Merge pull request #452 from ClusterCockpit/dev

Dev
This commit is contained in:
Jan Eitzinger
2025-12-18 11:28:07 +01:00
committed by GitHub
80 changed files with 4665 additions and 1012 deletions
+1 -1
View File
@@ -13,7 +13,7 @@
/var/checkpoints*
migrateTimestamps.pl
test_ccms_write_api.sh
test_ccms_write_api*
/web/frontend/public/build
/web/frontend/node_modules
+26
View File
@@ -0,0 +1,26 @@
# ClusterCockpit Backend - Agent Guidelines
## Build/Test Commands
- Build: `make` or `go build ./cmd/cc-backend`
- Run all tests: `make test` (runs: `go clean -testcache && go build ./... && go vet ./... && go test ./...`)
- Run single test: `go test -run TestName ./path/to/package`
- Run single test file: `go test ./path/to/package -run TestName`
- Frontend build: `cd web/frontend && npm install && npm run build`
- Generate GraphQL: `make graphql` (uses gqlgen)
- Generate Swagger: `make swagger` (uses swaggo/swag)
## Code Style
- **Formatting**: Use `gofumpt` for all Go files (strict requirement)
- **Copyright header**: All files must include copyright header (see existing files)
- **Package docs**: Document packages with comprehensive package-level comments explaining purpose, usage, configuration
- **Imports**: Standard library first, then external packages, then internal packages (grouped with blank lines)
- **Naming**: Use camelCase for private, PascalCase for exported; descriptive names (e.g., `JobRepository`, `handleError`)
- **Error handling**: Return errors, don't panic; use custom error types where appropriate; log with cclog package
- **Logging**: Use `cclog` package (e.g., `cclog.Errorf()`, `cclog.Warnf()`, `cclog.Debugf()`)
- **Testing**: Use standard `testing` package; use `testify/assert` for assertions; name tests `TestFunctionName`
- **Comments**: Document all exported functions/types with godoc-style comments
- **Structs**: Document fields with inline comments, especially for complex configurations
- **HTTP handlers**: Return proper status codes; use `handleError()` helper for consistent error responses
- **JSON**: Use struct tags for JSON marshaling; `DisallowUnknownFields()` for strict decoding
+2 -2
View File
@@ -50,12 +50,12 @@ frontend:
swagger:
$(info ===> GENERATE swagger)
@go run github.com/swaggo/swag/cmd/swag init --parseDependency -d ./internal/api -g rest.go -o ./api
@go tool github.com/swaggo/swag/cmd/swag init --parseDependency -d ./internal/api -g rest.go -o ./api
@mv ./api/docs.go ./internal/api/docs.go
graphql:
$(info ===> GENERATE graphql)
@go run github.com/99designs/gqlgen
@go tool github.com/99designs/gqlgen
clean:
$(info ===> CLEAN)
+19
View File
@@ -164,6 +164,13 @@ type JobMetricWithName {
metric: JobMetric!
}
type ClusterMetricWithName {
name: String!
unit: Unit
timestep: Int!
data: [NullableFloat!]!
}
type JobMetric {
unit: Unit
timestep: Int!
@@ -267,6 +274,11 @@ type NodeMetrics {
metrics: [JobMetricWithName!]!
}
type ClusterMetrics {
nodeCount: Int!
metrics: [ClusterMetricWithName!]!
}
type NodesResultList {
items: [NodeMetrics!]!
offset: Int
@@ -385,6 +397,13 @@ type Query {
page: PageRequest
resolution: Int
): NodesResultList!
clusterMetrics(
cluster: String!
metrics: [String!]
from: Time!
to: Time!
): ClusterMetrics!
}
type Mutation {
+1 -1
View File
@@ -33,6 +33,6 @@ func cliInit() {
flag.StringVar(&flagDelUser, "del-user", "", "Remove a existing user. Argument format: <username>")
flag.StringVar(&flagGenJWT, "jwt", "", "Generate and print a JWT for the user specified by its `username`")
flag.StringVar(&flagImportJob, "import-job", "", "Import a job. Argument format: `<path-to-meta.json>:<path-to-data.json>,...`")
flag.StringVar(&flagLogLevel, "loglevel", "warn", "Sets the logging level: `[debug, info (default), warn, err, crit]`")
flag.StringVar(&flagLogLevel, "loglevel", "warn", "Sets the logging level: `[debug, info , warn (default), err, crit]`")
flag.Parse()
}
+9 -1
View File
@@ -30,6 +30,7 @@ import (
"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/nats"
"github.com/ClusterCockpit/cc-backend/web"
ccconf "github.com/ClusterCockpit/cc-lib/ccConfig"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
@@ -262,11 +263,18 @@ func generateJWT(authHandle *auth.Authentication, username string) error {
return fmt.Errorf("generating JWT for user '%s': %w", user.Username, err)
}
cclog.Infof("JWT: Successfully generated JWT for user '%s': %s", user.Username, jwt)
cclog.Printf("JWT: Successfully generated JWT for user '%s': %s\n", user.Username, jwt)
return nil
}
func initSubsystems() error {
// Initialize nats client
natsConfig := ccconf.GetPackageConfig("nats")
if err := nats.Init(natsConfig); err != nil {
cclog.Warnf("initializing (optional) nats client: %s", err.Error())
}
nats.Connect()
// Initialize job archive
archiveCfg := ccconf.GetPackageConfig("archive")
if archiveCfg == nil {
+6
View File
@@ -31,6 +31,7 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/graph/generated"
"github.com/ClusterCockpit/cc-backend/internal/memorystore"
"github.com/ClusterCockpit/cc-backend/internal/routerConfig"
"github.com/ClusterCockpit/cc-backend/pkg/nats"
"github.com/ClusterCockpit/cc-backend/web"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/runtimeEnv"
@@ -363,6 +364,11 @@ func (s *Server) Shutdown(ctx context.Context) {
shutdownCtx, cancel := context.WithTimeout(ctx, 30*time.Second)
defer cancel()
nc := nats.GetClient()
if nc != nil {
nc.Close()
}
// First shut down the server gracefully (waiting for all ongoing requests)
if err := s.server.Shutdown(shutdownCtx); err != nil {
cclog.Errorf("Server shutdown error: %v", err)
+16 -1
View File
@@ -29,6 +29,11 @@
"max-age": "2000h"
}
},
"nats": {
"address": "nats://0.0.0.0:4222",
"username": "root",
"password": "root"
},
"clusters": [
{
"name": "fritz",
@@ -86,6 +91,16 @@
"interval": "1h",
"directory": "./var/archive"
},
"retention-in-memory": "48h"
"retention-in-memory": "48h",
"subscriptions": [
{
"subscribe-to": "hpc-nats",
"cluster-tag": "fritz"
},
{
"subscribe-to": "hpc-nats",
"cluster-tag": "alex"
}
]
}
}
+6 -2
View File
@@ -9,8 +9,12 @@
"apiAllowedIPs": ["*"],
"short-running-jobs-duration": 300,
"resampling": {
"trigger": 30,
"resolutions": [600, 300, 120, 60]
"minimumPoints": 600,
"trigger": 180,
"resolutions": [
240,
60
]
}
},
"cron": {
+5
View File
@@ -4,6 +4,11 @@ go 1.24.0
toolchain go1.24.1
tool (
github.com/99designs/gqlgen
github.com/swaggo/swag/cmd/swag
)
require (
github.com/99designs/gqlgen v0.17.84
github.com/ClusterCockpit/cc-lib v1.0.0
+7 -7
View File
@@ -253,7 +253,7 @@ func (api *RestAPI) getCompleteJobByID(rw http.ResponseWriter, r *http.Request)
return
}
job, err = api.JobRepository.FindById(r.Context(), id) // Get Job from Repo by ID
job, err = api.JobRepository.FindByID(r.Context(), id) // Get Job from Repo by ID
} else {
handleError(fmt.Errorf("the parameter 'id' is required"), http.StatusBadRequest, rw)
return
@@ -346,7 +346,7 @@ func (api *RestAPI) getJobByID(rw http.ResponseWriter, r *http.Request) {
return
}
job, err = api.JobRepository.FindById(r.Context(), id)
job, err = api.JobRepository.FindByID(r.Context(), id)
} else {
handleError(errors.New("the parameter 'id' is required"), http.StatusBadRequest, rw)
return
@@ -445,7 +445,7 @@ func (api *RestAPI) editMeta(rw http.ResponseWriter, r *http.Request) {
return
}
job, err := api.JobRepository.FindById(r.Context(), id)
job, err := api.JobRepository.FindByID(r.Context(), id)
if err != nil {
handleError(fmt.Errorf("finding job failed: %w", err), http.StatusNotFound, rw)
return
@@ -493,7 +493,7 @@ func (api *RestAPI) tagJob(rw http.ResponseWriter, r *http.Request) {
return
}
job, err := api.JobRepository.FindById(r.Context(), id)
job, err := api.JobRepository.FindByID(r.Context(), id)
if err != nil {
handleError(fmt.Errorf("finding job failed: %w", err), http.StatusNotFound, rw)
return
@@ -557,7 +557,7 @@ func (api *RestAPI) removeTagJob(rw http.ResponseWriter, r *http.Request) {
return
}
job, err := api.JobRepository.FindById(r.Context(), id)
job, err := api.JobRepository.FindByID(r.Context(), id)
if err != nil {
handleError(fmt.Errorf("finding job failed: %w", err), http.StatusNotFound, rw)
return
@@ -796,7 +796,7 @@ func (api *RestAPI) deleteJobByID(rw http.ResponseWriter, r *http.Request) {
return
}
err = api.JobRepository.DeleteJobById(id)
err = api.JobRepository.DeleteJobByID(id)
} else {
handleError(errors.New("the parameter 'id' is required"), http.StatusBadRequest, rw)
return
@@ -852,7 +852,7 @@ func (api *RestAPI) deleteJobByRequest(rw http.ResponseWriter, r *http.Request)
return
}
err = api.JobRepository.DeleteJobById(*job.ID)
err = api.JobRepository.DeleteJobByID(*job.ID)
if err != nil {
handleError(fmt.Errorf("deleting job failed: %w", err), http.StatusUnprocessableEntity, rw)
return
+231
View File
@@ -0,0 +1,231 @@
// 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 (
"bytes"
"database/sql"
"encoding/json"
"sync"
"time"
"github.com/ClusterCockpit/cc-backend/internal/archiver"
"github.com/ClusterCockpit/cc-backend/internal/config"
"github.com/ClusterCockpit/cc-backend/internal/importer"
"github.com/ClusterCockpit/cc-backend/internal/repository"
"github.com/ClusterCockpit/cc-backend/pkg/nats"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
)
// NatsAPI provides NATS subscription-based handlers for Job and Node operations.
// It mirrors the functionality of the REST API but uses NATS messaging.
type NatsAPI struct {
JobRepository *repository.JobRepository
// RepositoryMutex protects job creation operations from race conditions
// when checking for duplicate jobs during startJob calls.
RepositoryMutex sync.Mutex
}
// NewNatsAPI creates a new NatsAPI instance with default dependencies.
func NewNatsAPI() *NatsAPI {
return &NatsAPI{
JobRepository: repository.GetJobRepository(),
}
}
// StartSubscriptions registers all NATS subscriptions for Job and Node APIs.
// Returns an error if the NATS client is not available or subscription fails.
func (api *NatsAPI) StartSubscriptions() error {
client := nats.GetClient()
if client == nil {
cclog.Warn("NATS client not available, skipping API subscriptions")
return nil
}
if config.Keys.APISubjects != nil {
s := config.Keys.APISubjects
if err := client.Subscribe(s.SubjectJobStart, api.handleStartJob); err != nil {
return err
}
if err := client.Subscribe(s.SubjectJobStop, api.handleStopJob); err != nil {
return err
}
if err := client.Subscribe(s.SubjectNodeState, api.handleNodeState); err != nil {
return err
}
cclog.Info("NATS API subscriptions started")
}
return nil
}
// handleStartJob processes job start messages received via NATS.
// Expected JSON payload follows the schema.Job structure.
func (api *NatsAPI) handleStartJob(subject string, data []byte) {
req := schema.Job{
Shared: "none",
MonitoringStatus: schema.MonitoringStatusRunningOrArchiving,
}
dec := json.NewDecoder(bytes.NewReader(data))
dec.DisallowUnknownFields()
if err := dec.Decode(&req); err != nil {
cclog.Errorf("NATS %s: parsing request failed: %v", subject, err)
return
}
cclog.Debugf("NATS %s: %s", subject, req.GoString())
req.State = schema.JobStateRunning
if err := importer.SanityChecks(&req); err != nil {
cclog.Errorf("NATS %s: sanity check failed: %v", subject, err)
return
}
var unlockOnce sync.Once
api.RepositoryMutex.Lock()
defer unlockOnce.Do(api.RepositoryMutex.Unlock)
jobs, err := api.JobRepository.FindAll(&req.JobID, &req.Cluster, nil)
if err != nil && err != sql.ErrNoRows {
cclog.Errorf("NATS %s: checking for duplicate failed: %v", subject, err)
return
}
if err == nil {
for _, job := range jobs {
if (req.StartTime - job.StartTime) < secondsPerDay {
cclog.Errorf("NATS %s: job with jobId %d, cluster %s already exists (dbid: %d)",
subject, req.JobID, req.Cluster, job.ID)
return
}
}
}
id, err := api.JobRepository.Start(&req)
if err != nil {
cclog.Errorf("NATS %s: insert into database failed: %v", subject, err)
return
}
unlockOnce.Do(api.RepositoryMutex.Unlock)
for _, tag := range req.Tags {
if _, err := api.JobRepository.AddTagOrCreate(nil, id, tag.Type, tag.Name, tag.Scope); err != nil {
cclog.Errorf("NATS %s: adding tag to new job %d failed: %v", subject, id, err)
return
}
}
cclog.Infof("NATS: new job (id: %d): cluster=%s, jobId=%d, user=%s, startTime=%d",
id, req.Cluster, req.JobID, req.User, req.StartTime)
}
// handleStopJob processes job stop messages received via NATS.
// Expected JSON payload follows the StopJobAPIRequest structure.
func (api *NatsAPI) handleStopJob(subject string, data []byte) {
var req StopJobAPIRequest
dec := json.NewDecoder(bytes.NewReader(data))
dec.DisallowUnknownFields()
if err := dec.Decode(&req); err != nil {
cclog.Errorf("NATS %s: parsing request failed: %v", subject, err)
return
}
if req.JobID == nil {
cclog.Errorf("NATS %s: the field 'jobId' is required", subject)
return
}
job, err := api.JobRepository.Find(req.JobID, req.Cluster, req.StartTime)
if err != nil {
cachedJob, cachedErr := api.JobRepository.FindCached(req.JobID, req.Cluster, req.StartTime)
if cachedErr != nil {
cclog.Errorf("NATS %s: finding job failed: %v (cached lookup also failed: %v)",
subject, err, cachedErr)
return
}
job = cachedJob
}
if job.State != schema.JobStateRunning {
cclog.Errorf("NATS %s: jobId %d (id %d) on %s: job has already been stopped (state is: %s)",
subject, job.JobID, job.ID, job.Cluster, job.State)
return
}
if job.StartTime > req.StopTime {
cclog.Errorf("NATS %s: jobId %d (id %d) on %s: stopTime %d must be >= startTime %d",
subject, job.JobID, job.ID, job.Cluster, req.StopTime, job.StartTime)
return
}
if req.State != "" && !req.State.Valid() {
cclog.Errorf("NATS %s: jobId %d (id %d) on %s: invalid job state: %#v",
subject, job.JobID, job.ID, job.Cluster, req.State)
return
} else if req.State == "" {
req.State = schema.JobStateCompleted
}
job.Duration = int32(req.StopTime - job.StartTime)
job.State = req.State
api.JobRepository.Mutex.Lock()
defer api.JobRepository.Mutex.Unlock()
if err := api.JobRepository.Stop(*job.ID, job.Duration, job.State, job.MonitoringStatus); err != nil {
if err := api.JobRepository.StopCached(*job.ID, job.Duration, job.State, job.MonitoringStatus); err != nil {
cclog.Errorf("NATS %s: jobId %d (id %d) on %s: marking job as '%s' failed: %v",
subject, job.JobID, job.ID, job.Cluster, job.State, err)
return
}
}
cclog.Infof("NATS: archiving job (dbid: %d): cluster=%s, jobId=%d, user=%s, startTime=%d, duration=%d, state=%s",
job.ID, job.Cluster, job.JobID, job.User, job.StartTime, job.Duration, job.State)
if job.MonitoringStatus == schema.MonitoringStatusDisabled {
return
}
archiver.TriggerArchiving(job)
}
// handleNodeState processes node state update messages received via NATS.
// Expected JSON payload follows the UpdateNodeStatesRequest structure.
func (api *NatsAPI) handleNodeState(subject string, data []byte) {
var req UpdateNodeStatesRequest
dec := json.NewDecoder(bytes.NewReader(data))
dec.DisallowUnknownFields()
if err := dec.Decode(&req); err != nil {
cclog.Errorf("NATS %s: parsing request failed: %v", subject, err)
return
}
repo := repository.GetNodeRepository()
for _, node := range req.Nodes {
state := determineState(node.States)
nodeState := schema.NodeStateDB{
TimeStamp: time.Now().Unix(),
NodeState: state,
CpusAllocated: node.CpusAllocated,
MemoryAllocated: node.MemoryAllocated,
GpusAllocated: node.GpusAllocated,
HealthState: schema.MonitoringStateFull,
JobsRunning: node.JobsRunning,
}
repo.UpdateNodeState(node.Hostname, req.Cluster, &nodeState)
}
cclog.Debugf("NATS %s: updated %d node states for cluster %s", subject, len(req.Nodes), req.Cluster)
}
+8
View File
@@ -22,6 +22,8 @@ type ProgramConfig struct {
// Addresses from which secured admin API endpoints can be reached, can be wildcard "*"
APIAllowedIPs []string `json:"apiAllowedIPs"`
APISubjects *NATSConfig `json:"apiSubjects"`
// Drop root permissions once .env was read and the port was taken.
User string `json:"user"`
Group string `json:"group"`
@@ -87,6 +89,12 @@ type ResampleConfig struct {
Trigger int `json:"trigger"`
}
type NATSConfig struct {
SubjectJobStart string `json:"subjectJobStart"`
SubjectJobStop string `json:"subjectJobStop"`
SubjectNodeState string `json:"subjectNodeState"`
}
type IntRange struct {
From int `json:"from"`
To int `json:"to"`
File diff suppressed because it is too large Load Diff
+12
View File
@@ -13,6 +13,18 @@ import (
"github.com/ClusterCockpit/cc-lib/schema"
)
type ClusterMetricWithName struct {
Name string `json:"name"`
Unit *schema.Unit `json:"unit,omitempty"`
Timestep int `json:"timestep"`
Data []schema.Float `json:"data"`
}
type ClusterMetrics struct {
NodeCount int `json:"nodeCount"`
Metrics []*ClusterMetricWithName `json:"metrics"`
}
type Count struct {
Name string `json:"name"`
Count int `json:"count"`
+100 -134
View File
@@ -1,13 +1,15 @@
package graph
// This file will be automatically regenerated based on the schema, any resolver implementations
// This file will be automatically regenerated based on the schema, any resolver
// implementations
// will be copied through when generating and any unknown code will be moved to the end.
// Code generated by github.com/99designs/gqlgen version v0.17.81
// Code generated by github.com/99designs/gqlgen version v0.17.84
import (
"context"
"errors"
"fmt"
"math"
"regexp"
"slices"
"strconv"
@@ -376,7 +378,7 @@ func (r *queryResolver) Node(ctx context.Context, id string) (*schema.Node, erro
cclog.Warn("Error while parsing job id")
return nil, err
}
return repo.GetNodeById(numericId, false)
return repo.GetNodeByID(numericId, false)
}
// Nodes is the resolver for the nodes field.
