with peak %f for %s", *f.Cluster.Eq, peak, metric)
cclog.Debugf("Cluster %s filter found with peak %f for %s", *f.Cluster.Eq, peak, metric)
}
}
@@ -674,7 +674,7 @@ func (r *JobRepository) jobsMetricStatisticsHistogram(
}
}
// log.Debugf("Metric %s, Peak %f, Unit %s", metric, peak, unit)
// cclog.Debugf("Metric %s, Peak %f, Unit %s", metric, peak, unit)
// Make bins, see https://jereze.com/code/sql-histogram/ (Modified here)
start := time.Now()
@@ -686,7 +686,7 @@ func (r *JobRepository) jobsMetricStatisticsHistogram(
mainQuery := sq.Select(
fmt.Sprintf(`%s + 1 as bin`, binQuery),
fmt.Sprintf(`count(*) as count`),
`count(*) as count`,
// For Debug: // fmt.Sprintf(`CAST((%f / %d) as INTEGER ) * %s as min`, peak, *bins, binQuery),
// For Debug: // fmt.Sprintf(`CAST((%f / %d) as INTEGER ) * (%s + 1) as max`, peak, *bins, binQuery),
).From("job").Where(
@@ -709,7 +709,7 @@ func (r *JobRepository) jobsMetricStatisticsHistogram(
rows, err := mainQuery.RunWith(r.DB).Query()
if err != nil {
log.Errorf("Error while running mainQuery: %s", err)
cclog.Errorf("Error while running mainQuery: %s", err)
return nil, err
}
@@ -726,7 +726,7 @@ func (r *JobRepository) jobsMetricStatisticsHistogram(
for rows.Next() { // Fill Count if Bin-No. Matches (Not every Bin exists in DB!)
rpoint := model.MetricHistoPoint{}
if err := rows.Scan(&rpoint.Bin, &rpoint.Count); err != nil { // Required for Debug: &rpoint.Min, &rpoint.Max
log.Warnf("Error while scanning rows for %s", metric)
cclog.Warnf("Error while scanning rows for %s", metric)
return nil, err // FIXME: Totally bricks cc-backend if returned and if all metrics requested?
}
@@ -736,10 +736,10 @@ func (r *JobRepository) jobsMetricStatisticsHistogram(
e.Count = rpoint.Count
// Only Required For Debug: Check DB returned Min/Max against Backend Init above
// if rpoint.Min != nil {
// log.Warnf(">>>> Bin %d Min Set For %s to %d (Init'd with: %d)", *e.Bin, metric, *rpoint.Min, *e.Min)
// cclog.Warnf(">>>> Bin %d Min Set For %s to %d (Init'd with: %d)", *e.Bin, metric, *rpoint.Min, *e.Min)
// }
// if rpoint.Max != nil {
// log.Warnf(">>>> Bin %d Max Set For %s to %d (Init'd with: %d)", *e.Bin, metric, *rpoint.Max, *e.Max)
// cclog.Warnf(">>>> Bin %d Max Set For %s to %d (Init'd with: %d)", *e.Bin, metric, *rpoint.Max, *e.Max)
// }
break
}
@@ -749,7 +749,7 @@ func (r *JobRepository) jobsMetricStatisticsHistogram(
result := model.MetricHistoPoints{Metric: metric, Unit: unit, Stat: &footprintStat, Data: points}
log.Debugf("Timer jobsStatisticsHistogram %s", time.Since(start))
cclog.Debugf("Timer jobsStatisticsHistogram %s", time.Since(start))
return &result, nil
}
@@ -759,15 +759,14 @@ func (r *JobRepository) runningJobsMetricStatisticsHistogram(
filters []*model.JobFilter,
bins *int,
) []*model.MetricHistoPoints {
// Get Jobs
jobs, err := r.QueryJobs(ctx, filters, &model.PageRequest{Page: 1, ItemsPerPage: 500 + 1}, nil)
if err != nil {
log.Errorf("Error while querying jobs for footprint: %s", err)
cclog.Errorf("Error while querying jobs for footprint: %s", err)
return nil
}
if len(jobs) > 500 {
log.Errorf("too many jobs matched (max: %d)", 500)
cclog.Errorf("too many jobs matched (max: %d)", 500)
return nil
}
@@ -783,7 +782,7 @@ func (r *JobRepository) runningJobsMetricStatisticsHistogram(
}
if err := metricDataDispatcher.LoadAverages(job, metrics, avgs, ctx); err != nil {
log.Errorf("Error while loading averages for histogram: %s", err)
cclog.Errorf("Error while loading averages for histogram: %s", err)
return nil
}
}

3
internal/repository/stats_test.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -19,7 +19,6 @@ func TestBuildJobStatsQuery(t *testing.T) {
noErr(t, err)
fmt.Printf("SQL: %s\n", sql)
}
func TestJobStats(t *testing.T) {

