ClusterCockpit/cc-backend
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-backend synced 2024-11-10 00:47:26 +01:00
Code Issues Packages Releases Activity

BC: new schemas for basically everything

Browse Source
This commit is contained in:
Lou Knauer 2021-12-16 13:17:48 +01:00
parent 7fcc39a144
commit 89333666b3
14 changed files with 1631 additions and 549 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
api/rest.go
View File

@ -33,6 +33,7 @@ func (api *RestApi) MountRoutes(r *mux.Router) {
r.HandleFunc("/api/jobs/tag_job/{id}", api.tagJob).Methods(http.MethodPost, http.MethodPatch)
}
// TODO/FIXME: UPDATE API!
type StartJobApiRequest struct {
JobId int64 `json:"jobId"`
UserId string `json:"userId"`
@ -255,7 +256,7 @@ func (api *RestApi) stopJob(rw http.ResponseWriter, r *http.Request) {
return nil
}
log.Printf("archiving job... (id: %s): clusterId=%s, jobId=%s, userId=%s, startTime=%s, nodes=%v\n", job.ID, job.ClusterID, job.JobID, job.UserID, job.StartTime, job.Nodes)
log.Printf("archiving job... (id: %s): clusterId=%s, jobId=%d, userId=%s, startTime=%s\n", job.ID, job.Cluster, job.JobID, job.User, job.StartTime)
if api.AsyncArchiving {
rw.Header().Add("Content-Type", "application/json")
rw.WriteHeader(http.StatusOK)

6
gqlgen.yml
View File

@ -56,7 +56,7 @@ models:
- github.com/99designs/gqlgen/graphql.Int32
Job:
fields:
tags:
Tags:
resolver: true
JobMetric:
model: "github.com/ClusterCockpit/cc-jobarchive/schema.JobMetric"
@ -68,4 +68,8 @@ models:
model: "github.com/ClusterCockpit/cc-jobarchive/schema.Float"
JobMetricScope:
model: "github.com/ClusterCockpit/cc-jobarchive/schema.MetricScope"
JobResource:
model: "github.com/ClusterCockpit/cc-jobarchive/schema.JobResource"
Accelerator:
model: "github.com/ClusterCockpit/cc-jobarchive/schema.Accelerator"

1567
graph/generated/generated.go
View File

File diff suppressed because it is too large Load Diff

75
graph/model/models_gen.go
View File

@ -38,36 +38,42 @@ type IntRangeOutput struct {
}
type Job struct {
ID string `json:"id"`
JobID string `json:"jobId"`
UserID string `json:"userId"`
ProjectID string `json:"projectId"`
ClusterID string `json:"clusterId"`
StartTime time.Time `json:"startTime"`
Duration int `json:"duration"`
NumNodes int `json:"numNodes"`
Nodes []string `json:"nodes"`
HasProfile bool `json:"hasProfile"`
State JobState `json:"state"`
Tags []*JobTag `json:"tags"`
LoadAvg *float64 `json:"loadAvg"`
MemUsedMax *float64 `json:"memUsedMax"`
FlopsAnyAvg *float64 `json:"flopsAnyAvg"`
MemBwAvg *float64 `json:"memBwAvg"`
NetBwAvg *float64 `json:"netBwAvg"`
FileBwAvg *float64 `json:"fileBwAvg"`
ID string `json:"Id"`
JobID int `json:"JobId"`
User string `json:"User"`
Project string `json:"Project"`
Cluster string `json:"Cluster"`
StartTime time.Time `json:"StartTime"`
Duration int `json:"Duration"`
NumNodes int `json:"NumNodes"`
NumHWThreads int `json:"NumHWThreads"`
NumAcc int `json:"NumAcc"`
Smt int `json:"SMT"`
Exclusive int `json:"Exclusive"`
Partition string `json:"Partition"`
ArrayJobID int `json:"ArrayJobId"`
MonitoringStatus int `json:"MonitoringStatus"`
State JobState `json:"State"`
Tags []*JobTag `json:"Tags"`
Resources []*schema.JobResource `json:"Resources"`
LoadAvg *float64 `json:"LoadAvg"`
MemUsedMax *float64 `json:"MemUsedMax"`
FlopsAnyAvg *float64 `json:"FlopsAnyAvg"`
MemBwAvg *float64 `json:"MemBwAvg"`
NetBwAvg *float64 `json:"NetBwAvg"`
FileBwAvg *float64 `json:"FileBwAvg"`
}
type JobFilter struct {
Tags []string `json:"tags"`
JobID *StringInput `json:"jobId"`
UserID *StringInput `json:"userId"`
ProjectID *StringInput `json:"projectId"`
ClusterID *StringInput `json:"clusterId"`
User *StringInput `json:"user"`
Project *StringInput `json:"project"`
Cluster *StringInput `json:"cluster"`
Duration *IntRange `json:"duration"`
NumNodes *IntRange `json:"numNodes"`
StartTime *TimeRange `json:"startTime"`
IsRunning *bool `json:"isRunning"`
JobState []JobState `json:"jobState"`
FlopsAnyAvg *FloatRange `json:"flopsAnyAvg"`
MemBwAvg *FloatRange `json:"memBwAvg"`
LoadAvg *FloatRange `json:"loadAvg"`
@ -97,13 +103,14 @@ type JobsStatistics struct {
}
type MetricConfig struct {
Name string `json:"name"`
Unit string `json:"unit"`
Sampletime int `json:"sampletime"`
Peak int `json:"peak"`
Normal int `json:"normal"`
Caution int `json:"caution"`
Alert int `json:"alert"`
Name string `json:"Name"`
Unit string `json:"Unit"`
Timestep int `json:"Timestep"`
Peak int `json:"Peak"`
Normal int `json:"Normal"`
Caution int `json:"Caution"`
Alert int `json:"Alert"`
Scope string `json:"Scope"`
}
type MetricFootprints struct {
@ -196,16 +203,24 @@ type JobState string
const (
JobStateRunning JobState = "running"
JobStateCompleted JobState = "completed"
JobStateFailed JobState = "failed"
JobStateCanceled JobState = "canceled"
JobStateStopped JobState = "stopped"
JobStateTimeout JobState = "timeout"
)
var AllJobState = []JobState{
JobStateRunning,
JobStateCompleted,
JobStateFailed,
JobStateCanceled,
JobStateStopped,
JobStateTimeout,
}
func (e JobState) IsValid() bool {
switch e {
case JobStateRunning, JobStateCompleted:
case JobStateRunning, JobStateCompleted, JobStateFailed, JobStateCanceled, JobStateStopped, JobStateTimeout:
return true
}
return false