@@ -442,7 +444,7 @@ func (r *queryResolver) Job(ctx context.Context, id string) (*schema.Job, error)
return nil, err
}
job, err := r.Repo.FindById(ctx, numericId)
job, err := r.Repo.FindByID(ctx, numericId)
if err != nil {
cclog.Warn("Error while finding job by id")
return nil, err
@@ -804,140 +806,24 @@ func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, sub
return nil, errors.New("you need to be administrator or support staff for this query")
}
nodeRepo := repository.GetNodeRepository()
nodes, stateMap, countNodes, hasNextPage, nerr := nodeRepo.GetNodesForList(ctx, cluster, subCluster, stateFilter, nodeFilter, page)
if nerr != nil {
return nil, errors.New("Could not retrieve node list required for resolving NodeMetricsList")
}
if metrics == nil {
for _, mc := range archive.GetCluster(cluster).MetricConfig {
metrics = append(metrics, mc.Name)
}
}
// Build Filters
queryFilters := make([]*model.NodeFilter, 0)
if cluster != "" {
queryFilters = append(queryFilters, &model.NodeFilter{Cluster: &model.StringInput{Eq: &cluster}})
}
if subCluster != "" {
queryFilters = append(queryFilters, &model.NodeFilter{Subcluster: &model.StringInput{Eq: &subCluster}})
}
if nodeFilter != "" && stateFilter != "notindb" {
queryFilters = append(queryFilters, &model.NodeFilter{Hostname: &model.StringInput{Contains: &nodeFilter}})
}
if stateFilter != "all" && stateFilter != "notindb" {
var queryState schema.SchedulerState = schema.SchedulerState(stateFilter)
queryFilters = append(queryFilters, &model.NodeFilter{SchedulerState: &queryState})
}
// if healthFilter != "all" {
// filters = append(filters, &model.NodeFilter{HealthState: &healthFilter})
// }
// Special Case: Disable Paging for missing nodes filter, save IPP for later
var backupItems int
if stateFilter == "notindb" {
backupItems = page.ItemsPerPage
page.ItemsPerPage = -1
}
// Query Nodes From DB
nodeRepo := repository.GetNodeRepository()
rawNodes, serr := nodeRepo.QueryNodes(ctx, queryFilters, page, nil) // Order not Used
if serr != nil {
cclog.Warn("error while loading node database data (Resolver.NodeMetricsList)")
return nil, serr
}
// Intermediate Node Result Info
nodes := make([]string, 0)
stateMap := make(map[string]string)
for _, node := range rawNodes {
nodes = append(nodes, node.Hostname)
stateMap[node.Hostname] = string(node.NodeState)
}
// Setup Vars
var countNodes int
var cerr error
var hasNextPage bool
// Special Case: Find Nodes not in DB node table but in metricStore only
if stateFilter == "notindb" {
// Reapply Original Paging
page.ItemsPerPage = backupItems
// Get Nodes From Topology
var topoNodes []string
if subCluster != "" {
scNodes := archive.NodeLists[cluster][subCluster]
topoNodes = scNodes.PrintList()
} else {
subClusterNodeLists := archive.NodeLists[cluster]
for _, nodeList := range subClusterNodeLists {
topoNodes = append(topoNodes, nodeList.PrintList()...)
}
}
// Compare to all nodes from cluster/subcluster in DB
var missingNodes []string
for _, scanNode := range topoNodes {
if !slices.Contains(nodes, scanNode) {
missingNodes = append(missingNodes, scanNode)
}
}
// Filter nodes by name
if nodeFilter != "" {
filteredNodesByName := []string{}
for _, missingNode := range missingNodes {
if strings.Contains(missingNode, nodeFilter) {
filteredNodesByName = append(filteredNodesByName, missingNode)
}
}
missingNodes = filteredNodesByName
}
// Sort Missing Nodes Alphanumerically
slices.Sort(missingNodes)
// Total Missing
countNodes = len(missingNodes)
// Apply paging
if countNodes > page.ItemsPerPage {
start := (page.Page - 1) * page.ItemsPerPage
end := start + page.ItemsPerPage
if end > countNodes {
end = countNodes
hasNextPage = false
} else {
hasNextPage = true
}
nodes = missingNodes[start:end]
} else {
nodes = missingNodes
}
} else {
// DB Nodes: Count and Find Next Page
countNodes, cerr = nodeRepo.CountNodes(ctx, queryFilters)
if cerr != nil {
cclog.Warn("error while counting node database data (Resolver.NodeMetricsList)")
return nil, cerr
}
// Example Page 4 @ 10 IpP : Does item 41 exist?
// Minimal Page 41 @ 1 IpP : If len(result) is 1, Page 5 exists.
nextPage := &model.PageRequest{
ItemsPerPage: 1,
Page: ((page.Page * page.ItemsPerPage) + 1),
}
nextNodes, err := nodeRepo.QueryNodes(ctx, queryFilters, nextPage, nil) // Order not Used
if err != nil {
cclog.Warn("Error while querying next nodes")
return nil, err
}
hasNextPage = len(nextNodes) == 1
}
// Load Metric Data For Specified Nodes Only
data, err := metricDataDispatcher.LoadNodeListData(cluster, subCluster, nodes, metrics, scopes, *resolution, from, to, ctx)
if err != nil {
cclog.Warn("error while loading node data (Resolver.NodeMetricsList")
return nil, err
}
// Build Result
nodeMetricsList := make([]*model.NodeMetrics, 0, len(data))
for hostname, metrics := range data {
host := &model.NodeMetrics{
@@ -963,7 +849,6 @@ func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, sub
nodeMetricsList = append(nodeMetricsList, host)
}
// Final Return
nodeMetricsListResult := &model.NodesResultList{
Items: nodeMetricsList,
TotalNodes: &countNodes,
@@ -973,6 +858,85 @@ func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, sub
return nodeMetricsListResult, nil
}
// ClusterMetrics is the resolver for the clusterMetrics field.
func (r *queryResolver) ClusterMetrics(ctx context.Context, cluster string, metrics []string, from time.Time, to time.Time) (*model.ClusterMetrics, error) {
user := repository.GetUserFromContext(ctx)
if user != nil && !user.HasAnyRole([]schema.Role{schema.RoleAdmin, schema.RoleSupport}) {
return nil, errors.New("you need to be administrator or support staff for this query")
}
if metrics == nil {
for _, mc := range archive.GetCluster(cluster).MetricConfig {
metrics = append(metrics, mc.Name)
}
}
// 'nodes' == nil -> Defaults to all nodes of cluster for existing query workflow
scopes := []schema.MetricScope{"node"}
data, err := metricDataDispatcher.LoadNodeData(cluster, metrics, nil, scopes, from, to, ctx)
if err != nil {
cclog.Warn("error while loading node data")
return nil, err
}
clusterMetricData := make([]*model.ClusterMetricWithName, 0)
clusterMetrics := model.ClusterMetrics{NodeCount: 0, Metrics: clusterMetricData}
collectorTimestep := make(map[string]int)
collectorUnit := make(map[string]schema.Unit)
collectorData := make(map[string][]schema.Float)
for _, metrics := range data {
clusterMetrics.NodeCount += 1
for metric, scopedMetrics := range metrics {
_, ok := collectorData[metric]
if !ok {
collectorData[metric] = make([]schema.Float, 0)
for _, scopedMetric := range scopedMetrics {
// Collect Info
collectorTimestep[metric] = scopedMetric.Timestep
collectorUnit[metric] = scopedMetric.Unit
// Collect Initial Data
for _, ser := range scopedMetric.Series {
for _, val := range ser.Data {
collectorData[metric] = append(collectorData[metric], val)
}
}
}
} else {
// Sum up values by index
for _, scopedMetric := range scopedMetrics {
// For This Purpose (Cluster_Wide-Sum of Node Metrics) OK
for _, ser := range scopedMetric.Series {
for i, val := range ser.Data {
collectorData[metric][i] += val
}
}
}
}
}
}
for metricName, data := range collectorData {
cu := collectorUnit[metricName]
roundedData := make([]schema.Float, 0)
for _, val := range data {
roundedData = append(roundedData, schema.Float((math.Round(float64(val)*100.0) / 100.0)))
}
cm := model.ClusterMetricWithName{
Name: metricName,
Unit: &cu,
Timestep: collectorTimestep[metricName],
Data: roundedData,
}
clusterMetrics.Metrics = append(clusterMetrics.Metrics, &cm)
}
return &clusterMetrics, nil
}
// NumberOfNodes is the resolver for the numberOfNodes field.
func (r *subClusterResolver) NumberOfNodes(ctx context.Context, obj *schema.SubCluster) (int, error) {
nodeList, err := archive.ParseNodeList(obj.Nodes)
@@ -1003,10 +967,12 @@ func (r *Resolver) Query() generated.QueryResolver { return &queryResolver{r} }
// SubCluster returns generated.SubClusterResolver implementation.
func (r *Resolver) SubCluster() generated.SubClusterResolver { return &subClusterResolver{r} }
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 }
type (
clusterResolver struct{ *Resolver }
jobResolver struct{ *Resolver }
metricValueResolver struct{ *Resolver }
mutationResolver struct{ *Resolver }
nodeResolver struct{ *Resolver }
queryResolver struct{ *Resolver }
subClusterResolver struct{ *Resolver }
)
+3 -7
View File
@@ -2,12 +2,14 @@
// 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
import (
"context"
"fmt"
"math"
"slices"
"github.com/99designs/gqlgen/graphql"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
@@ -185,11 +187,5 @@ func (r *queryResolver) jobsFootprints(ctx context.Context, filter []*model.JobF
func requireField(ctx context.Context, name string) bool {
fields := graphql.CollectAllFields(ctx)
for _, f := range fields {
if f == name {
return true
}
}
return false
return slices.Contains(fields, name)
}
+7 -7
View File
@@ -111,18 +111,22 @@ func InitDB() error {
continue
}
id, err := r.TransactionAddNamed(t,
id, jobErr := r.TransactionAddNamed(t,
repository.NamedJobInsert, jobMeta)
if err != nil {
cclog.Errorf("repository initDB(): %v", err)
if jobErr != nil {
cclog.Errorf("repository initDB(): %v", jobErr)
errorOccured++
continue
}
// Job successfully inserted, increment counter
i += 1
for _, tag := range jobMeta.Tags {
tagstr := tag.Name + ":" + tag.Type
tagID, ok := tags[tagstr]
if !ok {
var err error
tagID, err = r.TransactionAdd(t,
addTagQuery,
tag.Name, tag.Type)
@@ -138,10 +142,6 @@ func InitDB() error {
setTagQuery,
id, tagID)
}
if err == nil {
i += 1
}
}
if errorOccured > 0 {
+13 -3
View File
@@ -6,12 +6,18 @@
package memorystore
import (
"errors"
"math"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/ClusterCockpit/cc-lib/util"
)
var (
ErrInvalidTimeRange = errors.New("[METRICSTORE]> invalid time range: 'from' must be before 'to'")
ErrEmptyCluster = errors.New("[METRICSTORE]> cluster name cannot be empty")
)
type APIMetricData struct {
Error *string `json:"error,omitempty"`
Data schema.FloatArray `json:"data,omitempty"`
@@ -109,10 +115,14 @@ func (data *APIMetricData) PadDataWithNull(ms *MemoryStore, from, to int64, metr
}
func FetchData(req APIQueryRequest) (*APIQueryResponse, error) {
req.WithData = true
req.WithData = true
req.WithData = true
if req.From > req.To {
return nil, ErrInvalidTimeRange
}
if req.Cluster == "" && req.ForAllNodes != nil {
return nil, ErrEmptyCluster
}
req.WithData = true
ms := GetMemoryStore()
response := APIQueryResponse{
+15 -10
View File
@@ -32,17 +32,14 @@ func Archiving(wg *sync.WaitGroup, ctx context.Context) {
return
}
ticks := func() <-chan time.Time {
if d <= 0 {
return nil
}
return time.NewTicker(d).C
}()
ticker := time.NewTicker(d)
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return
case <-ticks:
case <-ticker.C:
t := time.Now().Add(-d)
cclog.Infof("[METRICSTORE]> start archiving checkpoints (older than %s)...", t.Format(time.RFC3339))
n, err := ArchiveCheckpoints(Keys.Checkpoints.RootDir,
@@ -165,23 +162,31 @@ func archiveCheckpoints(dir string, archiveDir string, from int64, deleteInstead
n := 0
for _, checkpoint := range files {
// Use closure to ensure file is closed immediately after use,
// avoiding file descriptor leak from defer in loop
err := func() error {
filename := filepath.Join(dir, checkpoint)
r, err := os.Open(filename)
if err != nil {
return n, err
return err
}
defer r.Close()
w, err := zw.Create(checkpoint)
if err != nil {
return n, err
return err
}
if _, err = io.Copy(w, r); err != nil {
return n, err
return err
}
if err = os.Remove(filename); err != nil {
return err
}
return nil
}()
if err != nil {
return n, err
}
n += 1
+7 -12
View File
@@ -24,9 +24,8 @@ import (
"github.com/linkedin/goavro/v2"
)
var NumAvroWorkers int = 4
var NumAvroWorkers int = DefaultAvroWorkers
var startUp bool = true
var ErrNoNewData error = errors.New("no data in the pool")
func (as *AvroStore) ToCheckpoint(dir string, dumpAll bool) (int, error) {
levels := make([]*AvroLevel, 0)
@@ -464,19 +463,15 @@ func generateRecord(data map[string]schema.Float) map[string]any {
}
func correctKey(key string) string {
// Replace any invalid characters in the key
// For example, replace spaces with underscores
key = strings.ReplaceAll(key, ":", "___")
key = strings.ReplaceAll(key, ".", "__")
key = strings.ReplaceAll(key, "_", "_0x5F_")
key = strings.ReplaceAll(key, ":", "_0x3A_")
key = strings.ReplaceAll(key, ".", "_0x2E_")
return key
}
func ReplaceKey(key string) string {
// Replace any invalid characters in the key
// For example, replace spaces with underscores
key = strings.ReplaceAll(key, "___", ":")
key = strings.ReplaceAll(key, "__", ".")
key = strings.ReplaceAll(key, "_0x2E_", ".")
key = strings.ReplaceAll(key, "_0x3A_", ":")
key = strings.ReplaceAll(key, "_0x5F_", "_")
return key
}
+3 -3
View File
@@ -42,7 +42,7 @@ func DataStaging(wg *sync.WaitGroup, ctx context.Context) {
metricName := ""
for _, selectorName := range val.Selector {
metricName += selectorName + Delimiter
metricName += selectorName + SelectorDelimiter
}
metricName += val.MetricName
@@ -54,7 +54,7 @@ func DataStaging(wg *sync.WaitGroup, ctx context.Context) {
var selector []string
selector = append(selector, val.Cluster, val.Node, strconv.FormatInt(freq, 10))
if !testEq(oldSelector, selector) {
if !stringSlicesEqual(oldSelector, selector) {
// Get the Avro level for the metric
avroLevel = avroStore.root.findAvroLevelOrCreate(selector)
@@ -71,7 +71,7 @@ func DataStaging(wg *sync.WaitGroup, ctx context.Context) {
}()
}
func testEq(a, b []string) bool {
func stringSlicesEqual(a, b []string) bool {
if len(a) != len(b) {
return false
}
+4 -5
View File
@@ -13,12 +13,11 @@ import (
var (
LineProtocolMessages = make(chan *AvroStruct)
Delimiter = "ZZZZZ"
// SelectorDelimiter separates hierarchical selector components in metric names for Avro encoding
SelectorDelimiter = "_SEL_"
)
// CheckpointBufferMinutes should always be in minutes.
// Its controls the amount of data to hold for given amount of time.
var CheckpointBufferMinutes = 3
var CheckpointBufferMinutes = DefaultCheckpointBufferMin
type AvroStruct struct {
MetricName string
@@ -73,7 +72,7 @@ func (l *AvroLevel) findAvroLevelOrCreate(selector []string) *AvroLevel {
}
}
// The level does not exist, take write lock for unqiue access:
// The level does not exist, take write lock for unique access:
l.lock.Lock()
// While this thread waited for the write lock, another thread
// could have created the child node.
+3 -11
View File
@@ -12,15 +12,12 @@ import (
"github.com/ClusterCockpit/cc-lib/schema"
)
// Default buffer capacity.
// `buffer.data` will only ever grow up to it's capacity and a new link
// BufferCap is the default buffer capacity.
// buffer.data will only ever grow up to its capacity and a new link
// in the buffer chain will be created if needed so that no copying
// of data or reallocation needs to happen on writes.
const (
BufferCap int = 512
)
const BufferCap int = DefaultBufferCapacity
// So that we can reuse allocations
var bufferPool sync.Pool = sync.Pool{
New: func() any {
return &buffer{
@@ -75,7 +72,6 @@ func (b *buffer) write(ts int64, value schema.Float) (*buffer, error) {
newbuf := newBuffer(ts, b.frequency)
newbuf.prev = b
b.next = newbuf
b.close()
b = newbuf
idx = 0
}
@@ -103,8 +99,6 @@ func (b *buffer) firstWrite() int64 {
return b.start + (b.frequency / 2)
}
func (b *buffer) close() {}
// Return all known values from `from` to `to`. Gaps of information are represented as NaN.
// Simple linear interpolation is done between the two neighboring cells if possible.
// If values at the start or end are missing, instead of NaN values, the second and thrid
@@ -139,8 +133,6 @@ func (b *buffer) read(from, to int64, data []schema.Float) ([]schema.Float, int6
data[i] += schema.NaN
} else if t < b.start {
data[i] += schema.NaN
// } else if b.data[idx].IsNaN() {
// data[i] += interpolate(idx, b.data)
} else {
data[i] += b.data[idx]
}
+12 -34
View File
@@ -28,15 +28,10 @@ import (
"github.com/linkedin/goavro/v2"
)
// File operation constants
const (
// CheckpointFilePerms defines default permissions for checkpoint files
CheckpointFilePerms = 0o644
// CheckpointDirPerms defines default permissions for checkpoint directories
CheckpointDirPerms = 0o755
// GCTriggerInterval determines how often GC is forced during checkpoint loading
// GC is triggered every GCTriggerInterval*NumWorkers loaded hosts
GCTriggerInterval = 100
GCTriggerInterval = DefaultGCTriggerInterval
)
// Whenever changed, update MarshalJSON as well!
@@ -71,17 +66,14 @@ func Checkpointing(wg *sync.WaitGroup, ctx context.Context) {
return
}
ticks := func() <-chan time.Time {
if d <= 0 {
return nil
}
return time.NewTicker(d).C
}()
ticker := time.NewTicker(d)
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return
case <-ticks:
case <-ticker.C:
cclog.Infof("[METRICSTORE]> start checkpointing (starting at %s)...", lastCheckpoint.Format(time.RFC3339))
now := time.Now()
n, err := ms.ToCheckpoint(Keys.Checkpoints.RootDir,
@@ -98,33 +90,23 @@ func Checkpointing(wg *sync.WaitGroup, ctx context.Context) {
} else {
go func() {
defer wg.Done()
d, _ := time.ParseDuration("1m")
select {
case <-ctx.Done():
return
case <-time.After(time.Duration(CheckpointBufferMinutes) * time.Minute):
// This is the first tick untill we collect the data for given minutes.
GetAvroStore().ToCheckpoint(Keys.Checkpoints.RootDir, false)
// log.Printf("Checkpointing %d avro files", count)
}
ticks := func() <-chan time.Time {
if d <= 0 {
return nil
}
return time.NewTicker(d).C
}()
ticker := time.NewTicker(DefaultAvroCheckpointInterval)
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return
case <-ticks:
// Regular ticks of 1 minute to write data.
case <-ticker.C:
GetAvroStore().ToCheckpoint(Keys.Checkpoints.RootDir, false)
// log.Printf("Checkpointing %d avro files", count)
}
}
}()
@@ -329,7 +311,7 @@ func (m *MemoryStore) FromCheckpoint(dir string, from int64, extension string) (
lvl := m.root.findLevelOrCreate(host[:], len(m.Metrics))
nn, err := lvl.fromCheckpoint(m, filepath.Join(dir, host[0], host[1]), from, extension)
if err != nil {
cclog.Fatalf("[METRICSTORE]> error while loading checkpoints: %s", err.Error())
cclog.Errorf("[METRICSTORE]> error while loading checkpoints for %s/%s: %s", host[0], host[1], err.Error())
atomic.AddInt32(&errs, 1)
}
atomic.AddInt32(&n, int32(nn))
@@ -506,8 +488,8 @@ func (l *Level) loadAvroFile(m *MemoryStore, f *os.File, from int64) error {
for key, floatArray := range metricsData {
metricName := ReplaceKey(key)
if strings.Contains(metricName, Delimiter) {
subString := strings.Split(metricName, Delimiter)
if strings.Contains(metricName, SelectorDelimiter) {
subString := strings.Split(metricName, SelectorDelimiter)
lvl := l
@@ -557,12 +539,10 @@ func (l *Level) createBuffer(m *MemoryStore, metricName string, floatArray schem
next: nil,
archived: true,
}
b.close()
minfo, ok := m.Metrics[metricName]
if !ok {
return nil
// return errors.New("Unkown metric: " + name)
}
prev := l.metrics[minfo.offset]
@@ -616,17 +596,15 @@ func (l *Level) loadFile(cf *CheckpointFile, m *MemoryStore) error {
b := &buffer{
frequency: metric.Frequency,
start: metric.Start,
data: metric.Data[0:n:n], // Space is wasted here :(
data: metric.Data[0:n:n],
prev: nil,
next: nil,
archived: true,
}
b.close()
minfo, ok := m.Metrics[name]
if !ok {
continue
// return errors.New("Unkown metric: " + name)
}
prev := l.metrics[minfo.offset]
+10 -14
View File
@@ -7,6 +7,16 @@ package memorystore
import (
"fmt"
"time"
)
const (
DefaultMaxWorkers = 10
DefaultBufferCapacity = 512
DefaultGCTriggerInterval = 100
DefaultAvroWorkers = 4
DefaultCheckpointBufferMin = 3
DefaultAvroCheckpointInterval = time.Minute
)
var InternalCCMSFlag bool = false
@@ -31,20 +41,6 @@ type MetricStoreConfig struct {
RootDir string `json:"directory"`
DeleteInstead bool `json:"delete-instead"`
} `json:"archive"`
Nats []*NatsConfig `json:"nats"`
}
type NatsConfig struct {
// Address of the nats server
Address string `json:"address"`
// Username/Password, optional
Username string `json:"username"`
Password string `json:"password"`
// Creds file path
Credsfilepath string `json:"creds-file-path"`
Subscriptions []struct {
// Channel name
SubscribeTo string `json:"subscribe-to"`
+1 -1
View File
@@ -46,7 +46,7 @@ func (l *Level) findLevelOrCreate(selector []string, nMetrics int) *Level {
}
}
// The level does not exist, take write lock for unqiue access:
// The level does not exist, take write lock for unique access:
l.lock.Lock()
// While this thread waited for the write lock, another thread
// could have created the child node.