196
internal/repository/tags.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -9,8 +9,8 @@ import (
"strings"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
sq "github.com/Masterminds/squirrel"
)
@@ -18,7 +18,7 @@ import (
func (r *JobRepository) AddTag(user *schema.User, job int64, tag int64) ([]*schema.Tag, error) {
j, err := r.FindByIdWithUser(user, job)
if err != nil {
log.Warn("Error while finding job by id")
cclog.Warn("Error while finding job by id")
return nil, err
}
@@ -26,30 +26,61 @@ func (r *JobRepository) AddTag(user *schema.User, job int64, tag int64) ([]*sche
if _, err := q.RunWith(r.stmtCache).Exec(); err != nil {
s, _, _ := q.ToSql()
log.Errorf("Error adding tag with %s: %v", s, err)
cclog.Errorf("Error adding tag with %s: %v", s, err)
return nil, err
}
tags, err := r.GetTags(user, &job)
if err != nil {
log.Warn("Error while getting tags for job")
cclog.Warn("Error while getting tags for job")
return nil, err
}
archiveTags, err := r.getArchiveTags(&job)
if err != nil {
log.Warn("Error while getting tags for job")
cclog.Warn("Error while getting tags for job")
return nil, err
}
return tags, archive.UpdateTags(j, archiveTags)
}
// Removes a tag from a job by tag id
func (r *JobRepository) AddTagDirect(job int64, tag int64) ([]*schema.Tag, error) {
j, err := r.FindByIdDirect(job)
if err != nil {
cclog.Warn("Error while finding job by id")
return nil, err
}
q := sq.Insert("jobtag").Columns("job_id", "tag_id").Values(job, tag)
if _, err := q.RunWith(r.stmtCache).Exec(); err != nil {
s, _, _ := q.ToSql()
cclog.Errorf("Error adding tag with %s: %v", s, err)
return nil, err
}
tags, err := r.GetTagsDirect(&job)
if err != nil {
cclog.Warn("Error while getting tags for job")
return nil, err
}
archiveTags, err := r.getArchiveTags(&job)
if err != nil {
cclog.Warn("Error while getting tags for job")
return nil, err
}
return tags, archive.UpdateTags(j, archiveTags)
}
// Removes a tag from a job by tag id.
// Used by GraphQL API
func (r *JobRepository) RemoveTag(user *schema.User, job, tag int64) ([]*schema.Tag, error) {
j, err := r.FindByIdWithUser(user, job)
if err != nil {
log.Warn("Error while finding job by id")
cclog.Warn("Error while finding job by id")
return nil, err
}
@@ -57,19 +88,19 @@ func (r *JobRepository) RemoveTag(user *schema.User, job, tag int64) ([]*schema.
if _, err := q.RunWith(r.stmtCache).Exec(); err != nil {
s, _, _ := q.ToSql()
log.Errorf("Error removing tag with %s: %v", s, err)
cclog.Errorf("Error removing tag with %s: %v", s, err)
return nil, err
}
tags, err := r.GetTags(user, &job)
if err != nil {
log.Warn("Error while getting tags for job")
cclog.Warn("Error while getting tags for job")
return nil, err
}
archiveTags, err := r.getArchiveTags(&job)
if err != nil {
log.Warn("Error while getting tags for job")
cclog.Warn("Error while getting tags for job")
return nil, err
}
@@ -77,18 +108,19 @@ func (r *JobRepository) RemoveTag(user *schema.User, job, tag int64) ([]*schema.
}
// Removes a tag from a job by tag info
// Used by REST API
func (r *JobRepository) RemoveJobTagByRequest(user *schema.User, job int64, tagType string, tagName string, tagScope string) ([]*schema.Tag, error) {
// Get Tag ID to delete
tagID, exists := r.TagId(tagType, tagName, tagScope)
if !exists {
log.Warnf("Tag does not exist (name, type, scope): %s, %s, %s", tagName, tagType, tagScope)
return nil, fmt.Errorf("Tag does not exist (name, type, scope): %s, %s, %s", tagName, tagType, tagScope)
cclog.Warnf("Tag does not exist (name, type, scope): %s, %s, %s", tagName, tagType, tagScope)
return nil, fmt.Errorf("tag does not exist (name, type, scope): %s, %s, %s", tagName, tagType, tagScope)
}
// Get Job
j, err := r.FindByIdWithUser(user, job)
if err != nil {
log.Warn("Error while finding job by id")
cclog.Warn("Error while finding job by id")
return nil, err
}
@@ -97,63 +129,70 @@ func (r *JobRepository) RemoveJobTagByRequest(user *schema.User, job int64, tagT
if _, err := q.RunWith(r.stmtCache).Exec(); err != nil {
s, _, _ := q.ToSql()
log.Errorf("Error removing tag from table 'jobTag' with %s: %v", s, err)
cclog.Errorf("Error removing tag from table 'jobTag' with %s: %v", s, err)
return nil, err
}
tags, err := r.GetTags(user, &job)
if err != nil {
log.Warn("Error while getting tags for job")
cclog.Warn("Error while getting tags for job")
return nil, err
}
archiveTags, err := r.getArchiveTags(&job)
if err != nil {
log.Warn("Error while getting tags for job")
cclog.Warn("Error while getting tags for job")
return nil, err
}
return tags, archive.UpdateTags(j, archiveTags)
}
func (r *JobRepository) removeTagFromArchiveJobs(jobIds []int64) {
for _, j := range jobIds {
tags, err := r.getArchiveTags(&j)
if err != nil {
cclog.Warnf("Error while getting tags for job %d", j)
continue
}
job, err := r.FindByIdDirect(j)
if err != nil {
cclog.Warnf("Error while getting job %d", j)
continue
}
archive.UpdateTags(job, tags)
}
}
// Removes a tag from db by tag info
// Used by REST API. Does not update tagged jobs in Job archive.
func (r *JobRepository) RemoveTagByRequest(tagType string, tagName string, tagScope string) error {
// Get Tag ID to delete
tagID, exists := r.TagId(tagType, tagName, tagScope)
if !exists {
log.Warnf("Tag does not exist (name, type, scope): %s, %s, %s", tagName, tagType, tagScope)
return fmt.Errorf("Tag does not exist (name, type, scope): %s, %s, %s", tagName, tagType, tagScope)
cclog.Warnf("Tag does not exist (name, type, scope): %s, %s, %s", tagName, tagType, tagScope)
return fmt.Errorf("tag does not exist (name, type, scope): %s, %s, %s", tagName, tagType, tagScope)
}
// Handle Delete JobTagTable
qJobTag := sq.Delete("jobtag").Where("jobtag.tag_id = ?", tagID)
if _, err := qJobTag.RunWith(r.stmtCache).Exec(); err != nil {
s, _, _ := qJobTag.ToSql()
log.Errorf("Error removing tag from table 'jobTag' with %s: %v", s, err)
return err
}
// Handle Delete TagTable
qTag := sq.Delete("tag").Where("tag.id = ?", tagID)
if _, err := qTag.RunWith(r.stmtCache).Exec(); err != nil {
s, _, _ := qTag.ToSql()
log.Errorf("Error removing tag from table 'tag' with %s: %v", s, err)
return err
}
return nil
return r.RemoveTagById(tagID)
}
// Removes a tag from db by tag id
// Used by GraphQL API.
func (r *JobRepository) RemoveTagById(tagID int64) error {
jobIds, err := r.FindJobIdsByTag(tagID)
if err != nil {
return err
}
// Handle Delete JobTagTable
qJobTag := sq.Delete("jobtag").Where("jobtag.tag_id = ?", tagID)
if _, err := qJobTag.RunWith(r.stmtCache).Exec(); err != nil {
s, _, _ := qJobTag.ToSql()
log.Errorf("Error removing tag from table 'jobTag' with %s: %v", s, err)
cclog.Errorf("Error removing tag from table 'jobTag' with %s: %v", s, err)
return err
}
@@ -162,10 +201,13 @@ func (r *JobRepository) RemoveTagById(tagID int64) error {
if _, err := qTag.RunWith(r.stmtCache).Exec(); err != nil {
s, _, _ := qTag.ToSql()
log.Errorf("Error removing tag from table 'tag' with %s: %v", s, err)
cclog.Errorf("Error removing tag from table 'tag' with %s: %v", s, err)
return err
}
// asynchronously update archive jobs
go r.removeTagFromArchiveJobs(jobIds)
return nil
}
@@ -181,7 +223,7 @@ func (r *JobRepository) CreateTag(tagType string, tagName string, tagScope strin
res, err := q.RunWith(r.stmtCache).Exec()
if err != nil {
s, _, _ := q.ToSql()
log.Errorf("Error inserting tag with %s: %v", s, err)
cclog.Errorf("Error inserting tag with %s: %v", s, err)
return 0, err
}
@@ -230,7 +272,7 @@ func (r *JobRepository) CountTags(user *schema.User) (tags []schema.Tag, counts
// Handle Job Ownership
if user != nil && user.HasAnyRole([]schema.Role{schema.RoleAdmin, schema.RoleSupport}) { // ADMIN || SUPPORT: Count all jobs
// log.Debug("CountTags: User Admin or Support -> Count all Jobs for Tags")
// cclog.Debug("CountTags: User Admin or Support -> Count all Jobs for Tags")
// Unchanged: Needs to be own case still, due to UserRole/NoRole compatibility handling in else case
} else if user != nil && user.HasRole(schema.RoleManager) { // MANAGER: Count own jobs plus project's jobs
// Build ("project1", "project2", ...) list of variable length directly in SQL string
@@ -291,6 +333,38 @@ func (r *JobRepository) AddTagOrCreate(user *schema.User, jobId int64, tagType s
return tagId, nil
}
// used in auto tagger plugins
func (r *JobRepository) AddTagOrCreateDirect(jobId int64, tagType string, tagName string) (tagId int64, err error) {
tagScope := "global"
tagId, exists := r.TagId(tagType, tagName, tagScope)
if !exists {
tagId, err = r.CreateTag(tagType, tagName, tagScope)
if err != nil {
return 0, err
}
}
if _, err := r.AddTagDirect(jobId, tagId); err != nil {
return 0, err
}
return tagId, nil
}
func (r *JobRepository) HasTag(jobId int64, tagType string, tagName string) bool {
var id int64
q := sq.Select("id").From("tag").Join("jobtag ON jobtag.tag_id = tag.id").
Where("jobtag.job_id = ?", jobId).Where("tag.tag_type = ?", tagType).
Where("tag.tag_name = ?", tagName)
err := q.RunWith(r.stmtCache).QueryRow().Scan(&id)
if err != nil {
return false
} else {
return true
}
}
// TagId returns the database id of the tag with the specified type and name.
func (r *JobRepository) TagId(tagType string, tagName string, tagScope string) (tagId int64, exists bool) {
exists = true
@@ -322,7 +396,7 @@ func (r *JobRepository) GetTags(user *schema.User, job *int64) ([]*schema.Tag, e
rows, err := q.RunWith(r.stmtCache).Query()
if err != nil {
s, _, _ := q.ToSql()
log.Errorf("Error get tags with %s: %v", s, err)
cclog.Errorf("Error get tags with %s: %v", s, err)
return nil, err
}
@@ -330,7 +404,7 @@ func (r *JobRepository) GetTags(user *schema.User, job *int64) ([]*schema.Tag, e
for rows.Next() {
tag := &schema.Tag{}
if err := rows.Scan(&tag.ID, &tag.Type, &tag.Name, &tag.Scope); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
// Handle Scope Filtering: Tag Scope is Global, Private (== Username) or User is auth'd to view Admin Tags
@@ -346,6 +420,32 @@ func (r *JobRepository) GetTags(user *schema.User, job *int64) ([]*schema.Tag, e
return tags, nil
}
func (r *JobRepository) GetTagsDirect(job *int64) ([]*schema.Tag, error) {
q := sq.Select("id", "tag_type", "tag_name", "tag_scope").From("tag")
if job != nil {
q = q.Join("jobtag ON jobtag.tag_id = tag.id").Where("jobtag.job_id = ?", *job)
}
rows, err := q.RunWith(r.stmtCache).Query()
if err != nil {
s, _, _ := q.ToSql()
cclog.Errorf("Error get tags with %s: %v", s, err)
return nil, err
}
tags := make([]*schema.Tag, 0)
for rows.Next() {
tag := &schema.Tag{}
if err := rows.Scan(&tag.ID, &tag.Type, &tag.Name, &tag.Scope); err != nil {
cclog.Warn("Error while scanning rows")
return nil, err
}
tags = append(tags, tag)
}
return tags, nil
}
// GetArchiveTags returns a list of all tags *regardless of scope* for archiving if job is nil or of the tags that the job with that database ID has.
func (r *JobRepository) getArchiveTags(job *int64) ([]*schema.Tag, error) {
q := sq.Select("id", "tag_type", "tag_name", "tag_scope").From("tag")
@@ -356,7 +456,7 @@ func (r *JobRepository) getArchiveTags(job *int64) ([]*schema.Tag, error) {
rows, err := q.RunWith(r.stmtCache).Query()
if err != nil {
s, _, _ := q.ToSql()
log.Errorf("Error get tags with %s: %v", s, err)
cclog.Errorf("Error get tags with %s: %v", s, err)
return nil, err
}
@@ -364,7 +464,7 @@ func (r *JobRepository) getArchiveTags(job *int64) ([]*schema.Tag, error) {
for rows.Next() {
tag := &schema.Tag{}
if err := rows.Scan(&tag.ID, &tag.Type, &tag.Name, &tag.Scope); err != nil {
log.Warn("Error while scanning rows")
cclog.Warn("Error while scanning rows")
return nil, err
}
tags = append(tags, tag)
@@ -388,7 +488,7 @@ func (r *JobRepository) ImportTag(jobId int64, tagType string, tagName string, t
if _, err := q.RunWith(r.stmtCache).Exec(); err != nil {
s, _, _ := q.ToSql()
log.Errorf("Error adding tag on import with %s: %v", s, err)
cclog.Errorf("Error adding tag on import with %s: %v", s, err)
return err
}