44
graph/resolver.go
View File

@ -2,6 +2,7 @@ package graph
import (
"context"
"encoding/json"
"errors"
"fmt"
"regexp"
@ -22,7 +23,12 @@ type Resolver struct {
DB *sqlx.DB
}
var JobTableCols []string = []string{"id", "job_id", "user_id", "project_id", "cluster_id", "start_time", "duration", "job_state", "num_nodes", "node_list", "flops_any_avg", "mem_bw_avg", "net_bw_avg", "file_bw_avg", "load_avg"}
var JobTableCols []string = []string{
"id", "job_id", "cluster", "start_time",
"user", "project", "partition", "array_job_id", "duration", "job_state", "resources",
"num_nodes", "num_hwthreads", "num_acc", "smt", "exclusive", "monitoring_status",
"load_avg", "mem_used_max", "flops_any_avg", "mem_bw_avg", "net_bw_avg", "file_bw_avg",
}
type Scannable interface {
Scan(dest ...interface{}) error
@ -30,13 +36,18 @@ type Scannable interface {
// Helper function for scanning jobs with the `jobTableCols` columns selected.
func ScanJob(row Scannable) (*model.Job, error) {
job := &model.Job{HasProfile: true}
job := &model.Job{}
var nodeList string
var rawResources []byte
if err := row.Scan(
&job.ID, &job.JobID, &job.UserID, &job.ProjectID, &job.ClusterID,
&job.StartTime, &job.Duration, &job.State, &job.NumNodes, &nodeList,
&job.FlopsAnyAvg, &job.MemBwAvg, &job.NetBwAvg, &job.FileBwAvg, &job.LoadAvg); err != nil {
&job.ID, &job.JobID, &job.Cluster, &job.StartTime,
&job.User, &job.Project, &job.Partition, &job.ArrayJobID, &job.Duration, &job.State, &rawResources,
&job.NumNodes, &job.NumHWThreads, &job.NumAcc, &job.Smt, &job.Exclusive, &job.MonitoringStatus,
&job.LoadAvg, &job.MemUsedMax, &job.FlopsAnyAvg, &job.MemBwAvg, &job.NetBwAvg, &job.FileBwAvg); err != nil {
return nil, err
}
if err := json.Unmarshal(rawResources, &job.Resources); err != nil {
return nil, err
}
@ -44,7 +55,6 @@ func ScanJob(row Scannable) (*model.Job, error) {
job.Duration = int(time.Since(job.StartTime).Seconds())
}
job.Nodes = strings.Split(nodeList, ",")
return job, nil
}
@ -130,14 +140,14 @@ func buildWhereClause(filter *model.JobFilter, query sq.SelectBuilder) sq.Select
if filter.JobID != nil {
query = buildStringCondition("job.job_id", filter.JobID, query)
}
if filter.UserID != nil {
query = buildStringCondition("job.user_id", filter.UserID, query)
if filter.User != nil {
query = buildStringCondition("job.user", filter.User, query)
}
if filter.ProjectID != nil {
query = buildStringCondition("job.project_id", filter.ProjectID, query)
if filter.Project != nil {
query = buildStringCondition("job.project", filter.Project, query)
}
if filter.ClusterID != nil {
query = buildStringCondition("job.cluster_id", filter.ClusterID, query)
if filter.Cluster != nil {
query = buildStringCondition("job.cluster", filter.Cluster, query)
}
if filter.StartTime != nil {
query = buildTimeCondition("job.start_time", filter.StartTime, query)
@ -145,12 +155,8 @@ func buildWhereClause(filter *model.JobFilter, query sq.SelectBuilder) sq.Select
if filter.Duration != nil {
query = buildIntCondition("job.duration", filter.Duration, query)
}
if filter.IsRunning != nil {
if *filter.IsRunning {
query = query.Where("job.job_state = 'running'")
} else {
query = query.Where("job.job_state = 'completed'")
}
if filter.JobState != nil {
query = query.Where("job.job_state IN ?", filter.JobState)
}
if filter.NumNodes != nil {
query = buildIntCondition("job.num_nodes", filter.NumNodes, query)