+14 -107
View File
@@ -11,113 +11,36 @@ import (
"sync"
"time"
"github.com/ClusterCockpit/cc-backend/pkg/nats"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/ClusterCockpit/cc-lib/schema"
"github.com/influxdata/line-protocol/v2/lineprotocol"
"github.com/nats-io/nats.go"
)
// Each connection is handled in it's own goroutine. This is a blocking function.
// func ReceiveRaw(ctx context.Context,
// listener net.Listener,
// handleLine func(*lineprotocol.Decoder, string) error,
// ) error {
// var wg sync.WaitGroup
// wg.Add(1)
// go func() {
// defer wg.Done()
// <-ctx.Done()
// if err := listener.Close(); err != nil {
// log.Printf("listener.Close(): %s", err.Error())
// }
// }()
// for {
// conn, err := listener.Accept()
// if err != nil {
// if errors.Is(err, net.ErrClosed) {
// break
// }
// log.Printf("listener.Accept(): %s", err.Error())
// }
// wg.Add(2)
// go func() {
// defer wg.Done()
// defer conn.Close()
// dec := lineprotocol.NewDecoder(conn)
// connctx, cancel := context.WithCancel(context.Background())
// defer cancel()
// go func() {
// defer wg.Done()
// select {
// case <-connctx.Done():
// conn.Close()
// case <-ctx.Done():
// conn.Close()
// }
// }()
// if err := handleLine(dec, "default"); err != nil {
// if errors.Is(err, net.ErrClosed) {
// return
// }
// log.Printf("%s: %s", conn.RemoteAddr().String(), err.Error())
// errmsg := make([]byte, 128)
// errmsg = append(errmsg, `error: `...)
// errmsg = append(errmsg, err.Error()...)
// errmsg = append(errmsg, '\n')
// conn.Write(errmsg)
// }
// }()
// }
// wg.Wait()
// return nil
// }
// ReceiveNats connects to a nats server and subscribes to "updates". This is a
// blocking function. handleLine will be called for each line recieved via
// nats. Send `true` through the done channel for gracefull termination.
func ReceiveNats(conf *(NatsConfig),
ms *MemoryStore,
func ReceiveNats(ms *MemoryStore,
workers int,
ctx context.Context,
) error {
var opts []nats.Option
if conf.Username != "" && conf.Password != "" {
opts = append(opts, nats.UserInfo(conf.Username, conf.Password))
}
nc := nats.GetClient()
if conf.Credsfilepath != "" {
opts = append(opts, nats.UserCredentials(conf.Credsfilepath))
if nc == nil {
cclog.Warn("NATS client not initialized")
return nil
}
nc, err := nats.Connect(conf.Address, opts...)
if err != nil {
return err
}
defer nc.Close()
var wg sync.WaitGroup
var subs []*nats.Subscription
msgs := make(chan *nats.Msg, workers*2)
msgs := make(chan []byte, workers*2)
for _, sc := range conf.Subscriptions {
for _, sc := range Keys.Subscriptions {
clusterTag := sc.ClusterTag
var sub *nats.Subscription
if workers > 1 {
wg.Add(workers)
for range workers {
go func() {
for m := range msgs {
dec := lineprotocol.NewDecoderWithBytes(m.Data)
dec := lineprotocol.NewDecoderWithBytes(m)
if err := DecodeLine(dec, ms, clusterTag); err != nil {
cclog.Errorf("error: %s", err.Error())
}
@@ -127,37 +50,23 @@ func ReceiveNats(conf *(NatsConfig),
}()
}
sub, err = nc.Subscribe(sc.SubscribeTo, func(m *nats.Msg) {
msgs <- m
nc.Subscribe(sc.SubscribeTo, func(subject string, data []byte) {
msgs <- data
})
} else {
sub, err = nc.Subscribe(sc.SubscribeTo, func(m *nats.Msg) {
dec := lineprotocol.NewDecoderWithBytes(m.Data)
nc.Subscribe(sc.SubscribeTo, func(subject string, data []byte) {
dec := lineprotocol.NewDecoderWithBytes(data)
if err := DecodeLine(dec, ms, clusterTag); err != nil {
cclog.Errorf("error: %s", err.Error())
}
})
}
if err != nil {
return err
}
cclog.Infof("NATS subscription to '%s' on '%s' established", sc.SubscribeTo, conf.Address)
subs = append(subs, sub)
cclog.Infof("NATS subscription to '%s' established", sc.SubscribeTo)
}
<-ctx.Done()
for _, sub := range subs {
err = sub.Unsubscribe()
if err != nil {
cclog.Errorf("NATS unsubscribe failed: %s", err.Error())
}
}
close(msgs)
wg.Wait()
nc.Close()
cclog.Print("NATS connection closed")
return nil
}
@@ -266,8 +175,6 @@ func DecodeLine(dec *lineprotocol.Decoder,
case "stype-id":
subTypeBuf = append(subTypeBuf, val...)
default:
// Ignore unkown tags (cc-metric-collector might send us a unit for example that we do not need)
// return fmt.Errorf("unkown tag: '%s' (value: '%s')", string(key), string(val))
}
}
+10 -24
View File
@@ -44,8 +44,6 @@ var (
shutdownFunc context.CancelFunc
)
type Metric struct {
Name string
Value schema.Float
@@ -71,8 +69,7 @@ func Init(rawConfig json.RawMessage, wg *sync.WaitGroup) {
// Set NumWorkers from config or use default
if Keys.NumWorkers <= 0 {
maxWorkers := 10
Keys.NumWorkers = min(runtime.NumCPU()/2+1, maxWorkers)
Keys.NumWorkers = min(runtime.NumCPU()/2+1, DefaultMaxWorkers)
}
cclog.Debugf("[METRICSTORE]> Using %d workers for checkpoint/archive operations\n", Keys.NumWorkers)
@@ -144,21 +141,10 @@ func Init(rawConfig json.RawMessage, wg *sync.WaitGroup) {
// Store the shutdown function for later use by Shutdown()
shutdownFunc = shutdown
if Keys.Nats != nil {
for _, natsConf := range Keys.Nats {
// TODO: When multiple nats configs share a URL, do a single connect.
wg.Add(1)
nc := natsConf
go func() {
// err := ReceiveNats(conf.Nats, decodeLine, runtime.NumCPU()-1, ctx)
err := ReceiveNats(nc, ms, 1, ctx)
err = ReceiveNats(ms, 1, ctx)
if err != nil {
cclog.Fatal(err)
}
wg.Done()
}()
}
}
}
// InitMetrics creates a new, initialized instance of a MemoryStore.
@@ -244,18 +230,18 @@ func Retention(wg *sync.WaitGroup, ctx context.Context) {
return
}
ticks := func() <-chan time.Time {
d := d / 2
if d <= 0 {
return nil
tickInterval := d / 2
if tickInterval <= 0 {
return
}
return time.NewTicker(d).C
}()
ticker := time.NewTicker(tickInterval)
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return
case <-ticks:
case <-ticker.C:
t := time.Now().Add(-d)
cclog.Infof("[METRICSTORE]> start freeing buffers (older than %s)...\n", t.Format(time.RFC3339))
freed, err := ms.Free(nil, t.Unix())
@@ -332,7 +318,7 @@ func (m *MemoryStore) Read(selector util.Selector, metric string, from, to, reso
minfo, ok := m.Metrics[metric]
if !ok {
return nil, 0, 0, 0, errors.New("[METRICSTORE]> unkown metric: " + metric)
return nil, 0, 0, 0, errors.New("[METRICSTORE]> unknown metric: " + metric)
}
n, data := 0, make([]schema.Float, (to-from)/minfo.Frequency+1)
+1 -1
View File
@@ -77,7 +77,7 @@ func (m *MemoryStore) Stats(selector util.Selector, metric string, from, to int6
minfo, ok := m.Metrics[metric]
if !ok {
return nil, 0, 0, errors.New("unkown metric: " + metric)
return nil, 0, 0, errors.New("unknown metric: " + metric)
}
n, samples := 0, 0
+9 -8
View File
@@ -376,7 +376,7 @@ func (r *JobRepository) DeleteJobsBefore(startTime int64, omitTagged bool) (int,
return cnt, err
}
func (r *JobRepository) DeleteJobById(id int64) error {
func (r *JobRepository) DeleteJobByID(id int64) error {
// Invalidate cache entries before deletion
r.cache.Del(fmt.Sprintf("metadata:%d", id))
r.cache.Del(fmt.Sprintf("energyFootprint:%d", id))
@@ -577,10 +577,10 @@ func (r *JobRepository) StopJobsExceedingWalltimeBy(seconds int) error {
return nil
}
func (r *JobRepository) FindJobIdsByTag(tagId int64) ([]int64, error) {
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()
Where(sq.Eq{"jobtag.tag_id": tagID}).Distinct()
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
cclog.Error("Error while running query")
@@ -589,15 +589,15 @@ func (r *JobRepository) FindJobIdsByTag(tagId int64) ([]int64, error) {
jobIds := make([]int64, 0, 100)
for rows.Next() {
var jobId int64
var jobID int64
if err := rows.Scan(&jobId); err != nil {
if err := rows.Scan(&jobID); err != nil {
rows.Close()
cclog.Warn("Error while scanning rows")
return nil, err
}
jobIds = append(jobIds, jobId)
jobIds = append(jobIds, jobID)
}
return jobIds, nil
@@ -731,10 +731,11 @@ func (r *JobRepository) UpdateEnergy(
metricEnergy := 0.0
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)
switch sc.MetricConfig[i].Energy {
case "energy": // this metric has energy as unit (Joules or Wh)
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)
case "power": // this metric has power as unit (Watt)
// Energy: Power (in Watts) * Time (in Seconds)
// Unit: (W * (s / 3600)) / 1000 = kWh
// Round 2 Digits: round(Energy * 100) / 100
+7 -6
View File
@@ -2,6 +2,7 @@
// 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 (
@@ -109,27 +110,27 @@ func (r *JobRepository) Start(job *schema.Job) (id int64, err error) {
// Stop updates the job with the database id jobId using the provided arguments.
func (r *JobRepository) Stop(
jobId int64,
jobID int64,
duration int32,
state schema.JobState,
monitoringStatus int32,
) (err error) {
// Invalidate cache entries as job state is changing
r.cache.Del(fmt.Sprintf("metadata:%d", jobId))
r.cache.Del(fmt.Sprintf("energyFootprint:%d", jobId))
r.cache.Del(fmt.Sprintf("metadata:%d", jobID))
r.cache.Del(fmt.Sprintf("energyFootprint:%d", jobID))
stmt := sq.Update("job").
Set("job_state", state).
Set("duration", duration).
Set("monitoring_status", monitoringStatus).
Where("job.id = ?", jobId)
Where("job.id = ?", jobID)
_, err = stmt.RunWith(r.stmtCache).Exec()
return err
}
func (r *JobRepository) StopCached(
jobId int64,
jobID int64,
duration int32,
state schema.JobState,
monitoringStatus int32,
@@ -140,7 +141,7 @@ func (r *JobRepository) StopCached(
Set("job_state", state).
Set("duration", duration).
Set("monitoring_status", monitoringStatus).
Where("job_cache.id = ?", jobId)
Where("job_cache.id = ?", jobID)
_, err = stmt.RunWith(r.stmtCache).Exec()
return err
+34 -33
View File
@@ -2,6 +2,7 @@
// 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 (
@@ -22,13 +23,13 @@ import (
// It returns a pointer to a schema.Job data structure and an error variable.
// To check if no job was found test err == sql.ErrNoRows
func (r *JobRepository) Find(
jobId *int64,
jobID *int64,
cluster *string,
startTime *int64,
) (*schema.Job, error) {
start := time.Now()
q := sq.Select(jobColumns...).From("job").
Where("job.job_id = ?", *jobId)
Where("job.job_id = ?", *jobID)
if cluster != nil {
q = q.Where("job.cluster = ?", *cluster)
@@ -44,12 +45,12 @@ func (r *JobRepository) Find(
}
func (r *JobRepository) FindCached(
jobId *int64,
jobID *int64,
cluster *string,
startTime *int64,
) (*schema.Job, error) {
q := sq.Select(jobCacheColumns...).From("job_cache").
Where("job_cache.job_id = ?", *jobId)
Where("job_cache.job_id = ?", *jobID)
if cluster != nil {
q = q.Where("job_cache.cluster = ?", *cluster)
@@ -63,19 +64,19 @@ func (r *JobRepository) FindCached(
return scanJob(q.RunWith(r.stmtCache).QueryRow())
}
// Find executes a SQL query to find a specific batch job.
// The job is queried using the batch job id, the cluster name,
// and the start time of the job in UNIX epoch time seconds.
// It returns a pointer to a schema.Job data structure and an error variable.
// To check if no job was found test err == sql.ErrNoRows
// FindAll executes a SQL query to find all batch jobs matching the given criteria.
// Jobs are queried using the batch job id, and optionally filtered by cluster name
// and start time (UNIX epoch time seconds).
// It returns a slice of pointers to schema.Job data structures and an error variable.
// An empty slice is returned if no matching jobs are found.
func (r *JobRepository) FindAll(
jobId *int64,
jobID *int64,
cluster *string,
startTime *int64,
) ([]*schema.Job, error) {
start := time.Now()
q := sq.Select(jobColumns...).From("job").
Where("job.job_id = ?", *jobId)
Where("job.job_id = ?", *jobID)
if cluster != nil {
q = q.Where("job.cluster = ?", *cluster)
@@ -139,13 +140,13 @@ func (r *JobRepository) GetJobList(limit int, offset int) ([]int64, error) {
return jl, nil
}
// FindById executes a SQL query to find a specific batch job.
// 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.
// To check if no job was found test err == sql.ErrNoRows
func (r *JobRepository) FindById(ctx context.Context, jobId int64) (*schema.Job, error) {
func (r *JobRepository) FindByID(ctx context.Context, jobID int64) (*schema.Job, error) {
q := sq.Select(jobColumns...).
From("job").Where("job.id = ?", jobId)
From("job").Where("job.id = ?", jobID)
q, qerr := SecurityCheck(ctx, q)
if qerr != nil {
@@ -155,14 +156,14 @@ func (r *JobRepository) FindById(ctx context.Context, jobId int64) (*schema.Job,
return scanJob(q.RunWith(r.stmtCache).QueryRow())
}
// FindByIdWithUser executes a SQL query to find a specific batch job.
// FindByIDWithUser executes a SQL query to find a specific batch job.
// The job is queried using the database id. The user is passed directly,
// instead as part of the context.
// It returns a pointer to a schema.Job data structure and an error variable.
// To check if no job was found test err == sql.ErrNoRows
func (r *JobRepository) FindByIdWithUser(user *schema.User, jobId int64) (*schema.Job, error) {
func (r *JobRepository) FindByIDWithUser(user *schema.User, jobID int64) (*schema.Job, error) {
q := sq.Select(jobColumns...).
From("job").Where("job.id = ?", jobId)
From("job").Where("job.id = ?", jobID)
q, qerr := SecurityCheckWithUser(user, q)
if qerr != nil {
@@ -172,24 +173,24 @@ func (r *JobRepository) FindByIdWithUser(user *schema.User, jobId int64) (*schem
return scanJob(q.RunWith(r.stmtCache).QueryRow())
}
// FindByIdDirect executes a SQL query to find a specific batch job.
// FindByIDDirect 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.
// To check if no job was found test err == sql.ErrNoRows
func (r *JobRepository) FindByIdDirect(jobId int64) (*schema.Job, error) {
func (r *JobRepository) FindByIDDirect(jobID int64) (*schema.Job, error) {
q := sq.Select(jobColumns...).
From("job").Where("job.id = ?", jobId)
From("job").Where("job.id = ?", jobID)
return scanJob(q.RunWith(r.stmtCache).QueryRow())
}
// FindByJobId executes a SQL query to find a specific batch job.
// FindByJobID executes a SQL query to find a specific batch job.
// The job is queried using the slurm id and the clustername.
// It returns a pointer to a schema.Job data structure and an error variable.
// To check if no job was found test err == sql.ErrNoRows
func (r *JobRepository) FindByJobId(ctx context.Context, jobId int64, startTime int64, cluster string) (*schema.Job, error) {
func (r *JobRepository) FindByJobID(ctx context.Context, jobID int64, startTime int64, cluster string) (*schema.Job, error) {
q := sq.Select(jobColumns...).
From("job").
Where("job.job_id = ?", jobId).
Where("job.job_id = ?", jobID).
Where("job.cluster = ?", cluster).
Where("job.start_time = ?", startTime)
@@ -205,10 +206,10 @@ func (r *JobRepository) FindByJobId(ctx context.Context, jobId int64, startTime
// The job is queried using the slurm id,a username and the cluster.
// It returns a bool.
// If job was found, user is owner: test err != sql.ErrNoRows
func (r *JobRepository) IsJobOwner(jobId int64, startTime int64, user string, cluster string) bool {
func (r *JobRepository) IsJobOwner(jobID int64, startTime int64, user string, cluster string) bool {
q := sq.Select("id").
From("job").
Where("job.job_id = ?", jobId).
Where("job.job_id = ?", jobID).
Where("job.hpc_user = ?", user).
Where("job.cluster = ?", cluster).
Where("job.start_time = ?", startTime)
@@ -269,19 +270,19 @@ func (r *JobRepository) FindConcurrentJobs(
queryString := fmt.Sprintf("cluster=%s", job.Cluster)
for rows.Next() {
var id, jobId, startTime sql.NullInt64
var id, jobID, startTime sql.NullInt64
if err = rows.Scan(&id, &jobId, &startTime); err != nil {
if err = rows.Scan(&id, &jobID, &startTime); err != nil {
cclog.Warn("Error while scanning rows")
return nil, err
}
if id.Valid {
queryString += fmt.Sprintf("&jobId=%d", int(jobId.Int64))
queryString += fmt.Sprintf("&jobId=%d", int(jobID.Int64))
items = append(items,
&model.JobLink{
ID: fmt.Sprint(id.Int64),
JobID: int(jobId.Int64),
JobID: int(jobID.Int64),
})
}
}
@@ -294,19 +295,19 @@ func (r *JobRepository) FindConcurrentJobs(
defer rows.Close()
for rows.Next() {
var id, jobId, startTime sql.NullInt64
var id, jobID, startTime sql.NullInt64
if err := rows.Scan(&id, &jobId, &startTime); err != nil {
if err := rows.Scan(&id, &jobID, &startTime); err != nil {
cclog.Warn("Error while scanning rows")
return nil, err
}
if id.Valid {
queryString += fmt.Sprintf("&jobId=%d", int(jobId.Int64))
queryString += fmt.Sprintf("&jobId=%d", int(jobID.Int64))
items = append(items,
&model.JobLink{
ID: fmt.Sprint(id.Int64),
JobID: int(jobId.Int64),
JobID: int(jobID.Int64),
})
}
}
+1 -1
View File
@@ -33,7 +33,7 @@ func TestFind(t *testing.T) {
func TestFindById(t *testing.T) {
r := setup(t)
job, err := r.FindById(getContext(t), 338)
job, err := r.FindByID(getContext(t), 338)
if err != nil {
t.Fatal(err)
}
+1 -1
View File
@@ -118,7 +118,7 @@ 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!")
cclog.Info("Legacy database without version or missing database file!")
} else {
return err
}
+137 -3
View File
@@ -10,6 +10,8 @@ import (
"database/sql"
"encoding/json"
"fmt"
"slices"
"strings"
"sync"
"time"
@@ -106,7 +108,7 @@ func (r *NodeRepository) GetNode(hostname string, cluster string, withMeta bool)
return node, nil
}
func (r *NodeRepository) GetNodeById(id int64, withMeta bool) (*schema.Node, error) {
func (r *NodeRepository) GetNodeByID(id int64, withMeta bool) (*schema.Node, error) {
node := &schema.Node{}
var timestamp int
if err := sq.Select("node.hostname", "node.cluster", "node.subcluster", "node_state.node_state",
@@ -240,7 +242,6 @@ func (r *NodeRepository) QueryNodes(
page *model.PageRequest,
order *model.OrderByInput, // Currently unused!
) ([]*schema.Node, error) {
query, qerr := AccessCheck(ctx,
sq.Select("hostname", "cluster", "subcluster", "node_state", "health_state", "MAX(time_stamp) as time").
From("node").
@@ -309,7 +310,6 @@ func (r *NodeRepository) CountNodes(
ctx context.Context,
filters []*model.NodeFilter,
) (int, error) {
query, qerr := AccessCheck(ctx,
sq.Select("time_stamp", "count(*) as countRes").
From("node").