BIN
internal/repository/testdata/job.db vendored
View File

Binary file not shown.

21
internal/repository/transaction.go
View File

@@ -1,11 +1,11 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
import (
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/jmoiron/sqlx"
)
@@ -20,7 +20,7 @@ func (r *JobRepository) TransactionInit() (*Transaction, error) {
t.tx, err = r.DB.Beginx()
if err != nil {
log.Warn("Error while bundling transactions")
cclog.Warn("Error while bundling transactions")
return nil, err
}
return t, nil
@@ -30,14 +30,14 @@ func (r *JobRepository) TransactionCommit(t *Transaction) error {
var err error
if t.tx != nil {
if err = t.tx.Commit(); err != nil {
log.Warn("Error while committing transactions")
cclog.Warn("Error while committing transactions")
return err
}
}
t.tx, err = r.DB.Beginx()
if err != nil {
log.Warn("Error while bundling transactions")
cclog.Warn("Error while bundling transactions")
return err
}
@@ -46,7 +46,7 @@ func (r *JobRepository) TransactionCommit(t *Transaction) error {
func (r *JobRepository) TransactionEnd(t *Transaction) error {
if err := t.tx.Commit(); err != nil {
log.Warn("Error while committing SQL transactions")
cclog.Warn("Error while committing SQL transactions")
return err
}
return nil
@@ -59,13 +59,13 @@ func (r *JobRepository) TransactionAddNamed(
) (int64, error) {
res, err := t.tx.NamedExec(query, args)
if err != nil {
log.Errorf("Named Exec failed: %v", err)
cclog.Errorf("Named Exec failed: %v", err)
return 0, err
}
id, err := res.LastInsertId()
if err != nil {
log.Errorf("repository initDB(): %v", err)
cclog.Errorf("repository initDB(): %v", err)
return 0, err
}
@@ -73,16 +73,15 @@ func (r *JobRepository) TransactionAddNamed(
}
func (r *JobRepository) TransactionAdd(t *Transaction, query string, args ...interface{}) (int64, error) {
res, err := t.tx.Exec(query, args...)
if err != nil {
log.Errorf("TransactionAdd(), Exec() Error: %v", err)
cclog.Errorf("TransactionAdd(), Exec() Error: %v", err)
return 0, err
}
id, err := res.LastInsertId()
if err != nil {
log.Errorf("TransactionAdd(), LastInsertId() Error: %v", err)
cclog.Errorf("TransactionAdd(), LastInsertId() Error: %v", err)
return 0, err
}