128
graph/schema.graphqls
View File

@ -1,78 +1,102 @@
type Job {
id: ID! # Database ID, unique
jobId: String! # ID given to the job by the cluster scheduler
userId: String! # Username
projectId: String! # Project
clusterId: String! # Name of the cluster this job was running on
startTime: Time! # RFC3339 formated string
duration: Int! # For running jobs, the time it has already run
numNodes: Int! # Number of nodes this job was running on
nodes: [String!]! # List of hostnames
hasProfile: Boolean! # TODO: Could be removed?
state: JobState! # State of the job
tags: [JobTag!]! # List of tags this job has
Id: ID! # Database ID, unique
JobId: Int! # ID given to the job by the cluster scheduler
User: String! # Username
Project: String! # Project
Cluster: String! # Name of the cluster this job was running on
StartTime: Time! # RFC3339 formated string
Duration: Int! # For running jobs, the time it has already run
NumNodes: Int! # Number of nodes this job was running on
NumHWThreads: Int!
NumAcc: Int!
SMT: Int!
Exclusive: Int!
Partition: String!
ArrayJobId: Int!
MonitoringStatus: Int!
State: JobState! # State of the job
Tags: [JobTag!]! # List of tags this job has
Resources: [JobResource!]! # List of hosts/hwthreads/gpus/...
# Will be null for running jobs.
loadAvg: Float
memUsedMax: Float
flopsAnyAvg: Float
memBwAvg: Float
netBwAvg: Float
fileBwAvg: Float
LoadAvg: Float
MemUsedMax: Float
FlopsAnyAvg: Float
MemBwAvg: Float
NetBwAvg: Float
FileBwAvg: Float
}
type JobResource {
Hostname: String!
HWThreads: [Int!]
Accelerators: [Accelerator!]
}
type Accelerator {
Id: String!
Type: String!
Model: String!
}
# TODO: Extend by more possible states?
enum JobState {
running
completed
failed
canceled
stopped
timeout
}
type JobTag {
id: ID! # Database ID, unique
tagType: String! # Type
tagName: String! # Name
Id: ID! # Database ID, unique
TagType: String! # Type
TagName: String! # Name
}
type Cluster {
clusterID: String!
processorType: String!
socketsPerNode: Int!
coresPerSocket: Int!
threadsPerCore: Int!
flopRateScalar: Int!
flopRateSimd: Int!
memoryBandwidth: Int!
metricConfig: [MetricConfig!]!
filterRanges: FilterRanges!
ClusterID: String!
ProcessorType: String!
SocketsPerNode: Int!
CoresPerSocket: Int!
ThreadsPerCore: Int!
FlopRateScalar: Int!
FlopRateSimd: Int!
MemoryBandwidth: Int!
MetricConfig: [MetricConfig!]!
FilterRanges: FilterRanges!
}
type MetricConfig {
name: String!
unit: String!
sampletime: Int!
peak: Int!
normal: Int!
caution: Int!
alert: Int!
Name: String!
Unit: String!
Timestep: Int!
Peak: Int!
Normal: Int!
Caution: Int!
Alert: Int!
Scope: String!
}
type JobMetric {
unit: String!
scope: JobMetricScope!
timestep: Int!
series: [JobMetricSeries!]!
Unit: String!
Scope: JobMetricScope!
Timestep: Int!
Series: [JobMetricSeries!]!
}
type JobMetricSeries {
node_id: String!
statistics: JobMetricStatistics
data: [NullableFloat!]!
Hostname: String!
Id: Int
Statistics: JobMetricStatistics
Data: [NullableFloat!]!
}
type JobMetricStatistics {
avg: Float!
min: Float!
max: Float!
Avg: Float!
Min: Float!
Max: Float!
}
type JobMetricWithName {
@ -141,13 +165,13 @@ type FilterRanges {
input JobFilter {
tags: [ID!]
jobId: StringInput
userId: StringInput
projectId: StringInput
clusterId: StringInput
user: StringInput
project: StringInput
cluster: StringInput
duration: IntRange
numNodes: IntRange
startTime: TimeRange
isRunning: Boolean
jobState: [JobState!]
flopsAnyAvg: FloatRange
memBwAvg: FloatRange
loadAvg: FloatRange