@@ -553,6 +553,140 @@ func (r *NodeRepository) CountStatesTimed(ctx context.Context, filters []*model.
return timedStates, nil
}
func (r *NodeRepository) GetNodesForList(
ctx context.Context,
cluster string,
subCluster string,
stateFilter string,
nodeFilter string,
page *model.PageRequest,
) ([]string, map[string]string, int, bool, error) {
// Init Return Vars
nodes := make([]string, 0)
stateMap := make(map[string]string)
countNodes := 0
hasNextPage := false
// Build Filters
queryFilters := make([]*model.NodeFilter, 0)
if cluster != "" {
queryFilters = append(queryFilters, &model.NodeFilter{Cluster: &model.StringInput{Eq: &cluster}})
}
if subCluster != "" {
queryFilters = append(queryFilters, &model.NodeFilter{Subcluster: &model.StringInput{Eq: &subCluster}})
}
if nodeFilter != "" && stateFilter != "notindb" {
queryFilters = append(queryFilters, &model.NodeFilter{Hostname: &model.StringInput{Contains: &nodeFilter}})
}
if stateFilter != "all" && stateFilter != "notindb" {
var queryState schema.SchedulerState = schema.SchedulerState(stateFilter)
queryFilters = append(queryFilters, &model.NodeFilter{SchedulerState: &queryState})
}
// if healthFilter != "all" {
// filters = append(filters, &model.NodeFilter{HealthState: &healthFilter})
// }
// Special Case: Disable Paging for missing nodes filter, save IPP for later
var backupItems int
if stateFilter == "notindb" {
backupItems = page.ItemsPerPage
page.ItemsPerPage = -1
}
// Query Nodes From DB
rawNodes, serr := r.QueryNodes(ctx, queryFilters, page, nil) // Order not Used
if serr != nil {
cclog.Warn("error while loading node database data (Resolver.NodeMetricsList)")
return nil, nil, 0, false, serr
}
// Intermediate Node Result Info
for _, node := range rawNodes {
if node == nil {
continue
}
nodes = append(nodes, node.Hostname)
stateMap[node.Hostname] = string(node.NodeState)
}
// Special Case: Find Nodes not in DB node table but in metricStore only
if stateFilter == "notindb" {
// Reapply Original Paging
page.ItemsPerPage = backupItems
// Get Nodes From Topology
var topoNodes []string
if subCluster != "" {
scNodes := archive.NodeLists[cluster][subCluster]
topoNodes = scNodes.PrintList()
} else {
subClusterNodeLists := archive.NodeLists[cluster]
for _, nodeList := range subClusterNodeLists {
topoNodes = append(topoNodes, nodeList.PrintList()...)
}
}
// Compare to all nodes from cluster/subcluster in DB
var missingNodes []string
for _, scanNode := range topoNodes {
if !slices.Contains(nodes, scanNode) {
missingNodes = append(missingNodes, scanNode)
}
}
// Filter nodes by name
if nodeFilter != "" {
filteredNodesByName := []string{}
for _, missingNode := range missingNodes {
if strings.Contains(missingNode, nodeFilter) {
filteredNodesByName = append(filteredNodesByName, missingNode)
}
}
missingNodes = filteredNodesByName
}
// Sort Missing Nodes Alphanumerically
slices.Sort(missingNodes)
// Total Missing
countNodes = len(missingNodes)
// Apply paging
if countNodes > page.ItemsPerPage {
start := (page.Page - 1) * page.ItemsPerPage
end := start + page.ItemsPerPage
if end > countNodes {
end = countNodes
hasNextPage = false
} else {
hasNextPage = true
}
nodes = missingNodes[start:end]
} else {
nodes = missingNodes
}
} else {
// DB Nodes: Count and Find Next Page
var cerr error
countNodes, cerr = r.CountNodes(ctx, queryFilters)
if cerr != nil {
cclog.Warn("error while counting node database data (Resolver.NodeMetricsList)")
return nil, nil, 0, false, cerr
}
// Example Page 4 @ 10 IpP : Does item 41 exist?
// Minimal Page 41 @ 1 IpP : If len(result) is 1, Page 5 exists.
nextPage := &model.PageRequest{
ItemsPerPage: 1,
Page: ((page.Page * page.ItemsPerPage) + 1),
}
nextNodes, err := r.QueryNodes(ctx, queryFilters, nextPage, nil) // Order not Used
if err != nil {
cclog.Warn("Error while querying next nodes")
return nil, nil, 0, false, err
}
hasNextPage = len(nextNodes) == 1
}
return nodes, stateMap, countNodes, hasNextPage, nil
}
func AccessCheck(ctx context.Context, query sq.SelectBuilder) (sq.SelectBuilder, error) {
user := GetUserFromContext(ctx)
return AccessCheckWithUser(user, query)
+1 -1
View File
@@ -55,7 +55,7 @@ func BenchmarkDB_FindJobById(b *testing.B) {
b.RunParallel(func(pb *testing.PB) {
for pb.Next() {
_, err := db.FindById(getContext(b), jobId)
_, err := db.FindByID(getContext(b), jobId)
noErr(b, err)
}
})
+10 -9
View File
@@ -2,6 +2,7 @@
// 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 (
@@ -18,7 +19,7 @@ import (
// AddTag adds the tag with id `tagId` to the job with the database id `jobId`.
// Requires user authentication for security checks.
func (r *JobRepository) AddTag(user *schema.User, job int64, tag int64) ([]*schema.Tag, error) {
j, err := r.FindByIdWithUser(user, job)
j, err := r.FindByIDWithUser(user, job)
if err != nil {
cclog.Warnf("Error finding job %d for user %s: %v", job, user.Username, err)
return nil, err
@@ -32,7 +33,7 @@ func (r *JobRepository) AddTag(user *schema.User, job int64, tag int64) ([]*sche
// AddTagDirect adds a tag without user security checks.
// Use only for internal/admin operations.
func (r *JobRepository) AddTagDirect(job int64, tag int64) ([]*schema.Tag, error) {
j, err := r.FindByIdDirect(job)
j, err := r.FindByIDDirect(job)
if err != nil {
cclog.Warnf("Error finding job %d: %v", job, err)
return nil, err
@@ -43,10 +44,10 @@ func (r *JobRepository) AddTagDirect(job int64, tag int64) ([]*schema.Tag, error
})
}
// Removes a tag from a job by tag id.
// Used by GraphQL API
// RemoveTag removes the tag with the database id `tag` from the job with the database id `job`.
// Requires user authentication for security checks. Used by GraphQL API.
func (r *JobRepository) RemoveTag(user *schema.User, job, tag int64) ([]*schema.Tag, error) {
j, err := r.FindByIdWithUser(user, job)
j, err := r.FindByIDWithUser(user, job)
if err != nil {
cclog.Warn("Error while finding job by id")
return nil, err
@@ -75,8 +76,8 @@ func (r *JobRepository) RemoveTag(user *schema.User, job, tag int64) ([]*schema.
return tags, archive.UpdateTags(j, archiveTags)
}
// Removes a tag from a job by tag info
// Used by REST API
// RemoveJobTagByRequest removes a tag from the job with the database id `job` by tag type, name, and scope.
// Requires user authentication for security checks. 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)
@@ -86,7 +87,7 @@ func (r *JobRepository) RemoveJobTagByRequest(user *schema.User, job int64, tagT
}
// Get Job
j, err := r.FindByIdWithUser(user, job)
j, err := r.FindByIDWithUser(user, job)
if err != nil {
cclog.Warn("Error while finding job by id")
return nil, err
@@ -124,7 +125,7 @@ func (r *JobRepository) removeTagFromArchiveJobs(jobIds []int64) {
continue
}
job, err := r.FindByIdDirect(j)
job, err := r.FindByIDDirect(j)
if err != nil {
cclog.Warnf("Error while getting job %d", j)
continue
+20 -6
View File
@@ -47,7 +47,9 @@ var routes []Route = []Route{
{"/monitoring/systems/list/{cluster}/{subcluster}", "monitoring/systems.tmpl", "Cluster <ID> <SID> Node List - ClusterCockpit", false, setupClusterListRoute},
{"/monitoring/node/{cluster}/{hostname}", "monitoring/node.tmpl", "Node <ID> - ClusterCockpit", false, setupNodeRoute},
{"/monitoring/analysis/{cluster}", "monitoring/analysis.tmpl", "Analysis - ClusterCockpit", true, setupAnalysisRoute},
{"/monitoring/status/{cluster}", "monitoring/status.tmpl", "Status of <ID> - ClusterCockpit", false, setupClusterStatusRoute},
{"/monitoring/status/{cluster}", "monitoring/status.tmpl", "<ID> Dashboard - ClusterCockpit", false, setupClusterStatusRoute},
{"/monitoring/status/detail/{cluster}", "monitoring/status.tmpl", "Status of <ID> - ClusterCockpit", false, setupClusterDetailRoute},
{"/monitoring/dashboard/{cluster}", "monitoring/dashboard.tmpl", "<ID> Dashboard - ClusterCockpit", false, setupDashboardRoute},
}
func setupHomeRoute(i InfoType, r *http.Request) InfoType {
@@ -117,11 +119,23 @@ func setupClusterStatusRoute(i InfoType, r *http.Request) InfoType {
vars := mux.Vars(r)
i["id"] = vars["cluster"]
i["cluster"] = vars["cluster"]
from, to := r.URL.Query().Get("from"), r.URL.Query().Get("to")
if from != "" || to != "" {
i["from"] = from
i["to"] = to
}
i["displayType"] = "DASHBOARD"
return i
}
func setupClusterDetailRoute(i InfoType, r *http.Request) InfoType {
vars := mux.Vars(r)
i["id"] = vars["cluster"]
i["cluster"] = vars["cluster"]
i["displayType"] = "DETAILS"
return i
}
func setupDashboardRoute(i InfoType, r *http.Request) InfoType {
vars := mux.Vars(r)
i["id"] = vars["cluster"]
i["cluster"] = vars["cluster"]
i["displayType"] = "PUBLIC" // Used in Main Template
return i
}
+1 -1
View File
@@ -43,7 +43,7 @@ func TestRegister(t *testing.T) {
func TestMatch(t *testing.T) {
r := setup(t)
job, err := r.FindByIdDirect(317)
job, err := r.FindByIDDirect(317)
noErr(t, err)
var tagger AppTagger
+1 -1
View File
@@ -98,7 +98,7 @@ func RunTaggers() error {
}
for _, id := range jl {
job, err := r.FindByIdDirect(id)
job, err := r.FindByIDDirect(id)
if err != nil {
cclog.Errorf("Error while getting job %s", err)
return err
+1 -1
View File
@@ -18,7 +18,7 @@ func TestInit(t *testing.T) {
func TestJobStartCallback(t *testing.T) {
Init()
r := setup(t)
job, err := r.FindByIdDirect(525)
job, err := r.FindByIDDirect(525)
noErr(t, err)
jobs := make([]*schema.Job, 0, 1)
+1 -1
View File
@@ -17,7 +17,7 @@ import (
func RegisterCompressionService(compressOlderThan int) {
cclog.Info("Register compression service")
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(0o5, 0, 0))),
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(5, 0, 0))),
gocron.NewTask(
func() {
var jobs []*schema.Job
+2 -2
View File
@@ -16,7 +16,7 @@ import (
func RegisterRetentionDeleteService(age int, includeDB bool, omitTagged bool) {
cclog.Info("Register retention delete service")
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(0o4, 0, 0))),
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(3, 0, 0))),
gocron.NewTask(
func() {
startTime := time.Now().Unix() - int64(age*24*3600)
@@ -43,7 +43,7 @@ func RegisterRetentionDeleteService(age int, includeDB bool, omitTagged bool) {
func RegisterRetentionMoveService(age int, includeDB bool, location string, omitTagged bool) {
cclog.Info("Register retention move service")
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(0o4, 0, 0))),
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(4, 0, 0))),
gocron.NewTask(
func() {
startTime := time.Now().Unix() - int64(age*24*3600)
+1 -1
View File
@@ -16,7 +16,7 @@ import (
func RegisterStopJobsExceedTime() {
cclog.Info("Register undead jobs service")
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(0o3, 0, 0))),
s.NewJob(gocron.DailyJob(1, gocron.NewAtTimes(gocron.NewAtTime(3, 0, 0))),
gocron.NewTask(
func() {
err := jobRepo.StopJobsExceedingWalltimeBy(config.Keys.StopJobsExceedingWalltime)
+84 -30
View File
@@ -18,6 +18,7 @@ import (
"path/filepath"
"strconv"
"strings"
"sync"
"text/tabwriter"
"time"
@@ -490,7 +491,46 @@ func (fsa *FsArchive) LoadClusterCfg(name string) (*schema.Cluster, error) {
func (fsa *FsArchive) Iter(loadMetricData bool) <-chan JobContainer {
ch := make(chan JobContainer)
go func() {
defer close(ch)
numWorkers := 4
jobPaths := make(chan string, numWorkers*2)
var wg sync.WaitGroup
for range numWorkers {
wg.Add(1)
go func() {
defer wg.Done()
for jobPath := range jobPaths {
job, err := loadJobMeta(filepath.Join(jobPath, "meta.json"))
if err != nil && !errors.Is(err, &jsonschema.ValidationError{}) {
cclog.Errorf("in %s: %s", jobPath, err.Error())
continue
}
if loadMetricData {
isCompressed := true
filename := filepath.Join(jobPath, "data.json.gz")
if !util.CheckFileExists(filename) {
filename = filepath.Join(jobPath, "data.json")
isCompressed = false
}
data, err := loadJobData(filename, isCompressed)
if err != nil && !errors.Is(err, &jsonschema.ValidationError{}) {
cclog.Errorf("in %s: %s", jobPath, err.Error())
}
ch <- JobContainer{Meta: job, Data: &data}
} else {
ch <- JobContainer{Meta: job, Data: nil}
}
}
}()
}
clustersDir, err := os.ReadDir(fsa.path)
if err != nil {
cclog.Fatalf("Reading clusters failed @ cluster dirs: %s", err.Error())
@@ -507,7 +547,6 @@ func (fsa *FsArchive) Iter(loadMetricData bool) <-chan JobContainer {
for _, lvl1Dir := range lvl1Dirs {
if !lvl1Dir.IsDir() {
// Could be the cluster.json file
continue
}
@@ -525,35 +564,17 @@ func (fsa *FsArchive) Iter(loadMetricData bool) <-chan JobContainer {
for _, startTimeDir := range startTimeDirs {
if startTimeDir.IsDir() {
job, err := loadJobMeta(filepath.Join(dirpath, startTimeDir.Name(), "meta.json"))
if err != nil && !errors.Is(err, &jsonschema.ValidationError{}) {
cclog.Errorf("in %s: %s", filepath.Join(dirpath, startTimeDir.Name()), err.Error())
jobPaths <- filepath.Join(dirpath, startTimeDir.Name())
}
}
}
}
}
if loadMetricData {
isCompressed := true
filename := filepath.Join(dirpath, startTimeDir.Name(), "data.json.gz")
if !util.CheckFileExists(filename) {
filename = filepath.Join(dirpath, startTimeDir.Name(), "data.json")
isCompressed = false
}
data, err := loadJobData(filename, isCompressed)
if err != nil && !errors.Is(err, &jsonschema.ValidationError{}) {
cclog.Errorf("in %s: %s", filepath.Join(dirpath, startTimeDir.Name()), err.Error())
}
ch <- JobContainer{Meta: job, Data: &data}
} else {
ch <- JobContainer{Meta: job, Data: nil}
}
}
}
}
}
}
close(ch)
close(jobPaths)
wg.Wait()
}()
return ch
}
@@ -603,19 +624,52 @@ func (fsa *FsArchive) ImportJob(
return err
}
var dataBuf bytes.Buffer
if err := EncodeJobData(&dataBuf, jobData); err != nil {
cclog.Error("Error while encoding job metricdata")
return err
}
if dataBuf.Len() > 2000 {
f, err = os.Create(path.Join(dir, "data.json.gz"))
if err != nil {
cclog.Error("Error while creating filepath for data.json.gz")
return err
}
gzipWriter := gzip.NewWriter(f)
if _, err := gzipWriter.Write(dataBuf.Bytes()); err != nil {
cclog.Error("Error while writing compressed job data")
gzipWriter.Close()
f.Close()
return err
}
if err := gzipWriter.Close(); err != nil {
cclog.Warn("Error while closing gzip writer")
f.Close()
return err
}
if err := f.Close(); err != nil {
cclog.Warn("Error while closing data.json.gz file")
return err
}
} else {
f, err = os.Create(path.Join(dir, "data.json"))
if err != nil {
cclog.Error("Error while creating filepath for data.json")
return err
}
if err := EncodeJobData(f, jobData); err != nil {
cclog.Error("Error while encoding job metricdata to data.json file")
if _, err := f.Write(dataBuf.Bytes()); err != nil {
cclog.Error("Error while writing job metricdata to data.json file")
f.Close()
return err
}
if err := f.Close(); err != nil {
cclog.Warn("Error while closing data.json file")
}
return err
}
}
return nil
}
func (fsa *FsArchive) StoreClusterCfg(name string, config *schema.Cluster) error {
+58 -24
View File
@@ -17,6 +17,7 @@ import (
"os"
"strconv"
"strings"
"sync"
"text/tabwriter"
"time"
@@ -467,7 +468,6 @@ func (s3a *S3Archive) StoreJobMeta(job *schema.Job) error {
func (s3a *S3Archive) ImportJob(jobMeta *schema.Job, jobData *schema.JobData) error {
ctx := context.Background()
// Upload meta.json
metaKey := getS3Key(jobMeta, "meta.json")
var metaBuf bytes.Buffer
if err := EncodeJobMeta(&metaBuf, jobMeta); err != nil {
@@ -485,18 +485,37 @@ func (s3a *S3Archive) ImportJob(jobMeta *schema.Job, jobData *schema.JobData) er
return err
}
// Upload data.json
dataKey := getS3Key(jobMeta, "data.json")
var dataBuf bytes.Buffer
if err := EncodeJobData(&dataBuf, jobData); err != nil {
cclog.Error("S3Archive ImportJob() > encoding data error")
return err
}
var dataKey string
var dataBytes []byte
if dataBuf.Len() > 2000 {
dataKey = getS3Key(jobMeta, "data.json.gz")
var compressedBuf bytes.Buffer
gzipWriter := gzip.NewWriter(&compressedBuf)
if _, err := gzipWriter.Write(dataBuf.Bytes()); err != nil {
cclog.Errorf("S3Archive ImportJob() > gzip write error: %v", err)
return err
}
if err := gzipWriter.Close(); err != nil {
cclog.Errorf("S3Archive ImportJob() > gzip close error: %v", err)
return err
}
dataBytes = compressedBuf.Bytes()
} else {
dataKey = getS3Key(jobMeta, "data.json")
dataBytes = dataBuf.Bytes()
}
_, err = s3a.client.PutObject(ctx, &s3.PutObjectInput{
Bucket: aws.String(s3a.bucket),
Key: aws.String(dataKey),
Body: bytes.NewReader(dataBuf.Bytes()),
Body: bytes.NewReader(dataBytes),
})
if err != nil {
cclog.Errorf("S3Archive ImportJob() > PutObject data error: %v", err)
@@ -795,29 +814,18 @@ func (s3a *S3Archive) Iter(loadMetricData bool) <-chan JobContainer {
ctx := context.Background()
defer close(ch)
for _, cluster := range s3a.clusters {
prefix := cluster + "/"
numWorkers := 4
metaKeys := make(chan string, numWorkers*2)
var wg sync.WaitGroup
paginator := s3.NewListObjectsV2Paginator(s3a.client, &s3.ListObjectsV2Input{
Bucket: aws.String(s3a.bucket),
Prefix: aws.String(prefix),
})
for paginator.HasMorePages() {
page, err := paginator.NextPage(ctx)
if err != nil {
cclog.Fatalf("S3Archive Iter() > list error: %s", err.Error())
}
for _, obj := range page.Contents {
if obj.Key == nil || !strings.HasSuffix(*obj.Key, "/meta.json") {
continue
}
// Load job metadata
for range numWorkers {
wg.Add(1)
go func() {
defer wg.Done()
for metaKey := range metaKeys {
result, err := s3a.client.GetObject(ctx, &s3.GetObjectInput{
Bucket: aws.String(s3a.bucket),
Key: obj.Key,
Key: aws.String(metaKey),
})
if err != nil {
cclog.Errorf("S3Archive Iter() > GetObject meta error: %v", err)
@@ -849,8 +857,34 @@ func (s3a *S3Archive) Iter(loadMetricData bool) <-chan JobContainer {
ch <- JobContainer{Meta: job, Data: nil}
}
}
}()
}
for _, cluster := range s3a.clusters {
prefix := cluster + "/"
paginator := s3.NewListObjectsV2Paginator(s3a.client, &s3.ListObjectsV2Input{
Bucket: aws.String(s3a.bucket),
Prefix: aws.String(prefix),
})
for paginator.HasMorePages() {
page, err := paginator.NextPage(ctx)
if err != nil {
cclog.Fatalf("S3Archive Iter() > list error: %s", err.Error())
}
for _, obj := range page.Contents {
if obj.Key == nil || !strings.HasSuffix(*obj.Key, "/meta.json") {
continue
}
metaKeys <- *obj.Key
}
}
}
close(metaKeys)
wg.Wait()
}()
return ch
+100 -24
View File
@@ -16,6 +16,7 @@ import (
"os"
"slices"
"strconv"
"sync"
"text/tabwriter"
"time"
@@ -60,6 +61,7 @@ CREATE TABLE IF NOT EXISTS jobs (
CREATE INDEX IF NOT EXISTS idx_jobs_cluster ON jobs(cluster);
CREATE INDEX IF NOT EXISTS idx_jobs_start_time ON jobs(start_time);
CREATE INDEX IF NOT EXISTS idx_jobs_order ON jobs(cluster, start_time);
CREATE INDEX IF NOT EXISTS idx_jobs_lookup ON jobs(cluster, job_id, start_time);
CREATE TABLE IF NOT EXISTS clusters (
@@ -361,16 +363,37 @@ func (sa *SqliteArchive) ImportJob(jobMeta *schema.Job, jobData *schema.JobData)
return err
}
var dataBytes []byte
var compressed bool
if dataBuf.Len() > 2000 {
var compressedBuf bytes.Buffer
gzipWriter := gzip.NewWriter(&compressedBuf)
if _, err := gzipWriter.Write(dataBuf.Bytes()); err != nil {
cclog.Errorf("SqliteArchive ImportJob() > gzip write error: %v", err)
return err
}
if err := gzipWriter.Close(); err != nil {
cclog.Errorf("SqliteArchive ImportJob() > gzip close error: %v", err)
return err
}
dataBytes = compressedBuf.Bytes()
compressed = true
} else {
dataBytes = dataBuf.Bytes()
compressed = false
}
now := time.Now().Unix()
_, err := sa.db.Exec(`
INSERT INTO jobs (job_id, cluster, start_time, meta_json, data_json, data_compressed, created_at, updated_at)
VALUES (?, ?, ?, ?, ?, 0, ?, ?)