58
internal/repository/user.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -13,13 +13,13 @@ import (
"strings"
"sync"
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
sq "github.com/Masterminds/squirrel"
"github.com/jmoiron/sqlx"
"golang.org/x/crypto/bcrypt"
"github.com/ClusterCockpit/cc-backend/internal/config"
)
var (
@@ -50,7 +50,7 @@ func (r *UserRepository) GetUser(username string) (*schema.User, error) {
if err := sq.Select("password", "ldap", "name", "roles", "email", "projects").From("hpc_user").
Where("hpc_user.username = ?", username).RunWith(r.DB).
QueryRow().Scan(&hashedPassword, &user.AuthSource, &name, &rawRoles, &email, &rawProjects); err != nil {
log.Warnf("Error while querying user '%v' from database", username)
cclog.Warnf("Error while querying user '%v' from database", username)
return nil, err
}
@@ -59,7 +59,7 @@ func (r *UserRepository) GetUser(username string) (*schema.User, error) {
user.Email = email.String
if rawRoles.Valid {
if err := json.Unmarshal([]byte(rawRoles.String), &user.Roles); err != nil {
log.Warn("Error while unmarshaling raw roles from DB")
cclog.Warn("Error while unmarshaling raw roles from DB")
return nil, err
}
}
@@ -76,14 +76,14 @@ func (r *UserRepository) GetLdapUsernames() ([]string, error) {
var users []string
rows, err := r.DB.Query(`SELECT username FROM hpc_user WHERE hpc_user.ldap = 1`)
if err != nil {
log.Warn("Error while querying usernames")
cclog.Warn("Error while querying usernames")
return nil, err
}
for rows.Next() {
var username string
if err := rows.Scan(&username); err != nil {
log.Warnf("Error while scanning for user '%s'", username)
cclog.Warnf("Error while scanning for user '%s'", username)
return nil, err
}
@@ -111,7 +111,7 @@ func (r *UserRepository) AddUser(user *schema.User) error {
if user.Password != "" {
password, err := bcrypt.GenerateFromPassword([]byte(user.Password), bcrypt.DefaultCost)
if err != nil {
log.Error("Error while encrypting new user password")
cclog.Error("Error while encrypting new user password")
return err
}
cols = append(cols, "password")
@@ -123,21 +123,21 @@ func (r *UserRepository) AddUser(user *schema.User) error {
}
if _, err := sq.Insert("hpc_user").Columns(cols...).Values(vals...).RunWith(r.DB).Exec(); err != nil {
log.Errorf("Error while inserting new user '%v' into DB", user.Username)
cclog.Errorf("Error while inserting new user '%v' into DB", user.Username)
return err
}
log.Infof("new user %#v created (roles: %s, auth-source: %d, projects: %s)", user.Username, rolesJson, user.AuthSource, projectsJson)
cclog.Infof("new user %#v created (roles: %s, auth-source: %d, projects: %s)", user.Username, rolesJson, user.AuthSource, projectsJson)
defaultMetricsCfg, err := config.LoadDefaultMetricsConfig()
if err != nil {
log.Errorf("Error loading default metrics config: %v", err)
cclog.Errorf("Error loading default metrics config: %v", err)
} else if defaultMetricsCfg != nil {
for _, cluster := range defaultMetricsCfg.Clusters {
metricsArray := config.ParseMetricsString(cluster.DefaultMetrics)
metricsJSON, err := json.Marshal(metricsArray)
if err != nil {
log.Errorf("Error marshaling default metrics for cluster %s: %v", cluster.Name, err)
cclog.Errorf("Error marshaling default metrics for cluster %s: %v", cluster.Name, err)
continue
}
confKey := "job_view_selectedMetrics:" + cluster.Name
@@ -145,9 +145,9 @@ func (r *UserRepository) AddUser(user *schema.User) error {
Columns("username", "confkey", "value").
Values(user.Username, confKey, string(metricsJSON)).
RunWith(r.DB).Exec(); err != nil {
log.Errorf("Error inserting default job view metrics for user %s and cluster %s: %v", user.Username, cluster.Name, err)
cclog.Errorf("Error inserting default job view metrics for user %s and cluster %s: %v", user.Username, cluster.Name, err)
} else {
log.Infof("Default job view metrics for user %s and cluster %s set to %s", user.Username, cluster.Name, string(metricsJSON))
cclog.Infof("Default job view metrics for user %s and cluster %s set to %s", user.Username, cluster.Name, string(metricsJSON))
}
}
}
@@ -160,7 +160,7 @@ func (r *UserRepository) UpdateUser(dbUser *schema.User, user *schema.User) erro
// TODO: Discuss updatable fields
if dbUser.Name != user.Name {
if _, err := sq.Update("hpc_user").Set("name", user.Name).Where("hpc_user.username = ?", dbUser.Username).RunWith(r.DB).Exec(); err != nil {
log.Errorf("error while updating name of user '%s'", user.Username)
cclog.Errorf("error while updating name of user '%s'", user.Username)
return err
}
}
@@ -179,10 +179,10 @@ func (r *UserRepository) UpdateUser(dbUser *schema.User, user *schema.User) erro
func (r *UserRepository) DelUser(username string) error {
_, err := r.DB.Exec(`DELETE FROM hpc_user WHERE hpc_user.username = ?`, username)
if err != nil {
log.Errorf("Error while deleting user '%s' from DB", username)
cclog.Errorf("Error while deleting user '%s' from DB", username)
return err
}
log.Infof("deleted user '%s' from DB", username)
cclog.Infof("deleted user '%s' from DB", username)
return nil
}
@@ -194,7 +194,7 @@ func (r *UserRepository) ListUsers(specialsOnly bool) ([]*schema.User, error) {
rows, err := q.RunWith(r.DB).Query()
if err != nil {
log.Warn("Error while querying user list")
cclog.Warn("Error while querying user list")
return nil, err
}
@@ -206,12 +206,12 @@ func (r *UserRepository) ListUsers(specialsOnly bool) ([]*schema.User, error) {
user := &schema.User{}
var name, email sql.NullString
if err := rows.Scan(&user.Username, &name, &email, &rawroles, &rawprojects); err != nil {
log.Warn("Error while scanning user list")
cclog.Warn("Error while scanning user list")
return nil, err
}
if err := json.Unmarshal([]byte(rawroles), &user.Roles); err != nil {
log.Warn("Error while unmarshaling raw role list")
cclog.Warn("Error while unmarshaling raw role list")
return nil, err
}
@@ -234,7 +234,7 @@ func (r *UserRepository) AddRole(
newRole := strings.ToLower(queryrole)
user, err := r.GetUser(username)
if err != nil {
log.Warnf("Could not load user '%s'", username)
cclog.Warnf("Could not load user '%s'", username)
return err
}
@@ -249,7 +249,7 @@ func (r *UserRepository) AddRole(
roles, _ := json.Marshal(append(user.Roles, newRole))
if _, err := sq.Update("hpc_user").Set("roles", roles).Where("hpc_user.username = ?", username).RunWith(r.DB).Exec(); err != nil {
log.Errorf("error while adding new role for user '%s'", user.Username)
cclog.Errorf("error while adding new role for user '%s'", user.Username)
return err
}
return nil
@@ -259,7 +259,7 @@ func (r *UserRepository) RemoveRole(ctx context.Context, username string, queryr
oldRole := strings.ToLower(queryrole)
user, err := r.GetUser(username)
if err != nil {
log.Warnf("Could not load user '%s'", username)
cclog.Warnf("Could not load user '%s'", username)
return err
}
@@ -285,7 +285,7 @@ func (r *UserRepository) RemoveRole(ctx context.Context, username string, queryr
mroles, _ := json.Marshal(newroles)
if _, err := sq.Update("hpc_user").Set("roles", mroles).Where("hpc_user.username = ?", username).RunWith(r.DB).Exec(); err != nil {
log.Errorf("Error while removing role for user '%s'", user.Username)
cclog.Errorf("Error while removing role for user '%s'", user.Username)
return err
}
return nil
@@ -364,10 +364,10 @@ const ContextUserKey ContextKey = "user"
func GetUserFromContext(ctx context.Context) *schema.User {
x := ctx.Value(ContextUserKey)
if x == nil {
log.Warnf("no user retrieved from context")
cclog.Warnf("no user retrieved from context")
return nil
}
// log.Infof("user retrieved from context: %v", x.(*schema.User))
// cclog.Infof("user retrieved from context: %v", x.(*schema.User))
return x.(*schema.User)
}
@@ -385,11 +385,11 @@ func (r *UserRepository) FetchUserInCtx(ctx context.Context, username string) (*
if err == sql.ErrNoRows {
/* This warning will be logged *often* for non-local users, i.e. users mentioned only in job-table or archive, */
/* since FetchUser will be called to retrieve full name and mail for every job in query/list */
// log.Warnf("User '%s' Not found in DB", username)
// cclog.Warnf("User '%s' Not found in DB", username)
return nil, nil
}
log.Warnf("Error while fetching user '%s'", username)
cclog.Warnf("Error while fetching user '%s'", username)
return nil, err
}

22
internal/repository/userConfig.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -10,9 +10,9 @@ import (
"time"
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/lrucache"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/lrucache"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/jmoiron/sqlx"
)
@@ -35,7 +35,7 @@ func GetUserCfgRepo() *UserCfgRepo {
lookupConfigStmt, err := db.DB.Preparex(`SELECT confkey, value FROM configuration WHERE configuration.username = ?`)
if err != nil {
log.Fatalf("User Config: Call 'db.DB.Preparex()' failed.\nError: %s\n", err.Error())
cclog.Fatalf("User Config: Call 'db.DB.Preparex()' failed.\nError: %s\n", err.Error())
}
userCfgRepoInstance = &UserCfgRepo{
@@ -70,7 +70,7 @@ func (uCfg *UserCfgRepo) GetUIConfig(user *schema.User) (map[string]interface{},
rows, err := uCfg.Lookup.Query(user.Username)
if err != nil {
log.Warnf("Error while looking up user uiconfig for user '%v'", user.Username)
cclog.Warnf("Error while looking up user uiconfig for user '%v'", user.Username)
return err, 0, 0
}
@@ -79,13 +79,13 @@ func (uCfg *UserCfgRepo) GetUIConfig(user *schema.User) (map[string]interface{},
for rows.Next() {
var key, rawval string
if err := rows.Scan(&key, &rawval); err != nil {
log.Warn("Error while scanning user uiconfig values")
cclog.Warn("Error while scanning user uiconfig values")
return err, 0, 0
}
var val interface{}
if err := json.Unmarshal([]byte(rawval), &val); err != nil {
log.Warn("Error while unmarshaling raw user uiconfig json")
cclog.Warn("Error while unmarshaling raw user uiconfig json")
return err, 0, 0
}
@@ -100,7 +100,7 @@ func (uCfg *UserCfgRepo) GetUIConfig(user *schema.User) (map[string]interface{},
return uiconfig, 24 * time.Hour, size
})
if err, ok := data.(error); ok {
log.Error("Error in returned dataset")
cclog.Error("Error in returned dataset")
return nil, err
}
@@ -117,7 +117,7 @@ func (uCfg *UserCfgRepo) UpdateConfig(
if user == nil {
var val interface{}
if err := json.Unmarshal([]byte(value), &val); err != nil {
log.Warn("Error while unmarshaling raw user config json")
cclog.Warn("Error while unmarshaling raw user config json")
return err
}
@@ -128,7 +128,7 @@ func (uCfg *UserCfgRepo) UpdateConfig(
}
if _, err := uCfg.DB.Exec(`REPLACE INTO configuration (username, confkey, value) VALUES (?, ?, ?)`, user.Username, key, value); err != nil {
log.Warnf("Error while replacing user config in DB for user '%v'", user.Username)
cclog.Warnf("Error while replacing user config in DB for user '%v'", user.Username)
return err
}