9
graph/schema.resolvers.go
View File

@ -15,9 +15,14 @@ import (
"github.com/ClusterCockpit/cc-jobarchive/graph/generated"
"github.com/ClusterCockpit/cc-jobarchive/graph/model"
"github.com/ClusterCockpit/cc-jobarchive/metricdata"
"github.com/ClusterCockpit/cc-jobarchive/schema"
sq "github.com/Masterminds/squirrel"
)
func (r *acceleratorResolver) ID(ctx context.Context, obj *schema.Accelerator) (string, error) {
panic(fmt.Errorf("not implemented"))
}
func (r *jobResolver) Tags(ctx context.Context, obj *model.Job) ([]*model.JobTag, error) {
query := sq.
Select("tag.id", "tag.tag_type", "tag.tag_name").
@ -232,6 +237,9 @@ func (r *queryResolver) NodeMetrics(ctx context.Context, cluster string, nodes [
return res, nil
}
// Accelerator returns generated.AcceleratorResolver implementation.
func (r *Resolver) Accelerator() generated.AcceleratorResolver { return &acceleratorResolver{r} }
// Job returns generated.JobResolver implementation.
func (r *Resolver) Job() generated.JobResolver { return &jobResolver{r} }
@ -241,6 +249,7 @@ func (r *Resolver) Mutation() generated.MutationResolver { return &mutationResol
// Query returns generated.QueryResolver implementation.
func (r *Resolver) Query() generated.QueryResolver { return &queryResolver{r} }
type acceleratorResolver struct{ *Resolver }
type jobResolver struct{ *Resolver }
type mutationResolver struct{ *Resolver }
type queryResolver struct{ *Resolver }

120
init-db.go
View File

@ -8,13 +8,61 @@ import (
"log"
"os"
"path/filepath"
"strings"
"time"
"github.com/ClusterCockpit/cc-jobarchive/schema"
"github.com/jmoiron/sqlx"
)
const JOBS_DB_SCHEMA string = `
DROP TABLE IF EXISTS job;
DROP TABLE IF EXISTS tag;
DROP TABLE IF EXISTS jobtag;
CREATE TABLE job (
id INTEGER PRIMARY KEY AUTOINCREMENT, -- Not needed in sqlite
job_id BIGINT NOT NULL,
cluster VARCHAR(255) NOT NULL,
start_time BITINT NOT NULL,
user VARCHAR(255) NOT NULL,
project VARCHAR(255) NOT NULL,
partition VARCHAR(255) NOT NULL,
array_job_id BIGINT NOT NULL,
duration INT,
job_state VARCHAR(255) CHECK(job_state IN ('running', 'completed', 'failed', 'canceled', 'stopped', 'timeout')) NOT NULL,
meta_data TEXT, -- json, but sqlite has no json type
resources TEXT NOT NULL, -- json, but sqlite has no json type
num_nodes INT NOT NULL,
num_hwthreads INT NOT NULL,
num_acc INT NOT NULL,
smt TINYINT CHECK(smt IN (0, 1 )) NOT NULL DEFAULT 1,
exclusive TINYINT CHECK(exclusive IN (0, 1, 2)) NOT NULL DEFAULT 1,
monitoring_status TINYINT CHECK(monitoring_status IN (0, 1 )) NOT NULL DEFAULT 1,
mem_used_max REAL NOT NULL DEFAULT 0.0,
flops_any_avg REAL NOT NULL DEFAULT 0.0,
mem_bw_avg REAL NOT NULL DEFAULT 0.0,
load_avg REAL NOT NULL DEFAULT 0.0,
net_bw_avg REAL NOT NULL DEFAULT 0.0,
net_data_vol_total REAL NOT NULL DEFAULT 0.0,
file_bw_avg REAL NOT NULL DEFAULT 0.0,
file_data_vol_total REAL NOT NULL DEFAULT 0.0);
CREATE TABLE tag (
id INTEGER PRIMARY KEY,
tag_type VARCHAR(255) NOT NULL,
tag_name VARCHAR(255) NOT NULL);
CREATE TABLE jobtag (
job_id INTEGER,
tag_id INTEGER,
PRIMARY KEY (job_id, tag_id),
FOREIGN KEY (job_id) REFERENCES job (id) ON DELETE CASCADE,
FOREIGN KEY (tag_id) REFERENCES tag (id) ON DELETE CASCADE);
`
// Delete the tables "job", "tag" and "jobtag" from the database and