VALUES (?, ?, ?, ?, ?, ?, ?, ?)
ON CONFLICT(job_id, cluster, start_time) DO UPDATE SET
meta_json = excluded.meta_json,
data_json = excluded.data_json,
data_compressed = excluded.data_compressed,
updated_at = excluded.updated_at
`, jobMeta.JobID, jobMeta.Cluster, jobMeta.StartTime, metaBuf.Bytes(), dataBuf.Bytes(), now, now)
`, jobMeta.JobID, jobMeta.Cluster, jobMeta.StartTime, metaBuf.Bytes(), dataBytes, compressed, now, now)
if err != nil {
cclog.Errorf("SqliteArchive ImportJob() > insert error: %v", err)
return err
@@ -526,48 +549,60 @@ func (sa *SqliteArchive) CompressLast(starttime int64) int64 {
return last
}
type sqliteJobRow struct {
metaBlob []byte
dataBlob []byte
compressed bool
}
func (sa *SqliteArchive) Iter(loadMetricData bool) <-chan JobContainer {
ch := make(chan JobContainer)
go func() {
defer close(ch)
rows, err := sa.db.Query("SELECT job_id, cluster, start_time, meta_json, data_json, data_compressed FROM jobs ORDER BY cluster, start_time")
if err != nil {
cclog.Fatalf("SqliteArchive Iter() > query error: %s", err.Error())
}
defer rows.Close()
const chunkSize = 1000
offset := 0
for rows.Next() {
var jobID int64
var cluster string
var startTime int64
var metaBlob []byte
var dataBlob []byte
var compressed bool
if err := rows.Scan(&jobID, &cluster, &startTime, &metaBlob, &dataBlob, &compressed); err != nil {
cclog.Errorf("SqliteArchive Iter() > scan error: %v", err)
continue
var query string
if loadMetricData {
query = "SELECT meta_json, data_json, data_compressed FROM jobs ORDER BY cluster, start_time LIMIT ? OFFSET ?"
} else {
query = "SELECT meta_json FROM jobs ORDER BY cluster, start_time LIMIT ? OFFSET ?"
}
job, err := DecodeJobMeta(bytes.NewReader(metaBlob))
numWorkers := 4
jobRows := make(chan sqliteJobRow, numWorkers*2)
var wg sync.WaitGroup
for range numWorkers {
wg.Add(1)
go func() {
defer wg.Done()
for row := range jobRows {
job, err := DecodeJobMeta(bytes.NewReader(row.metaBlob))
if err != nil {
cclog.Errorf("SqliteArchive Iter() > decode meta error: %v", err)
continue
}
if loadMetricData && dataBlob != nil {
var reader io.Reader = bytes.NewReader(dataBlob)
if compressed {
if loadMetricData && row.dataBlob != nil {
var reader io.Reader = bytes.NewReader(row.dataBlob)
if row.compressed {
gzipReader, err := gzip.NewReader(reader)
if err != nil {
cclog.Errorf("SqliteArchive Iter() > gzip error: %v", err)
ch <- JobContainer{Meta: job, Data: nil}
continue
}
defer gzipReader.Close()
reader = gzipReader
decompressed, err := io.ReadAll(gzipReader)
gzipReader.Close()
if err != nil {
cclog.Errorf("SqliteArchive Iter() > decompress error: %v", err)
ch <- JobContainer{Meta: job, Data: nil}
continue
}
reader = bytes.NewReader(decompressed)
}
key := fmt.Sprintf("%s:%d:%d", job.Cluster, job.JobID, job.StartTime)
@@ -583,6 +618,47 @@ func (sa *SqliteArchive) Iter(loadMetricData bool) <-chan JobContainer {
}
}
}()
}
for {
rows, err := sa.db.Query(query, chunkSize, offset)
if err != nil {
cclog.Fatalf("SqliteArchive Iter() > query error: %s", err.Error())
}
rowCount := 0
for rows.Next() {
var row sqliteJobRow
if loadMetricData {
if err := rows.Scan(&row.metaBlob, &row.dataBlob, &row.compressed); err != nil {
cclog.Errorf("SqliteArchive Iter() > scan error: %v", err)
continue
}
} else {
if err := rows.Scan(&row.metaBlob); err != nil {
cclog.Errorf("SqliteArchive Iter() > scan error: %v", err)
continue
}
row.dataBlob = nil
row.compressed = false
}
jobRows <- row
rowCount++
}
rows.Close()
if rowCount < chunkSize {
break
}
offset += chunkSize
}
close(jobRows)
wg.Wait()
}()
return ch
}
+55
View File
@@ -311,3 +311,58 @@ func TestSqliteConfigParsing(t *testing.T) {
t.Errorf("expected dbPath '/tmp/test.db', got '%s'", cfg.DBPath)
}
}
func TestSqliteIterChunking(t *testing.T) {
tmpfile := t.TempDir() + "/test.db"
defer os.Remove(tmpfile)
var sa SqliteArchive
_, err := sa.Init(json.RawMessage(`{"dbPath":"` + tmpfile + `"}`))
if err != nil {
t.Fatalf("init failed: %v", err)
}
defer sa.db.Close()
const totalJobs = 2500
for i := 1; i <= totalJobs; i++ {
job := &schema.Job{
JobID: int64(i),
Cluster: "test",
StartTime: int64(i * 1000),
NumNodes: 1,
Resources: []*schema.Resource{{Hostname: "node001"}},
}
if err := sa.StoreJobMeta(job); err != nil {
t.Fatalf("store failed: %v", err)
}
}
t.Run("IterWithoutData", func(t *testing.T) {
count := 0
for container := range sa.Iter(false) {
if container.Meta == nil {
t.Error("expected non-nil meta")
}
if container.Data != nil {
t.Error("expected nil data when loadMetricData is false")
}
count++
}
if count != totalJobs {
t.Errorf("expected %d jobs, got %d", totalJobs, count)
}
})
t.Run("IterWithData", func(t *testing.T) {
count := 0
for container := range sa.Iter(true) {
if container.Meta == nil {
t.Error("expected non-nil meta")
}
count++
}
if count != totalJobs {
t.Errorf("expected %d jobs, got %d", totalJobs, count)
}
})
}
+246
View File
@@ -0,0 +1,246 @@
// 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 nats provides a generic NATS messaging client for publish/subscribe communication.
//
// The package wraps the nats.go library with connection management, automatic reconnection
// handling, and subscription tracking. It supports multiple authentication methods including
// username/password and credential files.
//
// # Configuration
//
// Configure the client via JSON in the application config:
//
// {
// "nats": {
// "address": "nats://localhost:4222",
// "username": "user",
// "password": "secret"
// }
// }
//
// Or using a credentials file:
//
// {
// "nats": {
// "address": "nats://localhost:4222",
// "creds-file-path": "/path/to/creds.json"
// }
// }
//
// # Usage
//
// The package provides a singleton client initialized once and retrieved globally:
//
// nats.Init(rawConfig)
// nats.Connect()
//
// client := nats.GetClient()
// client.Subscribe("events", func(subject string, data []byte) {
// fmt.Printf("Received: %s\n", data)
// })
//
// client.Publish("events", []byte("hello"))
//
// # Thread Safety
//
// All Client methods are safe for concurrent use.
package nats
import (
"context"
"fmt"
"sync"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
"github.com/nats-io/nats.go"
)
var (
clientOnce sync.Once
clientInstance *Client
)
// Client wraps a NATS connection with subscription management.
type Client struct {
conn *nats.Conn
subscriptions []*nats.Subscription
mu sync.Mutex
}
// MessageHandler is a callback function for processing received messages.
type MessageHandler func(subject string, data []byte)
// Connect initializes the singleton NATS client using the global Keys config.
func Connect() {
clientOnce.Do(func() {
if Keys.Address == "" {
cclog.Warn("NATS: no address configured, skipping connection")
return
}
client, err := NewClient(nil)
if err != nil {
cclog.Errorf("NATS connection failed: %v", err)
return
}
clientInstance = client
})
}
// GetClient returns the singleton NATS client instance.
func GetClient() *Client {
if clientInstance == nil {
cclog.Warn("NATS client not initialized")
}
return clientInstance
}
// NewClient creates a new NATS client. If cfg is nil, uses the global Keys config.
func NewClient(cfg *NatsConfig) (*Client, error) {
if cfg == nil {
cfg = &Keys
}
if cfg.Address == "" {
return nil, fmt.Errorf("NATS address is required")
}
var opts []nats.Option
if cfg.Username != "" && cfg.Password != "" {
opts = append(opts, nats.UserInfo(cfg.Username, cfg.Password))
}
if cfg.CredsFilePath != "" {
opts = append(opts, nats.UserCredentials(cfg.CredsFilePath))
}
opts = append(opts, nats.DisconnectErrHandler(func(_ *nats.Conn, err error) {
if err != nil {
cclog.Warnf("NATS disconnected: %v", err)
}
}))
opts = append(opts, nats.ReconnectHandler(func(nc *nats.Conn) {
cclog.Infof("NATS reconnected to %s", nc.ConnectedUrl())
}))
opts = append(opts, nats.ErrorHandler(func(_ *nats.Conn, _ *nats.Subscription, err error) {
cclog.Errorf("NATS error: %v", err)
}))
nc, err := nats.Connect(cfg.Address, opts...)
if err != nil {
return nil, fmt.Errorf("NATS connect failed: %w", err)
}
cclog.Infof("NATS connected to %s", cfg.Address)
return &Client{
conn: nc,
subscriptions: make([]*nats.Subscription, 0),
}, nil
}
// Subscribe registers a handler for messages on the given subject.
func (c *Client) Subscribe(subject string, handler MessageHandler) error {
c.mu.Lock()
defer c.mu.Unlock()
sub, err := c.conn.Subscribe(subject, func(msg *nats.Msg) {
handler(msg.Subject, msg.Data)
})
if err != nil {
return fmt.Errorf("NATS subscribe to '%s' failed: %w", subject, err)
}
c.subscriptions = append(c.subscriptions, sub)
cclog.Infof("NATS subscribed to '%s'", subject)
return nil
}
// SubscribeQueue registers a handler with queue group for load-balanced message processing.
func (c *Client) SubscribeQueue(subject, queue string, handler MessageHandler) error {
c.mu.Lock()
defer c.mu.Unlock()
sub, err := c.conn.QueueSubscribe(subject, queue, func(msg *nats.Msg) {
handler(msg.Subject, msg.Data)
})
if err != nil {
return fmt.Errorf("NATS queue subscribe to '%s' (queue: %s) failed: %w", subject, queue, err)
}
c.subscriptions = append(c.subscriptions, sub)
cclog.Infof("NATS queue subscribed to '%s' (queue: %s)", subject, queue)
return nil
}
// SubscribeChan subscribes to a subject and delivers messages to the provided channel.
func (c *Client) SubscribeChan(subject string, ch chan *nats.Msg) error {
c.mu.Lock()
defer c.mu.Unlock()
sub, err := c.conn.ChanSubscribe(subject, ch)
if err != nil {
return fmt.Errorf("NATS chan subscribe to '%s' failed: %w", subject, err)
}
c.subscriptions = append(c.subscriptions, sub)
cclog.Infof("NATS chan subscribed to '%s'", subject)
return nil
}
// Publish sends data to the specified subject.
func (c *Client) Publish(subject string, data []byte) error {
if err := c.conn.Publish(subject, data); err != nil {
return fmt.Errorf("NATS publish to '%s' failed: %w", subject, err)
}
return nil
}
// Request sends a request and waits for a response with the given context timeout.
func (c *Client) Request(subject string, data []byte, timeout context.Context) ([]byte, error) {
msg, err := c.conn.RequestWithContext(timeout, subject, data)
if err != nil {
return nil, fmt.Errorf("NATS request to '%s' failed: %w", subject, err)
}
return msg.Data, nil
}
// Flush flushes the connection buffer to ensure all published messages are sent.
func (c *Client) Flush() error {
return c.conn.Flush()
}
// Close unsubscribes all subscriptions and closes the NATS connection.
func (c *Client) Close() {
c.mu.Lock()
defer c.mu.Unlock()
for _, sub := range c.subscriptions {
if err := sub.Unsubscribe(); err != nil {
cclog.Warnf("NATS unsubscribe failed: %v", err)
}
}
c.subscriptions = nil
if c.conn != nil {
c.conn.Close()
cclog.Info("NATS connection closed")
}
}
// IsConnected returns true if the client has an active connection.
func (c *Client) IsConnected() bool {
return c.conn != nil && c.conn.IsConnected()
}
// Connection returns the underlying NATS connection for advanced usage.
func (c *Client) Connection() *nats.Conn {
return c.conn
}
+63
View File
@@ -0,0 +1,63 @@
// 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 nats
import (
"bytes"
"encoding/json"
cclog "github.com/ClusterCockpit/cc-lib/ccLogger"
)
// NatsConfig holds the configuration for connecting to a NATS server.
type NatsConfig struct {
Address string `json:"address"` // NATS server address (e.g., "nats://localhost:4222")
Username string `json:"username"` // Username for authentication (optional)
Password string `json:"password"` // Password for authentication (optional)
CredsFilePath string `json:"creds-file-path"` // Path to credentials file (optional)
}
// Keys holds the global NATS configuration loaded via Init.
var Keys NatsConfig
const ConfigSchema = `{
"type": "object",
"description": "Configuration for NATS messaging client.",
"properties": {
"address": {
"description": "Address of the NATS server (e.g., 'nats://localhost:4222').",
"type": "string"
},
"username": {
"description": "Username for NATS authentication (optional).",
"type": "string"
},
"password": {
"description": "Password for NATS authentication (optional).",
"type": "string"
},
"creds-file-path": {
"description": "Path to NATS credentials file for authentication (optional).",
"type": "string"
}
},
"required": ["address"]
}`
// Init initializes the global Keys configuration from JSON.
func Init(rawConfig json.RawMessage) error {
var err error
if rawConfig != nil {
dec := json.NewDecoder(bytes.NewReader(rawConfig))
dec.DisallowUnknownFields()
if err = dec.Decode(&Keys); err != nil {
cclog.Errorf("Error while initializing nats client: %s", err.Error())
}
}
return err
}
+8 -5
View File
@@ -7,13 +7,16 @@ if [ -d './var' ]; then
./cc-backend -server -dev
else
make
./cc-backend --init
wget https://hpc-mover.rrze.uni-erlangen.de/HPC-Data/0x7b58aefb/eig7ahyo6fo2bais0ephuf2aitohv1ai/job-archive-dev.tar
tar xf job-archive-dev.tar
rm ./job-archive-dev.tar
cp ./configs/env-template.txt .env
cp ./configs/config-demo.json config.json
wget https://hpc-mover.rrze.uni-erlangen.de/HPC-Data/0x7b58aefb/eig7ahyo6fo2bais0ephuf2aitohv1ai/job-archive-demo.tar
tar xf job-archive-demo.tar
rm ./job-archive-demo.tar
./cc-backend -migrate-db
./cc-backend -dev -init-db -add-user demo:admin,api:demo
./cc-backend -server -dev
fi
-9
View File
@@ -1,9 +0,0 @@
//go:build tools
// +build tools
package tools
import (
_ "github.com/99designs/gqlgen"
_ "github.com/swaggo/swag/cmd/swag"
)
+12 -12
View File
@@ -48,7 +48,7 @@ func TestImportFileToSqlite(t *testing.T) {
}
// Perform import
imported, failed, err := importArchive(srcBackend, dstBackend)
imported, failed, err := importArchive(srcBackend, dstBackend, srcConfig)
if err != nil {
t.Errorf("Import failed: %s", err.Error())
}
@@ -111,13 +111,13 @@ func TestImportFileToFile(t *testing.T) {
}
// Create destination archive directory
if err := os.MkdirAll(dstArchive, 0755); err != nil {
if err := os.MkdirAll(dstArchive, 0o755); err != nil {
t.Fatalf("Failed to create destination directory: %s", err.Error())
}
// Write version file
versionFile := filepath.Join(dstArchive, "version.txt")
if err := os.WriteFile(versionFile, []byte("3"), 0644); err != nil {
if err := os.WriteFile(versionFile, []byte("3"), 0o644); err != nil {
t.Fatalf("Failed to write version file: %s", err.Error())
}
@@ -136,7 +136,7 @@ func TestImportFileToFile(t *testing.T) {
}
// Perform import
imported, failed, err := importArchive(srcBackend, dstBackend)
imported, failed, err := importArchive(srcBackend, dstBackend, srcConfig)
if err != nil {
t.Errorf("Import failed: %s", err.Error())
}
@@ -183,7 +183,7 @@ func TestImportDataIntegrity(t *testing.T) {
}
// Perform import
_, _, err = importArchive(srcBackend, dstBackend)
_, _, err = importArchive(srcBackend, dstBackend, srcConfig)
if err != nil {
t.Errorf("Import failed: %s", err.Error())
}
@@ -253,13 +253,13 @@ func TestImportEmptyArchive(t *testing.T) {
dstDb := filepath.Join(tmpdir, "dst-archive.db")
// Create empty source archive
if err := os.MkdirAll(srcArchive, 0755); err != nil {
if err := os.MkdirAll(srcArchive, 0o755); err != nil {
t.Fatalf("Failed to create source directory: %s", err.Error())
}
// Write version file
versionFile := filepath.Join(srcArchive, "version.txt")
if err := os.WriteFile(versionFile, []byte("3"), 0644); err != nil {
if err := os.WriteFile(versionFile, []byte("3"), 0o644); err != nil {
t.Fatalf("Failed to write version file: %s", err.Error())
}
@@ -277,7 +277,7 @@ func TestImportEmptyArchive(t *testing.T) {
}
// Perform import
imported, failed, err := importArchive(srcBackend, dstBackend)
imported, failed, err := importArchive(srcBackend, dstBackend, srcConfig)
if err != nil {
t.Errorf("Import from empty archive should not fail: %s", err.Error())
}
@@ -321,13 +321,13 @@ func TestImportDuplicateJobs(t *testing.T) {
}
// First import
imported1, _, err := importArchive(srcBackend, dstBackend)
imported1, _, err := importArchive(srcBackend, dstBackend, srcConfig)
if err != nil {
t.Fatalf("First import failed: %s", err.Error())
}
// Second import (should skip all jobs)
imported2, _, err := importArchive(srcBackend, dstBackend)
imported2, _, err := importArchive(srcBackend, dstBackend, srcConfig)
if err != nil {
t.Errorf("Second import failed: %s", err.Error())
}
@@ -366,7 +366,7 @@ func TestImportToEmptyFileDestination(t *testing.T) {
util.CopyDir(testDataPath, srcArchive)
// Setup empty destination directory
os.MkdirAll(dstArchive, 0755)
os.MkdirAll(dstArchive, 0o755)
// NOTE: NOT writing version.txt here!
// Initialize source
@@ -384,7 +384,7 @@ func TestImportToEmptyFileDestination(t *testing.T) {
}
// Perform import
imported, _, err := importArchive(srcBackend, dstBackend)
imported, _, err := importArchive(srcBackend, dstBackend, srcConfig)
if err != nil {
t.Errorf("Import failed: %v", err)
}
+265 -35
View File
@@ -5,12 +5,20 @@
package main
import (
"context"
"encoding/json"
"flag"
"fmt"
"io/fs"
"os"
"os/exec"
"os/signal"
"path/filepath"
"strconv"
"strings"
"sync"
"sync/atomic"
"syscall"
"time"
"github.com/ClusterCockpit/cc-backend/internal/config"
@@ -33,80 +41,289 @@ func parseDate(in string) int64 {
return 0
}
// parseArchivePath extracts the path from the source config JSON.
func parseArchivePath(srcConfig string) (string, error) {
var config struct {
Kind string `json:"kind"`
Path string `json:"path"`
}
if err := json.Unmarshal([]byte(srcConfig), &config); err != nil {
return "", fmt.Errorf("failed to parse source config: %w", err)
}
if config.Path == "" {
return "", fmt.Errorf("no path found in source config")
}
return config.Path, nil
}
// countJobsNative counts jobs using native Go filepath.WalkDir.