8
internal/repository/userConfig_test.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package repository
@@ -10,8 +10,8 @@ import (
"testing"
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
_ "github.com/mattn/go-sqlite3"
)
@@ -39,7 +39,7 @@ func setupUserTest(t *testing.T) *UserCfgRepo {
} } ]
}`
log.Init("info", true)
cclog.Init("info", true)
dbfilepath := "testdata/job.db"
err := MigrateDB("sqlite3", dbfilepath)
if err != nil {

25
internal/routerConfig/routes.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package routerConfig
@@ -16,10 +16,10 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/internal/util"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
"github.com/ClusterCockpit/cc-backend/web"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/ClusterCockpit/cc-lib/util"
"github.com/gorilla/mux"
)
@@ -57,23 +57,23 @@ func setupHomeRoute(i InfoType, r *http.Request) InfoType {
// startJobCount := time.Now()
stats, err := jobRepo.JobCountGrouped(r.Context(), nil, &groupBy)
if err != nil {
log.Warnf("failed to count jobs: %s", err.Error())
cclog.Warnf("failed to count jobs: %s", err.Error())
}
// log.Infof("Timer HOME ROUTE startJobCount: %s", time.Since(startJobCount))
// cclog.Infof("Timer HOME ROUTE startJobCount: %s", time.Since(startJobCount))
// startRunningJobCount := time.Now()
stats, err = jobRepo.AddJobCountGrouped(r.Context(), nil, &groupBy, stats, "running")
if err != nil {
log.Warnf("failed to count running jobs: %s", err.Error())
cclog.Warnf("failed to count running jobs: %s", err.Error())
}
// log.Infof("Timer HOME ROUTE startRunningJobCount: %s", time.Since(startRunningJobCount))
// cclog.Infof("Timer HOME ROUTE startRunningJobCount: %s", time.Since(startRunningJobCount))
i["clusters"] = stats
if util.CheckFileExists("./var/notice.txt") {
msg, err := os.ReadFile("./var/notice.txt")
if err != nil {
log.Warnf("failed to read notice.txt file: %s", err.Error())
cclog.Warnf("failed to read notice.txt file: %s", err.Error())
} else {
i["message"] = string(msg)
}
@@ -161,7 +161,7 @@ func setupNodeRoute(i InfoType, r *http.Request) InfoType {
i["hostname"] = vars["hostname"]
i["id"] = fmt.Sprintf("%s (%s)", vars["cluster"], vars["hostname"])
from, to := r.URL.Query().Get("from"), r.URL.Query().Get("to")
if from != "" || to != "" {
if from != "" && to != "" {
i["from"] = from
i["to"] = to
}
@@ -178,7 +178,7 @@ func setupTaglistRoute(i InfoType, r *http.Request) InfoType {
tags, counts, err := jobRepo.CountTags(repository.GetUserFromContext(r.Context()))
tagMap := make(map[string][]map[string]interface{})
if err != nil {
log.Warnf("GetTags failed: %s", err.Error())
cclog.Warnf("GetTags failed: %s", err.Error())
i["tagmap"] = tagMap
return i
}
@@ -297,6 +297,9 @@ func buildFilterPresets(query url.Values) map[string]interface{} {
}
}
}
if len(query["dbId"]) != 0 {
filterPresets["dbId"] = query["dbId"]
}
if query.Get("jobId") != "" {
if len(query["jobId"]) == 1 {
filterPresets["jobId"] = query.Get("jobId")