// repopulate them using the jobs found in `archive`.
func initDB(db *sqlx.DB, archive string) error {
@ -22,39 +70,7 @@ func initDB(db *sqlx.DB, archive string) error {
fmt.Println("Building database...")
// Basic database structure:
_, err := db.Exec(`
DROP TABLE IF EXISTS job;
DROP TABLE IF EXISTS tag;
DROP TABLE IF EXISTS jobtag;
CREATE TABLE job (
id INTEGER PRIMARY KEY,
job_id TEXT,
user_id TEXT,
project_id TEXT,
cluster_id TEXT,
start_time TIMESTAMP,
duration INTEGER,
job_state TEXT,
num_nodes INTEGER,
node_list TEXT,
metadata TEXT,
flops_any_avg REAL,
mem_bw_avg REAL,
net_bw_avg REAL,
file_bw_avg REAL,
load_avg REAL);
CREATE TABLE tag (
id INTEGER PRIMARY KEY,
tag_type TEXT,
tag_name TEXT);
CREATE TABLE jobtag (
job_id INTEGER,
tag_id INTEGER,
PRIMARY KEY (job_id, tag_id),
FOREIGN KEY (job_id) REFERENCES job (id) ON DELETE CASCADE ON UPDATE NO ACTION,
FOREIGN KEY (tag_id) REFERENCES tag (id) ON DELETE CASCADE ON UPDATE NO ACTION);`)
_, err := db.Exec(JOBS_DB_SCHEMA)
if err != nil {
return err
}
@ -64,9 +80,17 @@ func initDB(db *sqlx.DB, archive string) error {
return err
}
insertstmt, err := db.Prepare(`INSERT INTO job
(job_id, user_id, project_id, cluster_id, start_time, duration, job_state, num_nodes, node_list, metadata, flops_any_avg, mem_bw_avg, net_bw_avg, file_bw_avg, load_avg)
VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?);`)
insertstmt, err := db.Prepare(`INSERT INTO job (
job_id, cluster, start_time,
user, project, partition, array_job_id, duration, job_state, meta_data, resources,
num_nodes, num_hwthreads, num_acc, smt, exclusive, monitoring_status,
flops_any_avg, mem_bw_avg
) VALUES (
?, ?, ?,
?, ?, ?, ?, ?, ?, ?, ?,
?, ?, ?, ?, ?, ?,
?, ?
);`)
if err != nil {
return err
}
@ -149,7 +173,7 @@ func initDB(db *sqlx.DB, archive string) error {
// Create indexes after inserts so that they do not
// need to be continually updated.
if _, err := db.Exec(`
CREATE INDEX job_by_user ON job (user_id);
CREATE INDEX job_by_user ON job (user);
CREATE INDEX job_by_starttime ON job (start_time);`); err != nil {
return err
}
@ -167,19 +191,27 @@ func loadJob(tx *sql.Tx, stmt *sql.Stmt, tags map[string]int64, path string) err
}
defer f.Close()
var job schema.JobMeta
var job schema.JobMeta = schema.JobMeta{
Exclusive: 1,
}
if err := json.NewDecoder(bufio.NewReader(f)).Decode(&job); err != nil {
return err
}
// TODO: Other metrics...
flopsAnyAvg := loadJobStat(&job, "flops_any")
memBwAvg := loadJobStat(&job, "mem_bw")
netBwAvg := loadJobStat(&job, "net_bw")
fileBwAvg := loadJobStat(&job, "file_bw")
loadAvg := loadJobStat(&job, "load_one")
res, err := stmt.Exec(job.JobId, job.UserId, job.ProjectId, job.ClusterId, job.StartTime, job.Duration, job.JobState,
job.NumNodes, strings.Join(job.Nodes, ","), nil, flopsAnyAvg, memBwAvg, netBwAvg, fileBwAvg, loadAvg)
resources, err := json.Marshal(job.Resources)
if err != nil {
return err
}
res, err := stmt.Exec(
job.JobId, job.Cluster, job.StartTime,
job.User, job.Project, job.Partition, job.ArrayJobId, job.Duration, job.JobState, job.MetaData, string(resources),
job.NumNodes, job.NumHWThreads, job.NumAcc, job.SMT, job.Exclusive, job.MonitoringStatus,
flopsAnyAvg, memBwAvg)
if err != nil {
return err
}