// This is used as a fallback when fd/fdfind is not available.
func countJobsNative(archivePath string) (int, error) {
count := 0
err := filepath.WalkDir(archivePath, func(path string, d fs.DirEntry, err error) error {
if err != nil {
return nil // Skip directories we can't access
}
if !d.IsDir() && d.Name() == "meta.json" {
count++
}
return nil
})
if err != nil {
return 0, fmt.Errorf("failed to walk directory: %w", err)
}
return count, nil
}
// countJobsWithFd counts jobs using the external fd command.
func countJobsWithFd(fdPath, archivePath string) (int, error) {
fdCmd := exec.Command(fdPath, "meta.json", archivePath)
wcCmd := exec.Command("wc", "-l")
pipe, err := fdCmd.StdoutPipe()
if err != nil {
return 0, fmt.Errorf("failed to create pipe: %w", err)
}
wcCmd.Stdin = pipe
if err := fdCmd.Start(); err != nil {
return 0, fmt.Errorf("failed to start fd command: %w", err)
}
output, err := wcCmd.Output()
if err != nil {
return 0, fmt.Errorf("failed to run wc command: %w", err)
}
if err := fdCmd.Wait(); err != nil {
return 0, fmt.Errorf("fd command failed: %w", err)
}
countStr := strings.TrimSpace(string(output))
count, err := strconv.Atoi(countStr)
if err != nil {
return 0, fmt.Errorf("failed to parse count from wc output '%s': %w", countStr, err)
}
return count, nil
}
// countJobs counts the total number of jobs in the source archive.
// It tries to use external fd/fdfind command for speed, falling back to
// native Go filepath.WalkDir if neither is available.
// The srcConfig parameter should be the JSON configuration string containing the archive path.
func countJobs(srcConfig string) (int, error) {
archivePath, err := parseArchivePath(srcConfig)
if err != nil {
return 0, err
}
// Try fd first (common name)
if fdPath, err := exec.LookPath("fd"); err == nil {
return countJobsWithFd(fdPath, archivePath)
}
// Try fdfind (Debian/Ubuntu package name)
if fdPath, err := exec.LookPath("fdfind"); err == nil {
return countJobsWithFd(fdPath, archivePath)
}
// Fall back to native Go implementation
cclog.Debug("fd/fdfind not found, using native Go file walker")
return countJobsNative(archivePath)
}
// formatDuration formats a duration as a human-readable string.
func formatDuration(d time.Duration) string {
if d < time.Minute {
return fmt.Sprintf("%ds", int(d.Seconds()))
} else if d < time.Hour {
return fmt.Sprintf("%dm%ds", int(d.Minutes()), int(d.Seconds())%60)
}
return fmt.Sprintf("%dh%dm", int(d.Hours()), int(d.Minutes())%60)
}
// progressMeter displays import progress to the terminal.
type progressMeter struct {
total int
processed int32
imported int32
skipped int32
failed int32
startTime time.Time
done chan struct{}
}
func newProgressMeter(total int) *progressMeter {
return &progressMeter{
total: total,
startTime: time.Now(),
done: make(chan struct{}),
}
}
func (p *progressMeter) start() {
go func() {
ticker := time.NewTicker(500 * time.Millisecond)
defer ticker.Stop()
for {
select {
case <-ticker.C:
p.render()
case <-p.done:
p.render()
fmt.Println()
return
}
}
}()
}
func (p *progressMeter) render() {
processed := atomic.LoadInt32(&p.processed)
imported := atomic.LoadInt32(&p.imported)
skipped := atomic.LoadInt32(&p.skipped)
failed := atomic.LoadInt32(&p.failed)
elapsed := time.Since(p.startTime)
percent := float64(processed) / float64(p.total) * 100
if p.total == 0 {
percent = 0
}
var eta string
var throughput float64
if processed > 0 {
throughput = float64(processed) / elapsed.Seconds()
remaining := float64(p.total-int(processed)) / throughput
eta = formatDuration(time.Duration(remaining) * time.Second)
} else {
eta = "calculating..."
}
barWidth := 30
filled := int(float64(barWidth) * float64(processed) / float64(p.total))
if p.total == 0 {
filled = 0
}
var bar strings.Builder
for i := range barWidth {
if i < filled {
bar.WriteString("█")
} else {
bar.WriteString("░")
}
}
fmt.Printf("\r[%s] %5.1f%% | %d/%d | %.1f jobs/s | ETA: %s | ✓%d ○%d ✗%d ",
bar.String(), percent, processed, p.total, throughput, eta, imported, skipped, failed)
}
func (p *progressMeter) stop() {
close(p.done)
}
// importArchive imports all jobs from a source archive backend to a destination archive backend.
// It uses parallel processing with a worker pool to improve performance.
// The import can be interrupted by CTRL-C (SIGINT) and will terminate gracefully.
// Returns the number of successfully imported jobs, failed jobs, and any error encountered.
func importArchive(srcBackend, dstBackend archive.ArchiveBackend) (int, int, error) {
func importArchive(srcBackend, dstBackend archive.ArchiveBackend, srcConfig string) (int, int, error) {
cclog.Info("Starting parallel archive import...")
cclog.Info("Press CTRL-C to interrupt (will finish current jobs before exiting)")
// Use atomic counters for thread-safe updates
var imported int32
var failed int32
var skipped int32
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
sigChan := make(chan os.Signal, 1)
signal.Notify(sigChan, os.Interrupt, syscall.SIGTERM)
var interrupted atomic.Bool
go func() {
<-sigChan
cclog.Warn("Interrupt received, stopping import (finishing current jobs)...")
interrupted.Store(true)
cancel()
// Stop listening for further signals to allow force quit with second CTRL-C
signal.Stop(sigChan)
}()
cclog.Info("Counting jobs in source archive (this may take a long time) ...")
totalJobs, err := countJobs(srcConfig)
if err != nil {
return 0, 0, fmt.Errorf("failed to count jobs: %w", err)
}
cclog.Infof("Found %d jobs to process", totalJobs)
progress := newProgressMeter(totalJobs)
// Number of parallel workers
numWorkers := 4
cclog.Infof("Using %d parallel workers", numWorkers)
// Create channels for job distribution
jobs := make(chan archive.JobContainer, numWorkers*2)
// WaitGroup to track worker completion
var wg sync.WaitGroup
// Start worker goroutines
progress.start()
for i := range numWorkers {
wg.Add(1)
go func(workerID int) {
defer wg.Done()
for job := range jobs {
// Validate job metadata
if job.Meta == nil {
cclog.Warn("Skipping job with nil metadata")
atomic.AddInt32(&failed, 1)
atomic.AddInt32(&progress.failed, 1)
atomic.AddInt32(&progress.processed, 1)
continue
}
// Validate job data
if job.Data == nil {
cclog.Warnf("Job %d from cluster %s has no metric data, skipping",
job.Meta.JobID, job.Meta.Cluster)
atomic.AddInt32(&failed, 1)
atomic.AddInt32(&progress.failed, 1)
atomic.AddInt32(&progress.processed, 1)
continue
}
// Check if job already exists in destination
if dstBackend.Exists(job.Meta) {
cclog.Debugf("Job %d (cluster: %s, start: %d) already exists in destination, skipping",
job.Meta.JobID, job.Meta.Cluster, job.Meta.StartTime)
atomic.AddInt32(&skipped, 1)
atomic.AddInt32(&progress.skipped, 1)
atomic.AddInt32(&progress.processed, 1)
continue
}
// Import job to destination
if err := dstBackend.ImportJob(job.Meta, job.Data); err != nil {
cclog.Errorf("Failed to import job %d from cluster %s: %s",
job.Meta.JobID, job.Meta.Cluster, err.Error())
atomic.AddInt32(&failed, 1)
atomic.AddInt32(&progress.failed, 1)
atomic.AddInt32(&progress.processed, 1)
continue
}
// Successfully imported
newCount := atomic.AddInt32(&imported, 1)
if newCount%100 == 0 {
cclog.Infof("Progress: %d jobs imported, %d skipped, %d failed",
newCount, atomic.LoadInt32(&skipped), atomic.LoadInt32(&failed))
}
atomic.AddInt32(&progress.imported, 1)
atomic.AddInt32(&progress.processed, 1)
}
}(i)
}
// Feed jobs to workers
go func() {
// Import cluster configs first
defer close(jobs)
clusters := srcBackend.GetClusters()
for _, clusterName := range clusters {
if ctx.Err() != nil {
return
}
clusterCfg, err := srcBackend.LoadClusterCfg(clusterName)
if err != nil {
cclog.Errorf("Failed to load cluster config for %s: %v", clusterName, err)
@@ -121,20 +338,33 @@ func importArchive(srcBackend, dstBackend archive.ArchiveBackend) (int, int, err
}
for job := range srcBackend.Iter(true) {
jobs <- job
select {
case <-ctx.Done():
// Drain remaining items from iterator to avoid resource leak
// but don't process them
return
case jobs <- job:
}
}
close(jobs)
}()
// Wait for all workers to complete
wg.Wait()
progress.stop()
finalImported := int(atomic.LoadInt32(&imported))
finalFailed := int(atomic.LoadInt32(&failed))
finalSkipped := int(atomic.LoadInt32(&skipped))
finalImported := int(atomic.LoadInt32(&progress.imported))
finalFailed := int(atomic.LoadInt32(&progress.failed))
finalSkipped := int(atomic.LoadInt32(&progress.skipped))
cclog.Infof("Import completed: %d jobs imported, %d skipped, %d failed",
finalImported, finalSkipped, finalFailed)
elapsed := time.Since(progress.startTime)
if interrupted.Load() {
cclog.Warnf("Import interrupted after %s: %d jobs imported, %d skipped, %d failed",
formatDuration(elapsed), finalImported, finalSkipped, finalFailed)
return finalImported, finalFailed, fmt.Errorf("import interrupted by user")
}
cclog.Infof("Import completed in %s: %d jobs imported, %d skipped, %d failed",
formatDuration(elapsed), finalImported, finalSkipped, finalFailed)
if finalFailed > 0 {
return finalImported, finalFailed, fmt.Errorf("%d jobs failed to import", finalFailed)
@@ -150,7 +380,7 @@ func main() {
flag.StringVar(&srcPath, "s", "./var/job-archive", "Specify the source job archive path. Default is ./var/job-archive")
flag.BoolVar(&flagLogDateTime, "logdate", false, "Set this flag to add date and time to log messages")
flag.StringVar(&flagLogLevel, "loglevel", "warn", "Sets the logging level: `[debug,info,warn (default),err,fatal,crit]`")
flag.StringVar(&flagLogLevel, "loglevel", "info", "Sets the logging level: `[debug,info,warn (default),err,fatal,crit]`")
flag.StringVar(&flagConfigFile, "config", "./config.json", "Specify alternative path to `config.json`")
flag.StringVar(&flagRemoveCluster, "remove-cluster", "", "Remove cluster from archive and database")
flag.StringVar(&flagRemoveBefore, "remove-before", "", "Remove all jobs with start time before date (Format: 2006-Jan-04)")
@@ -188,7 +418,7 @@ func main() {
cclog.Info("Destination backend initialized successfully")
// Perform import
imported, failed, err := importArchive(srcBackend, dstBackend)
imported, failed, err := importArchive(srcBackend, dstBackend, flagSrcConfig)
if err != nil {
cclog.Errorf("Import completed with errors: %s", err.Error())
if failed > 0 {
+1
View File
@@ -74,5 +74,6 @@ export default [
entrypoint('node', 'src/node.entrypoint.js'),
entrypoint('analysis', 'src/analysis.entrypoint.js'),
entrypoint('status', 'src/status.entrypoint.js'),
entrypoint('dashpublic', 'src/dashpublic.entrypoint.js'),
entrypoint('config', 'src/config.entrypoint.js')
];
+3 -1
View File
@@ -7,6 +7,7 @@
- `isApi Bool!`: Is currently logged in user api authority
- `username String!`: Empty string if auth. is disabled, otherwise the username as string
- `ncontent String!`: The currently displayed message on the homescreen
- `clusters [String]`: The available clusternames
-->
<script>
@@ -22,6 +23,7 @@
isApi,
username,
ncontent,
clusters
} = $props();
</script>
@@ -30,7 +32,7 @@
<CardHeader>
<CardTitle class="mb-1">Admin Options</CardTitle>
</CardHeader>
<AdminSettings {ncontent}/>
<AdminSettings {ncontent} {clusters}/>
</Card>
{/if}
+579
View File
@@ -0,0 +1,579 @@
<!--
@component Main cluster status view component; renders current system-usage information
Properties:
- `presetCluster String`: The cluster to show status information for
-->
<script>
// import {
// getContext
// } from "svelte"
import {
queryStore,
gql,
getContextClient,
} from "@urql/svelte";
import {
init,
} from "./generic/utils.js";
import {
formatNumber,
} from "./generic/units.js";
import {
Row,
Col,
Card,
CardHeader,
CardBody,
Spinner,
Table,
Progress,
Icon,
Button
} from "@sveltestrap/sveltestrap";
import Roofline from "./generic/plots/Roofline.svelte";
import Pie, { colors } from "./generic/plots/Pie.svelte";
import Stacked from "./generic/plots/Stacked.svelte";
import DoubleMetric from "./generic/plots/DoubleMetricPlot.svelte";
import Refresher from "./generic/helper/Refresher.svelte";
/* Svelte 5 Props */
let {
presetCluster,
} = $props();
/*Const Init */
const { query: initq } = init();
const client = getContextClient();
// const useCbColors = getContext("cc-config")?.plotConfiguration_colorblindMode || false
/* States */
let from = $state(new Date(Date.now() - (5 * 60 * 1000)));
let clusterFrom = $state(new Date(Date.now() - (8 * 60 * 60 * 1000)));
let to = $state(new Date(Date.now()));
let stackedFrom = $state(Math.floor(Date.now() / 1000) - 14400);
let colWidthStates = $state(0);
let colWidthRoof = $state(0);
let colWidthTotals = $state(0);
let colWidthStacked = $state(0);
/* Derived */
// States for Stacked charts
const statesTimed = $derived(queryStore({
client: client,
query: gql`
query ($filter: [NodeFilter!], $type: String!) {
nodeStatesTimed(filter: $filter, type: $type) {
state
counts
times
}
}
`,
variables: {
filter: { cluster: { eq: presetCluster }, timeStart: stackedFrom},
type: "node",
},
requestPolicy: "network-only"
}));
// Note: nodeMetrics are requested on configured $timestep resolution
// Result: The latest 5 minutes (datapoints) for each node independent of job
const statusQuery = $derived(queryStore({
client: client,
query: gql`
query (
$cluster: String!
$metrics: [String!]
$from: Time!
$to: Time!
$clusterFrom: Time!
$jobFilter: [JobFilter!]!
$nodeFilter: [NodeFilter!]!
$paging: PageRequest!
$sorting: OrderByInput!
) {
# Node 5 Minute Averages for Roofline
nodeMetrics(
cluster: $cluster
metrics: $metrics
from: $from
to: $to
) {
host
subCluster
metrics {
name
metric {
series {
statistics {
avg
}
}
}
}
}
# Running Job Metric Average for Rooflines
jobsMetricStats(filter: $jobFilter, metrics: $metrics) {
id
jobId
duration
numNodes
numAccelerators
subCluster
stats {
name
data {
avg
}
}
}
# Get Jobs for Per-Node Counts
jobs(filter: $jobFilter, order: $sorting, page: $paging) {
items {
jobId
resources {
hostname
}
}
count
}
# Only counts shared nodes once
allocatedNodes(cluster: $cluster) {
name
count
}
# Get Current States fir Pie Charts
nodeStates(filter: $nodeFilter) {
state
count
}
# Get States for Node Roofline; $sorting unused in backend: Use placeholder
nodes(filter: $nodeFilter, order: $sorting) {
count
items {
hostname
cluster
subCluster
schedulerState
}
}
# totalNodes includes multiples if shared jobs: Info-Card Data
jobsStatistics(
filter: $jobFilter
page: $paging
sortBy: TOTALJOBS
groupBy: SUBCLUSTER
) {
id
totalJobs
totalUsers
totalCores
totalAccs
}
# ClusterMetrics for doubleMetricPlot
clusterMetrics(
cluster: $cluster
metrics: $metrics
from: $clusterFrom
to: $to
) {
nodeCount
metrics {
name
unit {
prefix
base
}
timestep
data
}
}
}
`,
variables: {
cluster: presetCluster,
metrics: ["flops_any", "mem_bw"], // Metrics For Cluster Plot and Roofline
from: from.toISOString(),
clusterFrom: clusterFrom.toISOString(),
to: to.toISOString(),
jobFilter: [{ state: ["running"] }, { cluster: { eq: presetCluster } }],
nodeFilter: { cluster: { eq: presetCluster }},
paging: { itemsPerPage: -1, page: 1 }, // Get all: -1
sorting: { field: "startTime", type: "col", order: "DESC" }
},
requestPolicy: "network-only"
}));
const clusterInfo = $derived.by(() => {
if ($initq?.data?.clusters) {
let rawInfos = {};
let subClusters = $initq?.data?.clusters?.find((c) => c.name == presetCluster)?.subClusters || [];
for (let subCluster of subClusters) {
// Allocations
if (!rawInfos['allocatedNodes']) rawInfos['allocatedNodes'] = $statusQuery?.data?.allocatedNodes?.find(({ name }) => name == subCluster.name)?.count || 0;
else rawInfos['allocatedNodes'] += $statusQuery?.data?.allocatedNodes?.find(({ name }) => name == subCluster.name)?.count || 0;
if (!rawInfos['allocatedCores']) rawInfos['allocatedCores'] = $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalCores || 0;
else rawInfos['allocatedCores'] += $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalCores || 0;
if (!rawInfos['allocatedAccs']) rawInfos['allocatedAccs'] = $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalAccs || 0;
else rawInfos['allocatedAccs'] += $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalAccs || 0;
// Infos
if (!rawInfos['processorTypes']) rawInfos['processorTypes'] = subCluster?.processorType ? new Set([subCluster.processorType]) : new Set([]);
else rawInfos['processorTypes'].add(subCluster.processorType);
if (!rawInfos['activeUsers']) rawInfos['activeUsers'] = $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalUsers || 0;
else rawInfos['activeUsers'] += $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalUsers || 0;
if (!rawInfos['runningJobs']) rawInfos['runningJobs'] = $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalJobs || 0;
else rawInfos['runningJobs'] += $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalJobs || 0;
if (!rawInfos['totalNodes']) rawInfos['totalNodes'] = subCluster?.numberOfNodes || 0;
else rawInfos['totalNodes'] += subCluster?.numberOfNodes || 0;
if (!rawInfos['totalCores']) rawInfos['totalCores'] = (subCluster?.socketsPerNode * subCluster?.coresPerSocket * subCluster?.numberOfNodes) || 0;
else rawInfos['totalCores'] += (subCluster?.socketsPerNode * subCluster?.coresPerSocket * subCluster?.numberOfNodes) || 0;
if (!rawInfos['totalAccs']) rawInfos['totalAccs'] = (subCluster?.numberOfNodes * subCluster?.topology?.accelerators?.length) || 0;
else rawInfos['totalAccs'] += (subCluster?.numberOfNodes * subCluster?.topology?.accelerators?.length) || 0;
// Units (Set Once)
if (!rawInfos['flopRateUnit']) rawInfos['flopRateUnit'] = subCluster.flopRateSimd.unit.prefix + subCluster.flopRateSimd.unit.base
if (!rawInfos['memBwRateUnit']) rawInfos['memBwRateUnit'] = subCluster.memoryBandwidth.unit.prefix + subCluster.memoryBandwidth.unit.base
// Get Maxima For Roofline Knee Render
if (!rawInfos['roofData']) {
rawInfos['roofData'] = {
flopRateScalar: {value: subCluster.flopRateScalar.value},
flopRateSimd: {value: subCluster.flopRateSimd.value},
memoryBandwidth: {value: subCluster.memoryBandwidth.value}
};
} else {
rawInfos['roofData']['flopRateScalar']['value'] = Math.max(rawInfos['roofData']['flopRateScalar']['value'], subCluster.flopRateScalar.value)
rawInfos['roofData']['flopRateSimd']['value'] = Math.max(rawInfos['roofData']['flopRateSimd']['value'], subCluster.flopRateSimd.value)
rawInfos['roofData']['memoryBandwidth']['value'] = Math.max(rawInfos['roofData']['memoryBandwidth']['value'], subCluster.memoryBandwidth.value)
}
}
// Keymetrics (Data on Cluster-Scope)
let rawFlops = $statusQuery?.data?.nodeMetrics?.reduce((sum, node) =>
sum + (node.metrics.find((m) => m.name == 'flops_any')?.metric?.series[0]?.statistics?.avg || 0),
0, // Initial Value
) || 0;
rawInfos['flopRate'] = Math.floor((rawFlops * 100) / 100)
let rawMemBw = $statusQuery?.data?.nodeMetrics?.reduce((sum, node) =>
sum + (node.metrics.find((m) => m.name == 'mem_bw')?.metric?.series[0]?.statistics?.avg || 0),
0, // Initial Value
) || 0;
rawInfos['memBwRate'] = Math.floor((rawMemBw * 100) / 100)
return rawInfos
} else {
return {};
}
});
const refinedStateData = $derived.by(() => {
return $statusQuery?.data?.nodeStates.
filter((e) => ['allocated', 'reserved', 'idle', 'mixed','down', 'unknown'].includes(e.state)).
sort((a, b) => b.count - a.count)
});
/* Functions */
function transformNodesStatsToData(subclusterData) {
let data = null
const x = [], y = []
if (subclusterData) {
for (let i = 0; i < subclusterData.length; i++) {
const flopsData = subclusterData[i].metrics.find((s) => s.name == "flops_any")
const memBwData = subclusterData[i].metrics.find((s) => s.name == "mem_bw")
const f = flopsData.metric.series[0].statistics.avg
const m = memBwData.metric.series[0].statistics.avg
let intensity = f / m
if (Number.isNaN(intensity) || !Number.isFinite(intensity)) {
intensity = 0.0 // Set to Float Zero: Will not show in Log-Plot (Always below render limit)
}
x.push(intensity)
y.push(f)
}
} else {
// console.warn("transformNodesStatsToData: metrics for 'mem_bw' and/or 'flops_any' missing!")