1
internal/tagger/apps/alf.txt Normal file
View File

@@ -0,0 +1 @@
alf

7
internal/tagger/apps/caracal.txt Normal file
View File

@@ -0,0 +1,7 @@
calc_rate
qmdffgen
dynamic
evbopt
explore
black_box
poly_qmdff

3
internal/tagger/apps/chroma.txt Normal file
View File

@@ -0,0 +1,3 @@
chroma
qdp
qmp

1
internal/tagger/apps/cp2k.txt Normal file
View File

@@ -0,0 +1 @@
cp2k

1
internal/tagger/apps/cpmd.txt Normal file
View File

@@ -0,0 +1 @@
cpmd

1
internal/tagger/apps/flame.txt Normal file
View File

@@ -0,0 +1 @@
flame

3
internal/tagger/apps/gromacs.txt Normal file
View File

@@ -0,0 +1,3 @@
gromacs
gmx
mdrun

1
internal/tagger/apps/julia.txt Normal file
View File

@@ -0,0 +1 @@
julia

1
internal/tagger/apps/lammps.txt Normal file
View File

@@ -0,0 +1 @@
lmp

1
internal/tagger/apps/matlab.txt Normal file
View File

@@ -0,0 +1 @@
matlab

1
internal/tagger/apps/openfoam.txt Normal file
View File

@@ -0,0 +1 @@
openfoam

1
internal/tagger/apps/orca.txt Normal file
View File

@@ -0,0 +1 @@
orca

4
internal/tagger/apps/python.txt Normal file
View File

@@ -0,0 +1,4 @@
python
pip
anaconda
conda

2
internal/tagger/apps/starccm.txt Normal file
View File

@@ -0,0 +1,2 @@
starccm+
-podkey

10
internal/tagger/apps/turbomole.txt Normal file
View File

@@ -0,0 +1,10 @@
dscf
grad
ridft
rdgrad
ricc2
statpt
aoforce
escf
egrad
odft

1
internal/tagger/apps/vasp.txt Normal file
View File

@@ -0,0 +1 @@
vasp

322
internal/tagger/classifyJob.go Normal file
View File

@@ -0,0 +1,322 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package tagger
import (
"bytes"
"embed"
"encoding/json"
"fmt"
"maps"
"os"
"strings"
"text/template"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/ClusterCockpit/cc-lib/util"
"github.com/expr-lang/expr"
"github.com/expr-lang/expr/vm"
)
//go:embed jobclasses/*
var jobclassFiles embed.FS
type Variable struct {
Name string `json:"name"`
Expr string `json:"expr"`
}
type ruleVariable struct {
name string
expr *vm.Program
}
type RuleFormat struct {
Name string `json:"name"`
Tag string `json:"tag"`
Parameters []string `json:"parameters"`
Metrics []string `json:"metrics"`
Requirements []string `json:"requirements"`
Variables []Variable `json:"variables"`
Rule string `json:"rule"`
Hint string `json:"hint"`
}
type ruleInfo struct {
env map[string]any
metrics []string
requirements []*vm.Program
variables []ruleVariable
rule *vm.Program
hint *template.Template
}
type JobClassTagger struct {
rules map[string]ruleInfo
parameters map[string]any
tagType string
cfgPath string
}
func (t *JobClassTagger) prepareRule(b []byte, fns string) {
var rule RuleFormat
if err := json.NewDecoder(bytes.NewReader(b)).Decode(&rule); err != nil {
cclog.Warn("Error while decoding raw job meta json")
return
}
ri := ruleInfo{}
ri.env = make(map[string]any)
ri.metrics = make([]string, 0)
ri.requirements = make([]*vm.Program, 0)
ri.variables = make([]ruleVariable, 0)
// check if all required parameters are available
for _, p := range rule.Parameters {
param, ok := t.parameters[p]
if !ok {
cclog.Warnf("prepareRule() > missing parameter %s in rule %s", p, fns)
return
}
ri.env[p] = param
}
// set all required metrics
ri.metrics = append(ri.metrics, rule.Metrics...)
// compile requirements
for _, r := range rule.Requirements {
req, err := expr.Compile(r, expr.AsBool())
if err != nil {
cclog.Errorf("error compiling requirement %s: %#v", r, err)
return
}
ri.requirements = append(ri.requirements, req)
}
// compile variables
for _, v := range rule.Variables {
req, err := expr.Compile(v.Expr, expr.AsFloat64())
if err != nil {
cclog.Errorf("error compiling requirement %s: %#v", v.Name, err)
return
}
ri.variables = append(ri.variables, ruleVariable{name: v.Name, expr: req})
}
// compile rule
exp, err := expr.Compile(rule.Rule, expr.AsBool())
if err != nil {
cclog.Errorf("error compiling rule %s: %#v", fns, err)
return
}
ri.rule = exp
// prepare hint template
ri.hint, err = template.New(fns).Parse(rule.Hint)
if err != nil {
cclog.Errorf("error processing template %s: %#v", fns, err)
}
cclog.Infof("prepareRule() > processing %s with %d requirements and %d variables", fns, len(ri.requirements), len(ri.variables))
t.rules[rule.Tag] = ri
}
func (t *JobClassTagger) EventMatch(s string) bool {
return strings.Contains(s, "jobclasses")
}
// FIXME: Only process the file that caused the event
func (t *JobClassTagger) EventCallback() {
files, err := os.ReadDir(t.cfgPath)
if err != nil {
cclog.Fatal(err)
}
if util.CheckFileExists(t.cfgPath + "/parameters.json") {
cclog.Info("Merge parameters")
b, err := os.ReadFile(t.cfgPath + "/parameters.json")
if err != nil {
cclog.Warnf("prepareRule() > open file error: %v", err)
}
var paramTmp map[string]any
if err := json.NewDecoder(bytes.NewReader(b)).Decode(&paramTmp); err != nil {
cclog.Warn("Error while decoding parameters.json")
}
maps.Copy(t.parameters, paramTmp)
}
for _, fn := range files {
fns := fn.Name()
if fns != "parameters.json" {
cclog.Debugf("Process: %s", fns)
filename := fmt.Sprintf("%s/%s", t.cfgPath, fns)
b, err := os.ReadFile(filename)
if err != nil {
cclog.Warnf("prepareRule() > open file error: %v", err)
return
}
t.prepareRule(b, fns)
}
}
}
func (t *JobClassTagger) initParameters() error {
cclog.Info("Initialize parameters")
b, err := jobclassFiles.ReadFile("jobclasses/parameters.json")
if err != nil {
cclog.Warnf("prepareRule() > open file error: %v", err)
return err
}
if err := json.NewDecoder(bytes.NewReader(b)).Decode(&t.parameters); err != nil {
cclog.Warn("Error while decoding parameters.json")
return err
}
return nil
}
func (t *JobClassTagger) Register() error {
t.cfgPath = "./var/tagger/jobclasses"
t.tagType = "jobClass"
err := t.initParameters()
if err != nil {
cclog.Warnf("error reading parameters.json: %v", err)
return err
}
files, err := jobclassFiles.ReadDir("jobclasses")
if err != nil {
return fmt.Errorf("error reading app folder: %#v", err)
}
t.rules = make(map[string]ruleInfo, 0)
for _, fn := range files {
fns := fn.Name()
if fns != "parameters.json" {
filename := fmt.Sprintf("jobclasses/%s", fns)
cclog.Infof("Process: %s", fns)
b, err := jobclassFiles.ReadFile(filename)
if err != nil {
cclog.Warnf("prepareRule() > open file error: %v", err)
return err
}
t.prepareRule(b, fns)
}
}
if util.CheckFileExists(t.cfgPath) {
t.EventCallback()
cclog.Infof("Setup file watch for %s", t.cfgPath)
util.AddListener(t.cfgPath, t)
}
return nil
}
func (t *JobClassTagger) Match(job *schema.Job) {
r := repository.GetJobRepository()
jobstats, err := archive.GetStatistics(job)
metricsList := archive.GetMetricConfigSubCluster(job.Cluster, job.SubCluster)
cclog.Infof("Enter match rule with %d rules for job %d", len(t.rules), job.JobID)
if err != nil {
cclog.Errorf("job classification failed for job %d: %#v", job.JobID, err)
return
}
for tag, ri := range t.rules {
env := make(map[string]any)
maps.Copy(env, ri.env)
cclog.Infof("Try to match rule %s for job %d", tag, job.JobID)
// Initialize environment
env["job"] = map[string]any{
"exclusive": job.Exclusive,
"duration": job.Duration,
"numCores": job.NumHWThreads,
"numNodes": job.NumNodes,
"jobState": job.State,
"numAcc": job.NumAcc,
"smt": job.SMT,
}
// add metrics to env
for _, m := range ri.metrics {
stats, ok := jobstats[m]
if !ok {
cclog.Errorf("job classification failed for job %d: missing metric '%s'", job.JobID, m)
return
}
env[m] = map[string]any{
"min": stats.Min,
"max": stats.Max,
"avg": stats.Avg,
"limits": map[string]float64{
"peak": metricsList[m].Peak,
"normal": metricsList[m].Normal,
"caution": metricsList[m].Caution,
"alert": metricsList[m].Alert,
},
}
}
// check rule requirements apply
for _, r := range ri.requirements {
ok, err := expr.Run(r, env)
if err != nil {
cclog.Errorf("error running requirement for rule %s: %#v", tag, err)
return
}
if !ok.(bool) {
cclog.Infof("requirement for rule %s not met", tag)
return
}
}
// validate rule expression
for _, v := range ri.variables {
value, err := expr.Run(v.expr, env)
if err != nil {
cclog.Errorf("error running rule %s: %#v", tag, err)
return
}
env[v.name] = value
}
// dump.P(env)
match, err := expr.Run(ri.rule, env)
if err != nil {
cclog.Errorf("error running rule %s: %#v", tag, err)
return
}
if match.(bool) {
cclog.Info("Rule matches!")
id := *job.ID
if !r.HasTag(id, t.tagType, tag) {
r.AddTagOrCreateDirect(id, t.tagType, tag)
}
// process hint template
var msg bytes.Buffer
if err := ri.hint.Execute(&msg, env); err != nil {
cclog.Errorf("Template error: %s", err.Error())
return
}
// FIXME: Handle case where multiple tags apply
r.UpdateMetadata(job, "message", msg.String())
} else {
cclog.Info("Rule does not match!")
}
}
}