63
metricdata/archive.go
View File

@ -11,7 +11,6 @@ import (
"path"
"path/filepath"
"strconv"
"strings"
"github.com/ClusterCockpit/cc-jobarchive/config"
"github.com/ClusterCockpit/cc-jobarchive/graph/model"
@ -21,19 +20,14 @@ import (
// For a given job, return the path of the `data.json`/`meta.json` file.
// TODO: Implement Issue ClusterCockpit/ClusterCockpit#97
func getPath(job *model.Job, file string, checkLegacy bool) (string, error) {
id, err := strconv.Atoi(strings.Split(job.JobID, ".")[0])
if err != nil {
return "", err
}
lvl1, lvl2 := fmt.Sprintf("%d", id/1000), fmt.Sprintf("%03d", id%1000)
lvl1, lvl2 := fmt.Sprintf("%d", job.JobID/1000), fmt.Sprintf("%03d", job.JobID%1000)
if !checkLegacy {
return filepath.Join(JobArchivePath, job.ClusterID, lvl1, lvl2, strconv.FormatInt(job.StartTime.Unix(), 10), file), nil
return filepath.Join(JobArchivePath, job.Cluster, lvl1, lvl2, strconv.FormatInt(job.StartTime.Unix(), 10), file), nil
}
legacyPath := filepath.Join(JobArchivePath, job.ClusterID, lvl1, lvl2, file)
legacyPath := filepath.Join(JobArchivePath, job.Cluster, lvl1, lvl2, file)
if _, err := os.Stat(legacyPath); errors.Is(err, os.ErrNotExist) {
return filepath.Join(JobArchivePath, job.ClusterID, lvl1, lvl2, strconv.FormatInt(job.StartTime.Unix(), 10), file), nil
return filepath.Join(JobArchivePath, job.Cluster, lvl1, lvl2, strconv.FormatInt(job.StartTime.Unix(), 10), file), nil
}
return legacyPath, nil
@ -87,13 +81,13 @@ func UpdateTags(job *model.Job, tags []*model.JobTag) error {
f.Close()
metaFile.Tags = make([]struct {
Name string "json:\"name\""
Type string "json:\"type\""
Name string "json:\"Name\""
Type string "json:\"Type\""
}, 0)
for _, tag := range tags {
metaFile.Tags = append(metaFile.Tags, struct {
Name string "json:\"name\""
Type string "json:\"type\""
Name string "json:\"Name\""
Type string "json:\"Type\""
}{
Name: tag.TagName,
Type: tag.TagType,
@ -143,7 +137,7 @@ func ArchiveJob(job *model.Job, ctx context.Context) (*schema.JobMeta, error) {
}
allMetrics := make([]string, 0)
metricConfigs := config.GetClusterConfig(job.ClusterID).MetricConfig
metricConfigs := config.GetClusterConfig(job.Cluster).MetricConfig
for _, mc := range metricConfigs {
allMetrics = append(allMetrics, mc.Name)
}
@ -153,13 +147,13 @@ func ArchiveJob(job *model.Job, ctx context.Context) (*schema.JobMeta, error) {
}
tags := []struct {
Name string `json:"name"`
Type string `json:"type"`
Name string `json:"Name"`
Type string `json:"Type"`
}{}
for _, tag := range job.Tags {
tags = append(tags, struct {
Name string `json:"name"`
Type string `json:"type"`
Name string `json:"Name"`
Type string `json:"Type"`
}{
Name: tag.TagName,
Type: tag.TagType,
@ -167,16 +161,25 @@ func ArchiveJob(job *model.Job, ctx context.Context) (*schema.JobMeta, error) {
}
metaData := &schema.JobMeta{
JobId: job.JobID,
UserId: job.UserID,
ClusterId: job.ClusterID,
NumNodes: job.NumNodes,
JobState: job.State.String(),
StartTime: job.StartTime.Unix(),
Duration: int64(job.Duration),
Nodes: job.Nodes,
Tags: tags,
Statistics: make(map[string]*schema.JobMetaStatistics),
JobId: int64(job.JobID),
User: job.User,
Project: job.Project,
Cluster: job.Cluster,
NumNodes: job.NumNodes,
NumHWThreads: job.NumHWThreads,
NumAcc: job.NumAcc,
Exclusive: int8(job.Exclusive),
MonitoringStatus: int8(job.MonitoringStatus),
SMT: int8(job.Smt),
Partition: job.Partition,
ArrayJobId: job.ArrayJobID,
JobState: string(job.State),
StartTime: job.StartTime.Unix(),
Duration: int64(job.Duration),
Resources: job.Resources,
MetaData: "", // TODO/FIXME: Handle `meta_data`!
Tags: tags,
Statistics: make(map[string]*schema.JobMetaStatistics),
}
for metric, data := range jobData {
@ -188,7 +191,7 @@ func ArchiveJob(job *model.Job, ctx context.Context) (*schema.JobMeta, error) {
}
metaData.Statistics[metric] = &schema.JobMetaStatistics{
Unit: config.GetMetricConfig(job.ClusterID, metric).Unit,
Unit: config.GetMetricConfig(job.Cluster, metric).Unit,
Avg: avg / float64(job.NumNodes),
Min: min,
Max: max,