}
if (x.length > 0 && y.length > 0) {
data = [null, [x, y]] // for dataformat see roofline.svelte
}
return data
}
function transformNodesStatsToInfo(subClusterData) {
let result = [];
if (subClusterData) { // && $nodesState?.data) {
// Use Nodes as Returned from CCMS, *NOT* as saved in DB via SlurmState-API!
for (let j = 0; j < subClusterData.length; j++) {
const nodeName = subClusterData[j]?.host ? subClusterData[j].host : "unknown"
const nodeMatch = $statusQuery?.data?.nodes?.items?.find((n) => n.hostname == nodeName && n.subCluster == subClusterData[j].subCluster);
const schedulerState = nodeMatch?.schedulerState ? nodeMatch.schedulerState : "notindb"
let numJobs = 0
if ($statusQuery?.data) {
const nodeJobs = $statusQuery?.data?.jobs?.items?.filter((job) => job.resources.find((res) => res.hostname == nodeName))
numJobs = nodeJobs?.length ? nodeJobs.length : 0
}
result.push({nodeName: nodeName, schedulerState: schedulerState, numJobs: numJobs})
};
};
return result
}
</script>
<Card style="height: 98vh;">
<CardBody class="align-content-center">
<Row>
<Col>
<Refresher
hideSelector
initially={60}
onRefresh={(interval) => {
from = new Date(Date.now() - 5 * 60 * 1000);
to = new Date(Date.now());
clusterFrom = new Date(Date.now() - (8 * 60 * 60 * 1000))
if (interval) stackedFrom += Math.floor(interval / 1000);
else stackedFrom += 1 // Workaround: TimeSelection not linked, just trigger new data on manual refresh
}}
/>
</Col>
<Col class="d-flex justify-content-end">
<Button outline class="mb-1" size="sm" color="light" href="/">
<Icon name="x"/>
</Button>
</Col>
</Row>
{#if $statusQuery.fetching || $statesTimed.fetching}
<Row class="justify-content-center">
<Col xs="auto">
<Spinner />
</Col>
</Row>
{:else if $statusQuery.error || $statesTimed.error}
<Row cols={{xs:1, md:2}}>
{#if $statusQuery.error}
<Col>
<Card color="danger"><CardBody>Error Requesting Status Data: {$statusQuery.error.message}</CardBody></Card>
</Col>
{/if}
{#if $statesTimed.error}
<Col>
<Card color="danger"><CardBody>Error Requesting Node Scheduler States: {$statesTimed.error.message}</CardBody></Card>
</Col>
{/if}
</Row>
{:else}
<Row cols={{xs:1, md:2}}>
<Col> <!-- General Cluster Info Card -->
<Card class="h-100">
<CardHeader class="text-center">
<h2 class="mb-0">Cluster {presetCluster.charAt(0).toUpperCase() + presetCluster.slice(1)}</h2>
</CardHeader>
<CardBody>
<h4>CPU(s)</h4><p><strong>{[...clusterInfo?.processorTypes].join(', ')}</strong></p>
</CardBody>
</Card>
</Col>
<Col> <!-- Utilization Info Card -->
<Card class="h-100">
<CardBody>
<Table borderless>
<tr class="py-2">
<td style="font-size:x-large;">{clusterInfo?.runningJobs} Running Jobs</td>
<td colspan="2" style="font-size:x-large;">{clusterInfo?.activeUsers} Active Users</td>
</tr>
<hr class="my-1"/>
<tr class="pt-2">
<td style="font-size: large;">
Flop Rate (<span style="cursor: help;" title="Flops[Any] = (Flops[Double] x 2) + Flops[Single]">Any</span>)
</td>
<td colspan="2" style="font-size: large;">
Memory BW Rate
</td>
</tr>
<tr class="pb-2">
<td style="font-size:x-large;">
{clusterInfo?.flopRate}
{clusterInfo?.flopRateUnit}
</td>
<td colspan="2" style="font-size:x-large;">
{clusterInfo?.memBwRate}
{clusterInfo?.memBwRateUnit}
</td>
</tr>
<hr class="my-1"/>
<tr class="py-2">
<th scope="col">Allocated Nodes</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={clusterInfo?.allocatedNodes}
max={clusterInfo?.totalNodes}
/>
</div></td
>
<td
>{clusterInfo?.allocatedNodes} / {clusterInfo?.totalNodes}
Nodes</td
>
</tr>
<tr class="py-2">
<th scope="col">Allocated Cores</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={clusterInfo?.allocatedCores}
max={clusterInfo?.totalCores}
/>
</div></td
>
<td
>{formatNumber(clusterInfo?.allocatedCores)} / {formatNumber(clusterInfo?.totalCores)}
Cores</td
>
</tr>
{#if clusterInfo?.totalAccs !== 0}
<tr class="py-2">
<th scope="col">Allocated Accelerators</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={clusterInfo?.allocatedAccs}
max={clusterInfo?.totalAccs}
/>
</div></td
>
<td
>{clusterInfo?.allocatedAccs} / {clusterInfo?.totalAccs}
Accelerators</td
>
</tr>
{/if}
</Table>
</CardBody>
</Card>
</Col>
<Col> <!-- Total Cluster Metric in Time SUMS-->
<div bind:clientWidth={colWidthTotals}>
<DoubleMetric
width={colWidthTotals}
timestep={$statusQuery?.data?.clusterMetrics[0]?.timestep || 60}
numNodes={$statusQuery?.data?.clusterMetrics?.nodeCount || 0}
metricData={$statusQuery?.data?.clusterMetrics?.metrics || []}
cluster={presetCluster}
fixLinewidth={2}
/>
</div>
</Col>
<Col> <!-- Nodes Roofline -->
<div bind:clientWidth={colWidthRoof}>
{#key $statusQuery?.data?.nodeMetrics}
<Roofline
colorBackground
useColors={false}
useLegend={false}
allowSizeChange
width={colWidthRoof - 10}
height={300}
cluster={presetCluster}
subCluster={clusterInfo?.roofData ? clusterInfo.roofData : null}
roofData={transformNodesStatsToData($statusQuery?.data?.nodeMetrics)}
nodesData={transformNodesStatsToInfo($statusQuery?.data?.nodeMetrics)}
fixTitle="Node Utilization"
yMinimum={1.0}
/>
{/key}
</div>
</Col>
<Col> <!-- Pie Last States -->
<Row>
{#if refinedStateData.length > 0}
<Col class="px-3 mt-2 mt-lg-0">
<div bind:clientWidth={colWidthStates}>
{#key refinedStateData}
<Pie
canvasId="hpcpie-slurm"
size={colWidthStates * 0.66}
sliceLabel="Nodes"
quantities={refinedStateData.map(
(sd) => sd.count,
)}
entities={refinedStateData.map(
(sd) => sd.state,
)}
fixColors={refinedStateData.map(
(sd) => colors['nodeStates'][sd.state],
)}
/>
{/key}
</div>
</Col>
<Col class="px-4 py-2">
{#key refinedStateData}
<Table>
<tr class="mb-2">
<th></th>
<th class="h4">State</th>
<th class="h4">Count</th>
</tr>
{#each refinedStateData as sd, i}
<tr>
<td><Icon name="circle-fill" style="color: {colors['nodeStates'][sd.state]}; font-size: 30px;"/></td>
<td class="h5">{sd.state.charAt(0).toUpperCase() + sd.state.slice(1)}</td>
<td class="h5">{sd.count}</td>
</tr>
{/each}
</Table>
{/key}
</Col>
{:else}
<Col>
<Card body color="warning" class="mx-4 my-2"
>Cannot render state status: No state data returned for <code>Pie Chart</code></Card
>
</Col>
{/if}
</Row>
</Col>
<Col> <!-- Stacked SchedState -->
<div bind:clientWidth={colWidthStacked}>
{#key $statesTimed?.data?.nodeStatesTimed}
<Stacked
data={$statesTimed?.data?.nodeStatesTimed}
width={colWidthStacked * 0.95}
xlabel="Time"
ylabel="Nodes"
yunit = "#Count"
title = "Cluster Status"
stateType = "Node"
/>
{/key}
</div>
</Col>
</Row>
{/if}
</CardBody>
</Card>
+1 -1
View File
@@ -120,7 +120,7 @@
href: "/monitoring/status/",
icon: "clipboard-data",
perCluster: true,
listOptions: false,
listOptions: true,
menu: "Info",
},
];
+16 -57
View File
@@ -3,80 +3,39 @@
Properties:
- `presetCluster String`: The cluster to show status information for
- `displayType String`: The type of status component to render
-->
<script>
import {
getContext
} from "svelte"
import {
init,
} from "./generic/utils.js";
import {
Row,
Col,
Card,
CardBody,
TabContent,
TabPane,
Spinner
CardBody
} from "@sveltestrap/sveltestrap";
import StatusDash from "./status/StatusDash.svelte";
import UsageDash from "./status/UsageDash.svelte";
import StatisticsDash from "./status/StatisticsDash.svelte";
import DashDetails from "./status/DashDetails.svelte";
import DashInternal from "./status/DashInternal.svelte";
/* Svelte 5 Props */
let {
presetCluster
presetCluster,
displayType
} = $props();
/*Const Init */
const { query: initq } = init();
const useCbColors = getContext("cc-config")?.plotConfiguration_colorblindMode || false
</script>
<!-- Loading indicator & Refresh -->
<Row cols={1} class="mb-2">
<Col>
<h3 class="mb-0">Current Status of Cluster "{presetCluster.charAt(0).toUpperCase() + presetCluster.slice(1)}"</h3>
</Col>
</Row>
{#if $initq.fetching}
<Row cols={1} class="text-center mt-3">
<Col>
<Spinner />
</Col>
</Row>
{:else if $initq.error}
<Row cols={1} class="text-center mt-3">
<Col>
<Card body color="danger">{$initq.error.message}</Card>
</Col>
</Row>
{#if displayType === 'DETAILS'}
<DashDetails {presetCluster}/>
{:else if displayType === 'DASHBOARD'}
<DashInternal {presetCluster}/>
{:else}
<Card class="overflow-auto" style="height: auto;">
<TabContent>
<TabPane tabId="status-dash" tab="Status" active>
<Row>
<Col>
<Card color="danger">
<CardBody>
<StatusDash clusters={$initq.data.clusters} {presetCluster} {useCbColors} useAltColors></StatusDash>
Unknown DisplayType for Status View!
</CardBody>
</TabPane>
<TabPane tabId="usage-dash" tab="Usage">
<CardBody>
<UsageDash {presetCluster} {useCbColors}></UsageDash>
</CardBody>
</TabPane>
<TabPane tabId="metric-dash" tab="Statistics">
<CardBody>
<StatisticsDash {presetCluster} {useCbColors}></StatisticsDash>
</CardBody>
</TabPane>
</TabContent>
</Card>
</Col>
</Row>
{/if}
+1
View File
@@ -10,6 +10,7 @@ mount(Config, {
isApi: isApi,
username: username,
ncontent: ncontent,
clusters: hClusters.map((c) => c.name) // Defined in Header Template
},
context: new Map([
['cc-config', clusterCockpitConfig],
+4 -2
View File
@@ -3,6 +3,7 @@
Properties:
- `ncontent String`: The homepage notice content
- `clusters [String]`: The available clusternames
-->
<script>
@@ -17,7 +18,8 @@
/* Svelte 5 Props */
let {
ncontent
ncontent,
clusters
} = $props();
/* Const Init*/
@@ -66,6 +68,6 @@
<Col>
<EditProject reloadUser={() => getUserList()} />
</Col>
<Options config={ccconfig}/>
<Options config={ccconfig} {clusters}/>
<NoticeEdit {ncontent}/>
</Row>
+29 -1
View File
@@ -1,10 +1,18 @@
<!--
@component Admin option select card
Properties:
- `clusters [String]`: The available clusternames
-->
<script>
import { getContext, onMount } from "svelte";
import { Col, Card, CardBody, CardTitle } from "@sveltestrap/sveltestrap";
import { Row, Col, Card, CardBody, CardTitle, Button, Icon } from "@sveltestrap/sveltestrap";
/* Svelte 5 Props */
let {
clusters,
} = $props();
/*Const Init */
const resampleConfig = getContext("resampling");
@@ -44,6 +52,26 @@
</Card>
</Col>
{#if clusters?.length > 0}
<Col>
<Card class="h-100">
<CardBody>
<CardTitle class="mb-3">Public Dashboard Links</CardTitle>
<Row>
{#each clusters as cluster}
<Col>
<Button color="info" class="mb-2 mb-xl-0" href={`/monitoring/dashboard/${cluster}`} target="_blank">
<Icon name="clipboard-pulse" class="mr-2"/>
{cluster.charAt(0).toUpperCase() + cluster.slice(1)} Public Dashboard
</Button>
</Col>
{/each}
</Row>
</CardBody>
</Card>
</Col>
{/if}
{#if resampleConfig}
<Col>
<Card class="h-100">
+13
View File
@@ -0,0 +1,13 @@
import { mount } from 'svelte';
// import {} from './header.entrypoint.js'
import DashPublic from './DashPublic.root.svelte'
mount(DashPublic, {
target: document.getElementById('svelte-app'),
props: {
presetCluster: presetCluster,
},
context: new Map([
['cc-config', clusterCockpitConfig]
])
})
@@ -14,6 +14,7 @@
let {
initially = null,
presetClass = "",
hideSelector = false,
onRefresh
} = $props();
@@ -36,7 +37,8 @@
});
</script>
<InputGroup class={presetClass}>
{#if !hideSelector}
<InputGroup class={presetClass}>
<Input
type="select"
title="Periodic refresh interval"
@@ -56,5 +58,6 @@
>
<Icon name="arrow-clockwise" /> Refresh
</Button>
</InputGroup>
</InputGroup>
{/if}
@@ -0,0 +1,642 @@
<!--
@component Main plot component, based on uPlot; metricdata values by time
Only width/height should change reactively.
Properties:
- `metric String`: The metric name
- `scope String?`: Scope of the displayed data [Default: node]
- `height Number?`: The plot height [Default: 300]
- `timestep Number`: The timestep used for X-axis rendering
- `series [GraphQL.Series]`: The metric data object
- `statisticsSeries [GraphQL.StatisticsSeries]?`: Min/Max/Median representation of metric data [Default: null]
- `cluster String?`: Cluster name of the parent job / data [Default: ""]
- `subCluster String`: Name of the subCluster of the parent job
- `isShared Bool?`: If this job used shared resources; for additional legend display [Default: false]
- `forNode Bool?`: If this plot is used for node data display; will render x-axis as negative time with $now as maximum [Default: false]
- `numhwthreads Number?`: Number of job HWThreads [Default: 0]
- `numaccs Number?`: Number of job Accelerators [Default: 0]
- `zoomState Object?`: The last zoom state to preserve on user zoom [Default: null]
- `thersholdState Object?`: The last threshold state to preserve on user zoom [Default: null]
- `extendedLegendData Object?`: Additional information to be rendered in an extended legend [Default: null]
- `onZoom Func`: Callback function to handle zoom-in event
-->
<script>
import uPlot from "uplot";
import { formatNumber, formatDurationTime } from "../units.js";
import { getContext, onMount, onDestroy } from "svelte";
import { Card } from "@sveltestrap/sveltestrap";
/* Svelte 5 Props */
let {
// metric,
width = 0,
height = 300,
fixLinewidth = null,
timestep,
numNodes,
metricData,
// useStatsSeries = false,
// statisticsSeries = null,
cluster = "",
forNode = true,
// zoomState = null,
// thresholdState = null,
enableFlip = false,
// onZoom
} = $props();
/* Const Init */
const clusterCockpitConfig = getContext("cc-config");
// const resampleConfig = getContext("resampling");
// const subClusterTopology = getContext("getHardwareTopology")(cluster, subCluster);
// const metricConfig = getContext("getMetricConfig")(cluster, subCluster, metric);
const lineColors = clusterCockpitConfig.plotConfiguration_colorScheme;
const lineWidth = fixLinewidth ? fixLinewidth : clusterCockpitConfig.plotConfiguration_lineWidth / window.devicePixelRatio;
// const cbmode = clusterCockpitConfig?.plotConfiguration_colorblindMode || false;
const renderSleepTime = 200;
// const normalLineColor = "#000000";
// const backgroundColors = {
// normal: "rgba(255, 255, 255, 1.0)",
// caution: cbmode ? "rgba(239, 230, 69, 0.3)" : "rgba(255, 128, 0, 0.3)",
// alert: cbmode ? "rgba(225, 86, 44, 0.3)" : "rgba(255, 0, 0, 0.3)",
// };
/* Var Init */
let timeoutId = null;
/* State Init */
let plotWrapper = $state(null);
let uplot = $state(null);
/* Derived */
// const usesMeanStatsSeries = $derived((statisticsSeries?.mean && statisticsSeries.mean.length != 0));
// const resampleTrigger = $derived(resampleConfig?.trigger ? Number(resampleConfig.trigger) : null);
// const resampleResolutions = $derived(resampleConfig?.resolutions ? [...resampleConfig.resolutions] : null);
// const resampleMinimum = $derived(resampleConfig?.resolutions ? Math.min(...resampleConfig.resolutions) : null);
// const thresholds = $derived(findJobAggregationThresholds(
// subClusterTopology,
// metricConfig,
// scope,
// numhwthreads,
// numaccs
// ));
const longestSeries = $derived.by(() => {
// if (useStatsSeries) {
// return usesMeanStatsSeries ? statisticsSeries?.mean?.length : statisticsSeries?.median?.length;
// } else {
return metricData.reduce((n, m) => Math.max(n, m.data.length), 0);
// }
});
const maxX = $derived(longestSeries * timestep);
// const maxY = $derived.by(() => {
// let pendingY = 0;
// // if (useStatsSeries) {
// // pendingY = statisticsSeries.max.reduce(
// // (max, x) => Math.max(max, x),
// // thresholds?.normal,
// // ) || thresholds?.normal
// // } else {
// pendingY = series.reduce(
// (max, series) => Math.max(max, series?.statistics?.max),
// thresholds?.normal,
// ) || thresholds?.normal;
// // }
// if (pendingY >= 10 * thresholds.peak) {
// // Hard y-range render limit if outliers in series data
// return (10 * thresholds.peak);
// } else {
// return pendingY;
// }
// });
// const plotBands = $derived.by(() => {
// if (useStatsSeries) {
// return [
// { series: [2, 3], fill: cbmode ? "rgba(0,0,255,0.1)" : "rgba(0,255,0,0.1)" },
// { series: [3, 1], fill: cbmode ? "rgba(0,255,0,0.1)" : "rgba(255,0,0,0.1)" },
// ];
// };
// return null;
// })
const plotData = $derived.by(() => {
let pendingData = [new Array(longestSeries)];
// X
if (forNode === true) {
// Negative Timestamp Buildup
for (let i = 0; i <= longestSeries; i++) {
pendingData[0][i] = (longestSeries - i) * timestep * -1;
}
} else {
// Positive Timestamp Buildup
for (let j = 0; j < longestSeries; j++) {
pendingData[0][j] = j * timestep;
};
};
// Y
// if (useStatsSeries) {
// pendingData.push(statisticsSeries.min);
// pendingData.push(statisticsSeries.max);
// if (usesMeanStatsSeries) {
// pendingData.push(statisticsSeries.mean);
// } else {
// pendingData.push(statisticsSeries.median);
// }
// } else {
for (let i = 0; i < metricData.length; i++) {
pendingData.push(metricData[i]?.data);
};
// };
return pendingData;
})
const plotSeries = $derived.by(() => {
let pendingSeries = [
// Note: X-Legend Will not be shown as soon as Y-Axis are in extendedMode
{
label: "Runtime",
value: (u, ts, sidx, didx) =>
(didx == null) ? null : formatDurationTime(ts, forNode),
}
];
// Y
// if (useStatsSeries) {
// pendingSeries.push({
// label: "min",
// scale: "y",
// width: lineWidth,
// stroke: cbmode ? "rgb(0,255,0)" : "red",
// });
// pendingSeries.push({
// label: "max",
// scale: "y",
// width: lineWidth,
// stroke: cbmode ? "rgb(0,0,255)" : "green",
// });
// pendingSeries.push({
// label: usesMeanStatsSeries ? "mean" : "median",
// scale: "y",
// width: lineWidth,
// stroke: "black",
// });
// } else {
for (let i = 0; i < metricData.length; i++) {
// Default
// if (!extendedLegendData) {
pendingSeries.push({
label: `${metricData[i]?.name} (${metricData[i]?.unit?.prefix}${metricData[i]?.unit?.base})`,
scale: `y${i+1}`,
width: lineWidth,
stroke: lineColor(i, metricData.length),
});
// }
// Extended Legend For NodeList
// else {
// pendingSeries.push({
// label:
// scope === "node"
// ? series[i].hostname
// : scope === "accelerator"
// ? 'Acc #' + (i + 1) // series[i].id.slice(9, 14) | Too Hardware Specific
// : scope + " #" + (i + 1),
// scale: "y",
// width: lineWidth,
// stroke: lineColor(i, series?.length),
// values: (u, sidx, idx) => {
// // "i" = "sidx - 1" : sidx contains x-axis-data
// if (idx == null)
// return {
// time: '-',
// value: '-',
// user: '-',
// job: '-'
// };
// if (series[i].id in extendedLegendData) {
// return {
// time: formatDurationTime(plotData[0][idx], forNode),
// value: plotData[sidx][idx],
// user: extendedLegendData[series[i].id].user,
// job: extendedLegendData[series[i].id].job,
// };
// } else {
// return {
// time: formatDurationTime(plotData[0][idx], forNode),
// value: plotData[sidx][idx],
// user: '-',
// job: '-',
// };
// }
// }
// });
// }
// };
};
return pendingSeries;
})
/* Effects */
// $effect(() => {
// if (!useStatsSeries && statisticsSeries != null) useStatsSeries = true;
// })
// This updates plot on all size changes if wrapper (== data) exists
$effect(() => {
if (plotWrapper) {
onSizeChange(width, height);
}
});
/* Functions */
function timeIncrs(timestep, maxX, forNode) {
if (forNode === true) {
return [60, 120, 240, 300, 360, 480, 600, 900, 1800, 3600, 7200, 14400, 21600]; // forNode fixed increments
} else {
let incrs = [];
for (let t = timestep; t < maxX; t *= 10)
incrs.push(t, t * 2, t * 3, t * 5);
return incrs;
}
}
// removed arg "subcluster": input metricconfig and topology now directly derived from subcluster
// function findJobAggregationThresholds(
// subClusterTopology,
// metricConfig,
// scope,
// numhwthreads,
// numaccs
// ) {
// if (!subClusterTopology || !metricConfig || !scope) {
// console.warn("Argument missing for findJobAggregationThresholds!");
// return null;
// }
// // handle special *-stat scopes
// if (scope.match(/(.*)-stat$/)) {
// const statParts = scope.split('-');
// scope = statParts[0]
// }
// if (metricConfig?.aggregation == "avg") {
// // Return as Configured
// return {
// normal: metricConfig.normal,
// caution: metricConfig.caution,
// alert: metricConfig.alert,
// peak: metricConfig.peak,
// };
// }
// if (metricConfig?.aggregation == "sum") {
// // Scale Thresholds
// let fraction;
// if (numaccs > 0) fraction = subClusterTopology.accelerators.length / numaccs;
// else if (numhwthreads > 0) fraction = subClusterTopology.core.length / numhwthreads;
// else fraction = 1; // Fallback
// let divisor;
// // Exclusive: Fraction = 1; Shared: Fraction > 1
// if (scope == 'node') divisor = fraction;
// // Cap divisor at number of available sockets or domains
// else if (scope == 'socket') divisor = (fraction < subClusterTopology.socket.length) ? subClusterTopology.socket.length : fraction;
// else if (scope == "memoryDomain") divisor = (fraction < subClusterTopology.memoryDomain.length) ? subClusterTopology.socket.length : fraction;
// // Use Maximum Division for Smallest Scopes
// else if (scope == "core") divisor = subClusterTopology.core.length;
// else if (scope == "hwthread") divisor = subClusterTopology.core.length; // alt. name for core
// else if (scope == "accelerator") divisor = subClusterTopology.accelerators.length;
// else {
// console.log('Unknown scope, return default aggregation thresholds for sum', scope)
// divisor = 1;
// }
// return {
// peak: metricConfig.peak / divisor,
// normal: metricConfig.normal / divisor,
// caution: metricConfig.caution / divisor,
// alert: metricConfig.alert / divisor,
// };
// }
// console.warn(
// "Missing or unkown aggregation mode (sum/avg) for metric:",
// metricConfig,
// );
// return null;
// }
// UPLOT PLUGIN // converts the legend into a simple tooltip
function legendAsTooltipPlugin({
className,
style = { backgroundColor: "rgba(255, 249, 196, 0.92)", color: "black" },
} = {}) {
let legendEl;
const dataSize = metricData.length;
function init(u, opts) {
legendEl = u.root.querySelector(".u-legend");
legendEl.classList.remove("u-inline");
className && legendEl.classList.add(className);
uPlot.assign(legendEl.style, {
minWidth: "100px",
textAlign: "left",
pointerEvents: "none",
display: "none",
position: "absolute",
left: 0,
top: 0,
zIndex: 100,
boxShadow: "2px 2px 10px rgba(0,0,0,0.5)",
...style,
});
// conditional hide series color markers:
if (
// useStatsSeries || // Min/Max/Median Self-Explanatory
dataSize === 1 || // Only one Y-Dataseries
dataSize > 8 // More than 8 Y-Dataseries
) {
const idents = legendEl.querySelectorAll(".u-marker");
for (let i = 0; i < idents.length; i++)
idents[i].style.display = "none";
}
const overEl = u.over;
overEl.style.overflow = "visible";
// move legend into plot bounds
overEl.appendChild(legendEl);
// show/hide tooltip on enter/exit
overEl.addEventListener("mouseenter", () => {
legendEl.style.display = null;
});
overEl.addEventListener("mouseleave", () => {
legendEl.style.display = "none";
});
// let tooltip exit plot
// overEl.style.overflow = "visible";
}
function update(u) {
const { left, top } = u.cursor;
const internalWidth = u?.over?.querySelector(".u-legend")?.offsetWidth ? u.over.querySelector(".u-legend").offsetWidth : 0;
if (enableFlip && (left < (width/2))) {
legendEl.style.transform = "translate(" + (left + 15) + "px, " + (top + 15) + "px)";
} else {
legendEl.style.transform = "translate(" + (left - internalWidth - 15) + "px, " + (top + 15) + "px)";
}
}
if (dataSize <= 12 ) { // || useStatsSeries) {
return {
hooks: {
init: init,
setCursor: update,
},
};
} else {
// Setting legend-opts show/live as object with false here will not work ...