127
internal/tagger/detectApp.go Normal file
View File

@@ -0,0 +1,127 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package tagger
import (
"bufio"
"embed"
"fmt"
"io/fs"
"os"
"path/filepath"
"regexp"
"strings"
"github.com/ClusterCockpit/cc-backend/internal/repository"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/ClusterCockpit/cc-lib/util"
)
//go:embed apps/*
var appFiles embed.FS
type appInfo struct {
tag string
strings []string
}
type AppTagger struct {
apps map[string]appInfo
tagType string
cfgPath string
}
func (t *AppTagger) scanApp(f fs.File, fns string) {
scanner := bufio.NewScanner(f)
ai := appInfo{tag: strings.TrimSuffix(fns, filepath.Ext(fns)), strings: make([]string, 0)}
for scanner.Scan() {
ai.strings = append(ai.strings, scanner.Text())
}
delete(t.apps, ai.tag)
t.apps[ai.tag] = ai
}
func (t *AppTagger) EventMatch(s string) bool {
return strings.Contains(s, "apps")
}
// FIXME: Only process the file that caused the event
func (t *AppTagger) EventCallback() {
files, err := os.ReadDir(t.cfgPath)
if err != nil {
cclog.Fatal(err)
}
for _, fn := range files {
fns := fn.Name()
cclog.Debugf("Process: %s", fns)
f, err := os.Open(fmt.Sprintf("%s/%s", t.cfgPath, fns))
if err != nil {
cclog.Errorf("error opening app file %s: %#v", fns, err)
}
t.scanApp(f, fns)
}
}
func (t *AppTagger) Register() error {
t.cfgPath = "./var/tagger/apps"
t.tagType = "app"
files, err := appFiles.ReadDir("apps")
if err != nil {
return fmt.Errorf("error reading app folder: %#v", err)
}
t.apps = make(map[string]appInfo, 0)
for _, fn := range files {
fns := fn.Name()
cclog.Debugf("Process: %s", fns)
f, err := appFiles.Open(fmt.Sprintf("apps/%s", fns))
if err != nil {
return fmt.Errorf("error opening app file %s: %#v", fns, err)
}
defer f.Close()
t.scanApp(f, fns)
}
if util.CheckFileExists(t.cfgPath) {
t.EventCallback()
cclog.Infof("Setup file watch for %s", t.cfgPath)
util.AddListener(t.cfgPath, t)
}
return nil
}
func (t *AppTagger) Match(job *schema.Job) {
r := repository.GetJobRepository()
metadata, err := r.FetchMetadata(job)
if err != nil {
cclog.Infof("Cannot fetch metadata for job: %d on %s", job.JobID, job.Cluster)
return
}
jobscript, ok := metadata["jobScript"]
if ok {
id := *job.ID
out:
for _, a := range t.apps {
tag := a.tag
for _, s := range a.strings {
matched, _ := regexp.MatchString(s, strings.ToLower(jobscript))
if matched {
if !r.HasTag(id, t.tagType, tag) {
r.AddTagOrCreateDirect(id, t.tagType, tag)
break out
}
}
}
}
} else {
cclog.Infof("Cannot extract job script for job: %d on %s", job.JobID, job.Cluster)
}
}

59
internal/tagger/detectApp_test.go Normal file
View File

@@ -0,0 +1,59 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package tagger
import (
"testing"
"github.com/ClusterCockpit/cc-backend/internal/repository"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
)
func setup(tb testing.TB) *repository.JobRepository {
tb.Helper()
cclog.Init("warn", true)
dbfile := "../repository/testdata/job.db"
err := repository.MigrateDB("sqlite3", dbfile)
noErr(tb, err)
repository.Connect("sqlite3", dbfile)
return repository.GetJobRepository()
}
func noErr(tb testing.TB, err error) {
tb.Helper()
if err != nil {
tb.Fatal("Error is not nil:", err)
}
}
func TestRegister(t *testing.T) {
var tagger AppTagger
err := tagger.Register()
noErr(t, err)
if len(tagger.apps) != 16 {
t.Errorf("wrong summary for diagnostic \ngot: %d \nwant: 16", len(tagger.apps))
}
}
func TestMatch(t *testing.T) {
r := setup(t)
job, err := r.FindByIdDirect(5)
noErr(t, err)
var tagger AppTagger
err = tagger.Register()
noErr(t, err)
tagger.Match(job)
if !r.HasTag(5, "app", "vasp") {
t.Errorf("missing tag vasp")
}
}

26
internal/tagger/jobclasses/highload.json Normal file
View File

@@ -0,0 +1,26 @@
{
"name": "Excessive CPU load",
"tag": "excessiveload",
"parameters": [
"excessivecpuload_threshold_factor",
"job_min_duration_seconds",
"sampling_interval_seconds"
],
"metrics": ["cpu_load"],
"requirements": [
"job.exclusive == 1",
"job.duration > job_min_duration_seconds"
],
"variables": [
{
"name": "load_threshold",
"expr": "cpu_load.limits.peak * excessivecpuload_threshold_factor"
},
{
"name": "load_perc",
"expr": "1.0 - (cpu_load.avg / cpu_load.limits.peak)"
}
],
"rule": "cpu_load.avg > load_threshold",
"hint": "This job was detected as excessiveload because the average cpu load {{.cpu_load.avg}} falls above the threshold {{.load_threshold}}."
}

22
internal/tagger/jobclasses/lowUtilization.json Normal file
View File

@@ -0,0 +1,22 @@
{
"name": "Low ressource utilization",
"tag": "lowutilization",
"parameters": ["job_min_duration_seconds"],
"metrics": ["flops_any", "mem_bw"],
"requirements": [
"job.exclusive == 1",
"job.duration > job_min_duration_seconds"
],
"variables": [
{
"name": "mem_bw_perc",
"expr": "1.0 - (mem_bw.avg / mem_bw.limits.peak)"
},
{
"name": "flops_any_perc",
"expr": "1.0 - (flops_any.avg / flops_any.limits.peak)"
}
],
"rule": "flops_any.avg < flops_any.limits.alert && mem_bw.avg < mem_bw.limits.alert",
"hint": "This job was detected as low utilization because the average flop rate {{.flops_any.avg}} falls below the threshold {{.flops_any.limits.alert}}."
}

26
internal/tagger/jobclasses/lowload.json Normal file
View File

@@ -0,0 +1,26 @@
{
"name": "Low CPU load",
"tag": "lowload",
"parameters": [
"lowcpuload_threshold_factor",
"job_min_duration_seconds",
"sampling_interval_seconds"
],
"metrics": ["cpu_load"],
"requirements": [
"job.exclusive == 1",
"job.duration > job_min_duration_seconds"
],
"variables": [
{
"name": "load_threshold",
"expr": "job.numCores * lowcpuload_threshold_factor"
},
{
"name": "load_perc",
"expr": "1.0 - (cpu_load.avg / cpu_load.limits.peak)"
}
],
"rule": "cpu_load.avg < cpu_load.limits.caution",
"hint": "This job was detected as lowload because the average cpu load {{.cpu_load}} falls below the threshold {{.cpu_load.limits.caution}}."
}

14
internal/tagger/jobclasses/parameters.json Normal file
View File

@@ -0,0 +1,14 @@
{
"lowcpuload_threshold_factor": 0.9,
"excessivecpuload_threshold_factor": 1.1,
"highmemoryusage_threshold_factor": 0.9,
"node_load_imbalance_threshold_factor": 0.1,
"core_load_imbalance_threshold_factor": 0.1,
"high_memory_load_threshold_factor": 0.9,
"lowgpuload_threshold_factor": 0.7,
"memory_leak_slope_threshold": 0.1,
"job_min_duration_seconds": 600.0,
"sampling_interval_seconds": 30.0,
"cpu_load_pre_cutoff_samples": 11.0,
"cpu_load_core_pre_cutoff_samples": 6.0
}

88
internal/tagger/tagger.go Normal file
View File

@@ -0,0 +1,88 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package tagger
import (
"sync"
"github.com/ClusterCockpit/cc-backend/internal/repository"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
type Tagger interface {
Register() error
Match(job *schema.Job)
}
var (
initOnce sync.Once
jobTagger *JobTagger
)
type JobTagger struct {
startTaggers []Tagger
stopTaggers []Tagger
}
func newTagger() {
jobTagger = &JobTagger{}
jobTagger.startTaggers = make([]Tagger, 0)
jobTagger.startTaggers = append(jobTagger.startTaggers, &AppTagger{})
jobTagger.stopTaggers = make([]Tagger, 0)
jobTagger.stopTaggers = append(jobTagger.stopTaggers, &JobClassTagger{})
for _, tagger := range jobTagger.startTaggers {
tagger.Register()
}
for _, tagger := range jobTagger.stopTaggers {
tagger.Register()
}
}
func Init() {
initOnce.Do(func() {
newTagger()
repository.RegisterJobJook(jobTagger)
})
}
func (jt *JobTagger) JobStartCallback(job *schema.Job) {
for _, tagger := range jt.startTaggers {
tagger.Match(job)
}
}
func (jt *JobTagger) JobStopCallback(job *schema.Job) {
for _, tagger := range jt.stopTaggers {
tagger.Match(job)
}
}
func RunTaggers() error {
newTagger()
r := repository.GetJobRepository()
jl, err := r.GetJobList()
if err != nil {
cclog.Errorf("Error while getting job list %s", err)
return err
}
for _, id := range jl {
job, err := r.FindByIdDirect(id)
if err != nil {
cclog.Errorf("Error while getting job %s", err)
return err
}
for _, tagger := range jobTagger.startTaggers {
tagger.Match(job)
}
for _, tagger := range jobTagger.stopTaggers {
cclog.Infof("Run stop tagger for job %d", job.ID)
tagger.Match(job)
}
}
return nil
}