43
metricdata/cc-metric-store.go
View File

@ -61,8 +61,13 @@ func (ccms *CCMetricStore) doRequest(job *model.Job, suffix string, metrics []st
from, to := job.StartTime.Unix(), job.StartTime.Add(time.Duration(job.Duration)*time.Second).Unix()
reqBody := ApiRequestBody{}
reqBody.Metrics = metrics
for _, node := range job.Nodes {
reqBody.Selectors = append(reqBody.Selectors, []string{job.ClusterID, node})
for _, node := range job.Resources {
if node.Accelerators != nil || node.HWThreads != nil {
// TODO/FIXME:
return nil, errors.New("todo: cc-metric-store resources: Accelerator/HWThreads")
}
reqBody.Selectors = append(reqBody.Selectors, []string{job.Cluster, node.Hostname})
}
reqBodyBytes, err := json.Marshal(reqBody)
@ -86,33 +91,38 @@ func (ccms *CCMetricStore) LoadData(job *model.Job, metrics []string, ctx contex
return nil, err
}
resdata := make([]map[string]ApiMetricData, 0, len(job.Nodes))
resdata := make([]map[string]ApiMetricData, 0, len(job.Resources))
if err := json.NewDecoder(res.Body).Decode(&resdata); err != nil {
return nil, err
}
var jobData schema.JobData = make(schema.JobData)
for _, metric := range metrics {
mc := config.GetMetricConfig(job.ClusterID, metric)
mc := config.GetMetricConfig(job.Cluster, metric)
metricData := &schema.JobMetric{
Scope: "node", // TODO: FIXME: Whatever...
Unit: mc.Unit,
Timestep: mc.Sampletime,
Series: make([]*schema.MetricSeries, 0, len(job.Nodes)),
Timestep: mc.Timestep,
Series: make([]*schema.MetricSeries, 0, len(job.Resources)),
}
for i, node := range job.Nodes {
for i, node := range job.Resources {
if node.Accelerators != nil || node.HWThreads != nil {
// TODO/FIXME:
return nil, errors.New("todo: cc-metric-store resources: Accelerator/HWThreads")
}
data := resdata[i][metric]
if data.Error != nil {
return nil, errors.New(*data.Error)
}
if data.Avg == nil || data.Min == nil || data.Max == nil {
return nil, fmt.Errorf("no data for node '%s' and metric '%s'", node, metric)
return nil, fmt.Errorf("no data for node '%s' and metric '%s'", node.Hostname, metric)
}
metricData.Series = append(metricData.Series, &schema.MetricSeries{
NodeID: node,
Data: data.Data,
Hostname: node.Hostname,
Data: data.Data,
Statistics: &schema.MetricStatistics{
Avg: *data.Avg,
Min: *data.Min,
@ -132,7 +142,7 @@ func (ccms *CCMetricStore) LoadStats(job *model.Job, metrics []string, ctx conte
return nil, err
}
resdata := make([]map[string]ApiStatsData, 0, len(job.Nodes))
resdata := make([]map[string]ApiStatsData, 0, len(job.Resources))
if err := json.NewDecoder(res.Body).Decode(&resdata); err != nil {
return nil, err
}
@ -140,17 +150,22 @@ func (ccms *CCMetricStore) LoadStats(job *model.Job, metrics []string, ctx conte
stats := map[string]map[string]schema.MetricStatistics{}
for _, metric := range metrics {
nodestats := map[string]schema.MetricStatistics{}
for i, node := range job.Nodes {
for i, node := range job.Resources {
if node.Accelerators != nil || node.HWThreads != nil {
// TODO/FIXME:
return nil, errors.New("todo: cc-metric-store resources: Accelerator/HWThreads")
}
data := resdata[i][metric]
if data.Error != nil {
return nil, errors.New(*data.Error)
}
if data.Samples == 0 {
return nil, fmt.Errorf("no data for node '%s' and metric '%s'", node, metric)
return nil, fmt.Errorf("no data for node '%s' and metric '%s'", node.Hostname, metric)
}
nodestats[node] = schema.MetricStatistics{
nodestats[node.Hostname] = schema.MetricStatistics{
Avg: float64(data.Avg),
Min: float64(data.Min),
Max: float64(data.Max),

33
metricdata/influxdb-v2.go
View File

@ -2,6 +2,7 @@ package metricdata
import (
"context"
"errors"
"fmt"
"log"
"os"
@ -46,9 +47,14 @@ func (idb *InfluxDBv2DataRepository) LoadData(job *model.Job, metrics []string,
}
fieldsCond := strings.Join(fieldsConds, " or ")
hostsConds := make([]string, 0, len(job.Nodes))
for _, h := range job.Nodes {
hostsConds = append(hostsConds, fmt.Sprintf(`r.host == "%s"`, h))
hostsConds := make([]string, 0, len(job.Resources))
for _, h := range job.Resources {
if h.HWThreads != nil || h.Accelerators != nil {
// TODO/FIXME...
return nil, errors.New("the InfluxDB metric data repository does not support HWThreads or Accelerators")
}
hostsConds = append(hostsConds, fmt.Sprintf(`r.host == "%s"`, h.Hostname))
}
hostsCond := strings.Join(hostsConds, " or ")
@ -72,18 +78,18 @@ func (idb *InfluxDBv2DataRepository) LoadData(job *model.Job, metrics []string,
field, host := row.Field(), row.ValueByKey("host").(string)
jobMetric, ok := jobData[field]
if !ok {
mc := config.GetMetricConfig(job.ClusterID, field)
mc := config.GetMetricConfig(job.Cluster, field)
jobMetric = &schema.JobMetric{
Scope: "node", // TODO: FIXME: Whatever...
Unit: mc.Unit,
Timestep: mc.Sampletime,
Series: make([]*schema.MetricSeries, 0, len(job.Nodes)),
Timestep: mc.Timestep,
Series: make([]*schema.MetricSeries, 0, len(job.Resources)),
}
jobData[field] = jobMetric
}
currentSeries = &schema.MetricSeries{
NodeID: host,
Hostname: host,
Statistics: nil,
Data: make([]schema.Float, 0),
}
@ -102,7 +108,7 @@ func (idb *InfluxDBv2DataRepository) LoadData(job *model.Job, metrics []string,
jobMetric := jobData[metric]
for node, stats := range nodes {
for _, series := range jobMetric.Series {
if series.NodeID == node {
if series.Hostname == node {
series.Statistics = &stats
}
}
@ -115,9 +121,14 @@ func (idb *InfluxDBv2DataRepository) LoadData(job *model.Job, metrics []string,
func (idb *InfluxDBv2DataRepository) LoadStats(job *model.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.Nodes))
for _, h := range job.Nodes {
hostsConds = append(hostsConds, fmt.Sprintf(`r.host == "%s"`, h))
hostsConds := make([]string, 0, len(job.Resources))
for _, h := range job.Resources {
if h.HWThreads != nil || h.Accelerators != nil {
// TODO/FIXME...
return nil, errors.New("the InfluxDB metric data repository does not support HWThreads or Accelerators")
}
hostsConds = append(hostsConds, fmt.Sprintf(`r.host == "%s"`, h.Hostname))
}
hostsCond := strings.Join(hostsConds, " or ")