31
internal/tagger/tagger_test.go Normal file
View File

@@ -0,0 +1,31 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package tagger
import (
"testing"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-lib/schema"
)
func TestInit(t *testing.T) {
Init()
}
func TestJobStartCallback(t *testing.T) {
Init()
r := setup(t)
job, err := r.FindByIdDirect(2)
noErr(t, err)
jobs := make([]*schema.Job, 0, 1)
jobs = append(jobs, job)
repository.CallJobStartHooks(jobs)
if !r.HasTag(2, "app", "python") {
t.Errorf("missing tag python")
}
}

35
internal/taskManager/commitJobService.go Normal file
View File

@@ -0,0 +1,35 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package taskManager
import (
"time"
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/repository"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/go-co-op/gocron/v2"
)
func RegisterCommitJobService() {
var frequency string
if config.Keys.CronFrequency != nil && config.Keys.CronFrequency.CommitJobWorker != "" {
frequency = config.Keys.CronFrequency.CommitJobWorker
} else {
frequency = "2m"
}
d, _ := time.ParseDuration(frequency)
cclog.Infof("Register commitJob service with %s interval", frequency)
s.NewJob(gocron.DurationJob(d),
gocron.NewTask(
func() {
start := time.Now()
cclog.Printf("Jobcache sync started at %s", start.Format(time.RFC3339))
jobs, _ := jobRepo.SyncJobs()
repository.CallJobStartHooks(jobs)
cclog.Printf("Jobcache sync and job callbacks are done and took %s", time.Since(start))
}))
}

12
internal/taskManager/compressionService.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package taskManager
@@ -8,13 +8,13 @@ import (
"time"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/go-co-op/gocron/v2"
)
func RegisterCompressionService(compressOlderThan int) {
log.Info("Register compression service")
cclog.Info("Register compression service")
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(05, 0, 0))),
gocron.NewTask(
@@ -26,7 +26,7 @@ func RegisterCompressionService(compressOlderThan int) {
startTime := time.Now().Unix() - int64(compressOlderThan*24*3600)
lastTime := ar.CompressLast(startTime)
if startTime == lastTime {
log.Info("Compression Service - Complete archive run")
cclog.Info("Compression Service - Complete archive run")
jobs, err = jobRepo.FindJobsBetween(0, startTime)
} else {
@@ -34,7 +34,7 @@ func RegisterCompressionService(compressOlderThan int) {
}
if err != nil {
log.Warnf("Error while looking for compression jobs: %v", err)
cclog.Warnf("Error while looking for compression jobs: %v", err)
}
ar.Compress(jobs)
}))

14
internal/taskManager/ldapSyncService.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package taskManager
@@ -8,29 +8,29 @@ import (
"time"
"github.com/ClusterCockpit/cc-backend/internal/auth"
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/go-co-op/gocron/v2"
)
func RegisterLdapSyncService(ds string) {
interval, err := parseDuration(ds)
if err != nil {
log.Warnf("Could not parse duration for sync interval: %v",
cclog.Warnf("Could not parse duration for sync interval: %v",
ds)
return
}
auth := auth.GetAuthInstance()
log.Info("Register LDAP sync service")
cclog.Info("Register LDAP sync service")
s.NewJob(gocron.DurationJob(interval),
gocron.NewTask(
func() {
t := time.Now()
log.Printf("ldap sync started at %s", t.Format(time.RFC3339))
cclog.Printf("ldap sync started at %s", t.Format(time.RFC3339))
if err := auth.LdapAuth.Sync(); err != nil {
log.Errorf("ldap sync failed: %s", err.Error())
cclog.Errorf("ldap sync failed: %s", err.Error())
}
log.Print("ldap sync done")
cclog.Print("ldap sync done")
}))
}

24
internal/taskManager/retentionService.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package taskManager
@@ -8,12 +8,12 @@ import (
"time"
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/go-co-op/gocron/v2"
)
func RegisterRetentionDeleteService(age int, includeDB bool) {
log.Info("Register retention delete service")
cclog.Info("Register retention delete service")
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(04, 0, 0))),
gocron.NewTask(
@@ -21,26 +21,26 @@ func RegisterRetentionDeleteService(age int, includeDB bool) {
startTime := time.Now().Unix() - int64(age*24*3600)
jobs, err := jobRepo.FindJobsBetween(0, startTime)
if err != nil {
log.Warnf("Error while looking for retention jobs: %s", err.Error())
cclog.Warnf("Error while looking for retention jobs: %s", err.Error())
}
archive.GetHandle().CleanUp(jobs)
if includeDB {
cnt, err := jobRepo.DeleteJobsBefore(startTime)
if err != nil {
log.Errorf("Error while deleting retention jobs from db: %s", err.Error())
cclog.Errorf("Error while deleting retention jobs from db: %s", err.Error())
} else {
log.Infof("Retention: Removed %d jobs from db", cnt)
cclog.Infof("Retention: Removed %d jobs from db", cnt)
}
if err = jobRepo.Optimize(); err != nil {
log.Errorf("Error occured in db optimization: %s", err.Error())
cclog.Errorf("Error occured in db optimization: %s", err.Error())
}
}
}))
}
func RegisterRetentionMoveService(age int, includeDB bool, location string) {
log.Info("Register retention move service")
cclog.Info("Register retention move service")
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(04, 0, 0))),
gocron.NewTask(
@@ -48,19 +48,19 @@ func RegisterRetentionMoveService(age int, includeDB bool, location string) {
startTime := time.Now().Unix() - int64(age*24*3600)
jobs, err := jobRepo.FindJobsBetween(0, startTime)
if err != nil {
log.Warnf("Error while looking for retention jobs: %s", err.Error())
cclog.Warnf("Error while looking for retention jobs: %s", err.Error())
}
archive.GetHandle().Move(jobs, location)
if includeDB {
cnt, err := jobRepo.DeleteJobsBefore(startTime)
if err != nil {
log.Errorf("Error while deleting retention jobs from db: %v", err)
cclog.Errorf("Error while deleting retention jobs from db: %v", err)
} else {
log.Infof("Retention: Removed %d jobs from db", cnt)
cclog.Infof("Retention: Removed %d jobs from db", cnt)
}
if err = jobRepo.Optimize(); err != nil {
log.Errorf("Error occured in db optimization: %v", err)
cclog.Errorf("Error occured in db optimization: %v", err)
}
}
}))

8
internal/taskManager/stopJobsExceedTime.go
View File

@@ -1,5 +1,5 @@
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
// All rights reserved.
// All rights reserved. This file is part of cc-backend.
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file.
package taskManager
@@ -8,19 +8,19 @@ import (
"runtime"
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/pkg/log"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/go-co-op/gocron/v2"
)
func RegisterStopJobsExceedTime() {
log.Info("Register undead jobs service")
cclog.Info("Register undead jobs service")
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(03, 0, 0))),
gocron.NewTask(
func() {
err := jobRepo.StopJobsExceedingWalltimeBy(config.Keys.StopJobsExceedingWalltime)
if err != nil {
log.Warnf("Error while looking for jobs exceeding their walltime: %s", err.Error())
cclog.Warnf("Error while looking for jobs exceeding their walltime: %s", err.Error())
}
runtime.GC()
}))

Some files were not shown because too many files have changed in this diff Show More