8
metricdata/metricdata.go
View File

@ -59,9 +59,9 @@ func Init(jobArchivePath string, disableArchive bool) error {
// Fetches the metric data for a job.
func LoadData(job *model.Job, metrics []string, ctx context.Context) (schema.JobData, error) {
if job.State == model.JobStateRunning || !useArchive {
repo, ok := metricDataRepos[job.ClusterID]
repo, ok := metricDataRepos[job.Cluster]
if !ok {
return nil, fmt.Errorf("no metric data repository configured for '%s'", job.ClusterID)
return nil, fmt.Errorf("no metric data repository configured for '%s'", job.Cluster)
}
return repo.LoadData(job, metrics, ctx)
@ -90,9 +90,9 @@ func LoadAverages(job *model.Job, metrics []string, data [][]schema.Float, ctx c
return loadAveragesFromArchive(job, metrics, data)
}
repo, ok := metricDataRepos[job.ClusterID]
repo, ok := metricDataRepos[job.Cluster]
if !ok {
return fmt.Errorf("no metric data repository configured for '%s'", job.ClusterID)
return fmt.Errorf("no metric data repository configured for '%s'", job.Cluster)
}
stats, err := repo.LoadStats(job, metrics, ctx)

77
schema/metrics.go
View File

@ -9,10 +9,10 @@ import (
type JobData map[string]*JobMetric
type JobMetric struct {
Unit string `json:"unit"`
Scope MetricScope `json:"scope"`
Timestep int `json:"timestep"`
Series []*MetricSeries `json:"series"`
Unit string `json:"Unit"`
Scope MetricScope `json:"Scope"`
Timestep int `json:"Timestep"`
Series []*MetricSeries `json:"Series"`
}
type MetricScope string
@ -41,38 +41,59 @@ func (e MetricScope) MarshalGQL(w io.Writer) {
}
type MetricStatistics struct {
Avg float64 `json:"avg"`
Min float64 `json:"min"`
Max float64 `json:"max"`
Avg float64 `json:"Avg"`
Min float64 `json:"Min"`
Max float64 `json:"Max"`
}
type MetricSeries struct {
NodeID string `json:"node_id"`
Statistics *MetricStatistics `json:"statistics"`
Data []Float `json:"data"`
Hostname string `json:"Hostname"`
Id int `json:"Id"`
Statistics *MetricStatistics `json:"Statistics"`
Data []Float `json:"Data"`
}
type JobMetaStatistics struct {
Unit string `json:"unit"`
Avg float64 `json:"avg"`
Min float64 `json:"min"`
Max float64 `json:"max"`
Unit string `json:"Unit"`
Avg float64 `json:"Avg"`
Min float64 `json:"Min"`
Max float64 `json:"Max"`
}
type Accelerator struct {
ID int `json:"Id"`
Type string `json:"Type"`
Model string `json:"Model"`
}
type JobResource struct {
Hostname string `json:"Hostname"`
HWThreads []int `json:"HWThreads,omitempty"`
Accelerators []Accelerator `json:"Accelerators,omitempty"`
}
// Format of `meta.json` files.
type JobMeta struct {
JobId string `json:"job_id"`
UserId string `json:"user_id"`
ProjectId string `json:"project_id"`
ClusterId string `json:"cluster_id"`
NumNodes int `json:"num_nodes"`
JobState string `json:"job_state"`
StartTime int64 `json:"start_time"`
Duration int64 `json:"duration"`
Nodes []string `json:"nodes"`
Tags []struct {
Name string `json:"name"`
Type string `json:"type"`
} `json:"tags"`
Statistics map[string]*JobMetaStatistics `json:"statistics"`
JobId int64 `json:"JobId"`
User string `json:"User"`
Project string `json:"Project"`
Cluster string `json:"Cluster"`
NumNodes int `json:"NumNodes"`
NumHWThreads int `json:"NumHWThreads"`
NumAcc int `json:"NumAcc"`
Exclusive int8 `json:"Exclusive"`
MonitoringStatus int8 `json:"MonitoringStatus"`
SMT int8 `json:"SMT"`
Partition string `json:"Partition"`
ArrayJobId int `json:"ArrayJobId"`
JobState string `json:"JobState"`
StartTime int64 `json:"StartTime"`
Duration int64 `json:"Duration"`
Resources []*JobResource `json:"Resources"`
MetaData string `json:"MetaData"`
Tags []struct {
Name string `json:"Name"`
Type string `json:"Type"`
} `json:"Tags"`
Statistics map[string]*JobMetaStatistics `json:"Statistics"`
}

4
server.go
View File

@ -308,12 +308,12 @@ func monitoringRoutes(router *mux.Router, resolver *graph.Resolver) {
}
templates.Render(rw, r, "monitoring/job/", &templates.Page{
Title: fmt.Sprintf("Job %s - ClusterCockpit", job.JobID),
Title: fmt.Sprintf("Job %d - ClusterCockpit", job.JobID),
Config: conf,
Infos: map[string]interface{}{
"id": id,
"jobId": job.JobID,
"clusterId": job.ClusterID,
"clusterId": job.Cluster,
},
})
})