return {};
}
}
// RETURN BG COLOR FROM THRESHOLD
// function backgroundColor() {
// if (
// clusterCockpitConfig.plotConfiguration_colorBackground == false ||
// // !thresholds ||
// !(series && series.every((s) => s.statistics != null))
// )
// return backgroundColors.normal;
// let cond =
// thresholds.alert < thresholds.caution
// ? (a, b) => a <= b
// : (a, b) => a >= b;
// let avg =
// series.reduce((sum, series) => sum + series.statistics.avg, 0) /
// series.length;
// if (Number.isNaN(avg)) return backgroundColors.normal;
// if (cond(avg, thresholds.alert)) return backgroundColors.alert;
// if (cond(avg, thresholds.caution)) return backgroundColors.caution;
// return backgroundColors.normal;
// }
function lineColor(i, n) {
if (n && n >= lineColors.length) return lineColors[i % lineColors.length];
else return lineColors[Math.floor((i / n) * lineColors.length)];
}
function render(ren_width, ren_height) {
// Set Options
const opts = {
width,
height,
title: 'Cluster Utilization',
plugins: [legendAsTooltipPlugin()],
series: plotSeries,
axes: [
{
scale: "x",
space: 35,
incrs: timeIncrs(timestep, maxX, forNode),
label: "Time",
values: (_, vals) => vals.map((v) => formatDurationTime(v, forNode)),
},
{
scale: "y1",
grid: { show: true },
label: `${metricData[0]?.name} (${metricData[0]?.unit?.prefix}${metricData[0]?.unit?.base})`,
values: (u, vals) => vals.map((v) => formatNumber(v)),
},
{
side: 1,
scale: "y2",
grid: { show: false },
label: `${metricData[1]?.name} (${metricData[1]?.unit?.prefix}${metricData[1]?.unit?.base})`,
values: (u, vals) => vals.map((v) => formatNumber(v)),
},
],
// bands: plotBands,
padding: [5, 10, -20, 0],
hooks: {
// init: [
// (u) => {
// /* IF Zoom Enabled */
// if (resampleConfig && !forNode) {
// u.over.addEventListener("dblclick", (e) => {
// // console.log('Dispatch: Zoom Reset')
// onZoom({
// lastZoomState: {
// x: { time: false },
// y: { auto: true }
// }
// });
// });
// };
// },
// ],
draw: [
(u) => {
// Draw plot type label:
let textl = `Cluster ${cluster}`
// let textl = `${scope}${plotSeries.length > 2 ? "s" : ""}${
// useStatsSeries
// ? (usesMeanStatsSeries ? ": min/mean/max" : ": min/median/max")
// : metricConfig != null && scope != metricConfig.scope
// ? ` (${metricConfig.aggregation})`
// : ""
// }`;
let textr = `Sums of ${numNodes} nodes`
//let textr = `${isShared && scope != "core" && scope != "accelerator" ? "[Shared]" : ""}`;
u.ctx.save();
u.ctx.textAlign = "start"; // 'end'
u.ctx.fillStyle = "black";
u.ctx.fillText(textl, u.bbox.left + 10, u.bbox.top + (forNode ? 0 : 10));
u.ctx.textAlign = "end";
u.ctx.fillStyle = "black";
u.ctx.fillText(
textr,
u.bbox.left + u.bbox.width - 10,
u.bbox.top + (forNode ? 0 : 10),
);
// u.ctx.fillText(text, u.bbox.left + u.bbox.width - 10, u.bbox.top + u.bbox.height - 10) // Recipe for bottom right
// if (!thresholds) {
u.ctx.restore();
return;
// }
// let y = u.valToPos(thresholds.normal, "y", true);
// u.ctx.save();
// u.ctx.lineWidth = lineWidth;
// u.ctx.strokeStyle = normalLineColor;
// u.ctx.setLineDash([5, 5]);
// u.ctx.beginPath();
// u.ctx.moveTo(u.bbox.left, y);
// u.ctx.lineTo(u.bbox.left + u.bbox.width, y);
// u.ctx.stroke();
// u.ctx.restore();
},
],
// setScale: [
// (u, key) => { // If ZoomResample is Configured && Not System/Node View
// if (resampleConfig && !forNode && key === 'x') {
// const numX = (u.series[0].idxs[1] - u.series[0].idxs[0])
// if (numX <= resampleTrigger && timestep !== resampleMinimum) {
// /* Get closest zoom level; prevents multiple iterative zoom requests for big zoom-steps (e.g. 600 -> 300 -> 120 -> 60) */
// // Which resolution to theoretically request to achieve 30 or more visible data points:
// const target = (numX * timestep) / resampleTrigger
// // Which configured resolution actually matches the closest to theoretical target:
// const closest = resampleResolutions.reduce(function(prev, curr) {
// return (Math.abs(curr - target) < Math.abs(prev - target) ? curr : prev);
// });
// // Prevents non-required dispatches
// if (timestep !== closest) {
// // console.log('Dispatch: Zoom with Res from / to', timestep, closest)
// onZoom({
// newRes: closest,
// lastZoomState: u?.scales,
// lastThreshold: thresholds?.normal
// });
// }
// } else {
// // console.log('Dispatch: Zoom Update States')
// onZoom({
// lastZoomState: u?.scales,
// lastThreshold: thresholds?.normal
// });
// };
// };
// },
// ]
},
scales: {
x: { time: false },
y1: { auto: true },
y1: { auto: true },
},
legend: {
// Display legend until max 12 Y-dataseries
show: true, // metricData.length <= 12 || useStatsSeries,
live: true // But This Plot always for 2 Data-Series
},
cursor: {
drag: { x: true, y: true },
}
};
// Handle Render
if (!uplot) {
opts.width = ren_width;
opts.height = ren_height;
// if (plotSync) {
// opts.cursor.sync = {
// key: plotSync.key,
// scales: ["x", null],
// }
// }
// if (zoomState && metricConfig?.aggregation == "avg") {
// opts.scales = {...zoomState}
// } else if (zoomState && metricConfig?.aggregation == "sum") {
// // Allow Zoom In === Ymin changed
// if (zoomState.y.min !== 0) { // scope change?: only use zoomState if thresholds match
// if ((thresholdState === thresholds?.normal)) { opts.scales = {...zoomState} };
// } // else: reset scaling to default
// }
uplot = new uPlot(opts, plotData, plotWrapper);
} else {
uplot.setSize({ width: ren_width, height: ren_height });
}
}
function onSizeChange(chg_width, chg_height) {
if (!uplot) return;
if (timeoutId != null) clearTimeout(timeoutId);
timeoutId = setTimeout(() => {
timeoutId = null;
render(chg_width, chg_height);
}, renderSleepTime);
}
/* On Mount */
onMount(() => {
if (plotWrapper) {
render(width, height);
}
});
/* On Destroy */
onDestroy(() => {
if (timeoutId != null) clearTimeout(timeoutId);
if (uplot) uplot.destroy();
});
</script>
<!-- Define $width Wrapper and NoData Card -->
{#if metricData[0]?.data && metricData[0]?.data?.length > 0}
<div bind:this={plotWrapper} bind:clientWidth={width}
class={forNode ? 'py-2 rounded' : 'rounded'}
></div>
{:else}
<Card body color="warning" class="mx-4"
>Cannot render plot: No series data returned for <code>{cluster}</code></Card
>
{/if}
+12 -1
View File
@@ -59,7 +59,15 @@
'rgb(135,133,0)',
'rgb(0,167,108)',
'rgb(189,189,189)',
]
],
nodeStates: {
allocated: "rgba(0, 128, 0, 0.75)",
down: "rgba(255, 0, 0, 0.75)",
idle: "rgba(0, 0, 255, 0.75)",
reserved: "rgba(255, 0, 255, 0.75)",
mixed: "rgba(255, 215, 0, 0.75)",
unknown: "rgba(0, 0, 0, 0.75)"
}
}
</script>
@@ -77,6 +85,7 @@
entities,
displayLegend = false,
useAltColors = false,
fixColors = null
} = $props();
/* Const Init */
@@ -98,6 +107,8 @@
c = [...colors['colorblind']];
} else if (useAltColors) {
c = [...colors['alternative']];
} else if (fixColors?.length > 0) {
c = [...fixColors];
} else {
c = [...colors['default']];
}
+168 -23
View File
@@ -34,14 +34,18 @@
nodesData = null,
cluster = null,
subCluster = null,
fixTitle = null,
yMinimum = null,
allowSizeChange = false,
useColors = true,
useLegend = true,
colorBackground = false,
width = 600,
height = 380,
} = $props();
/* Const Init */
const lineWidth = clusterCockpitConfig.plotConfiguration_lineWidth;
const lineWidth = 2 // clusterCockpitConfig.plotConfiguration_lineWidth;
const cbmode = clusterCockpitConfig?.plotConfiguration_colorblindMode || false;
/* Var Init */
@@ -293,7 +297,7 @@
} else {
// No Colors: Use Black
u.ctx.strokeStyle = "rgb(0, 0, 0)";
u.ctx.fillStyle = "rgba(0, 0, 0, 0.5)";
u.ctx.fillStyle = colorBackground ? "rgb(0, 0, 0)" : "rgba(0, 0, 0, 0.5)";
}
// Get Values
@@ -526,6 +530,7 @@
let plotTitle = "CPU Roofline Diagram";
if (jobsData) plotTitle = "Job Average Roofline Diagram";
if (nodesData) plotTitle = "Node Average Roofline Diagram";
if (fixTitle) plotTitle = fixTitle
if (roofData) {
const opts = {
@@ -534,7 +539,7 @@
width: width,
height: height,
legend: {
show: true,
show: useLegend,
},
cursor: {
dataIdx: (u, seriesIdx) => {
@@ -616,7 +621,7 @@
},
y: {
range: [
0.01,
yMinimum ? yMinimum : 0.01,
subCluster?.flopRateSimd?.value
? nearestThousand(subCluster.flopRateSimd.value)
: 10000,
@@ -646,7 +651,7 @@
hooks: {
// setSeries: [ (u, seriesIdx) => console.log('setSeries', seriesIdx) ],
// setLegend: [ u => console.log('setLegend', u.legend.idxs) ],
drawClear: [
drawClear: [ // drawClear hook which fires before anything exists, so will render under the grid
(u) => {
qt = qt || new Quadtree(0, 0, u.bbox.width, u.bbox.height);
qt.clear();
@@ -658,7 +663,7 @@
});
},
],
draw: [
drawAxes: [ // drawAxes hook, which fires after axes and grid have been rendered
(u) => {
// draw roofs when subCluster set
if (subCluster != null) {
@@ -668,6 +673,7 @@
u.ctx.lineWidth = lineWidth;
u.ctx.beginPath();
// Get Values
const ycut = 0.01 * subCluster.memoryBandwidth.value;
const scalarKnee =
(subCluster.flopRateScalar.value - ycut) /
@@ -675,19 +681,20 @@
const simdKnee =
(subCluster.flopRateSimd.value - ycut) /
subCluster.memoryBandwidth.value;
const scalarKneeX = u.valToPos(scalarKnee, "x", true), // Value, axis, toCanvasPixels
simdKneeX = u.valToPos(simdKnee, "x", true),
flopRateScalarY = u.valToPos(
subCluster.flopRateScalar.value,
"y",
true,
),
flopRateSimdY = u.valToPos(
subCluster.flopRateSimd.value,
"y",
true,
);
// Get Const Coords
const originX = u.valToPos(0.01, "x", true);
const originY = u.valToPos(yMinimum ? yMinimum : 0.01, "y", true);
const outerX = u.valToPos(1000, "x", true); // rightmost x in plot coords
const scalarKneeX = u.valToPos(scalarKnee, "x", true) // Value, axis, toCanvasPixels
const simdKneeX = u.valToPos(simdKnee, "x", true)
const flopRateScalarY = u.valToPos(subCluster.flopRateScalar.value, "y", true)
const flopRateSimdY = u.valToPos(subCluster.flopRateSimd.value, "y", true);
/* Render Lines */
if (
scalarKneeX <
width * window.devicePixelRatio -
@@ -727,10 +734,10 @@
y1,
x2,
y2,
u.valToPos(0.01, "x", true),
u.valToPos(1.0, "y", true), // X-Axis Start Coords
u.valToPos(1000, "x", true),
u.valToPos(1.0, "y", true), // X-Axis End Coords
originX, // x3; X-Axis Start Coord-X
originY, // y3; X-Axis Start Coord-Y
outerX, // x4; X-Axis End Coord-X
originY, // y4; X-Axis End Coord-Y
);
if (xAxisIntersect.x > x1) {
@@ -745,6 +752,144 @@
u.ctx.stroke();
// Reset grid lineWidth
u.ctx.lineWidth = 0.15;
/* Render Area */
if (colorBackground) {
u.ctx.beginPath();
// Additional Coords for Colored Regions
const yhalf = u.valToPos(ycut/2, "y", true)
const simdShift = u.valToPos(simdKnee*1.75, "x", true)
let upperBorderIntersect = lineIntersect(
x1,
y1,
x2,
y2,
originX, // x3; X-Axis Start Coord-X
flopRateSimdY*1.667, // y3; X-Axis Start Coord-Y
outerX, // x4; X-Axis End Coord-X
flopRateSimdY*1.667, // y4; X-Axis End Coord-Y
);
let lowerBorderIntersect = lineIntersect(
x1,
y1,
x2,
y2,
originX, // x3; X-Axis Start Coord-X
flopRateScalarY*1.1667, // y3; X-Axis Start Coord-Y
outerX, // x4; X-Axis End Coord-X
flopRateScalarY*1.1667, // y4; X-Axis End Coord-Y
);
let helperUpperBorderIntersect = lineIntersect(
x1,
yhalf,
simdShift,
y2,
originX, // x3; X-Axis Start Coord-X
flopRateSimdY*1.667, // y3; X-Axis Start Coord-Y
outerX, // x4; X-Axis End Coord-X
flopRateSimdY*1.667, // y4; X-Axis End Coord-Y
);
let helperLowerBorderIntersect = lineIntersect(
x1,
yhalf,
simdShift,
y2,
originX, // x3; X-Axis Start Coord-X
flopRateScalarY*1.1667, // y3; X-Axis Start Coord-Y
outerX, // x4; X-Axis End Coord-X
flopRateScalarY*1.1667, // y4; X-Axis End Coord-Y
);
let helperLowerBorderIntersectTop = lineIntersect(
x1,
yhalf,
simdShift,
y2,
scalarKneeX, // x3; X-Axis Start Coord-X
flopRateScalarY, // y3; X-Axis Start Coord-Y
outerX, // x4; X-Axis End Coord-X
flopRateScalarY, // y4; X-Axis End Coord-Y
);
// Diagonal Helper
u.ctx.moveTo(x1, yhalf);
u.ctx.lineTo(simdShift, y2);
// Upper Simd Helper
u.ctx.moveTo(upperBorderIntersect.x, flopRateSimdY*1.667);
u.ctx.lineTo(outerX, flopRateSimdY*1.667);
// Lower Scalar Helper
u.ctx.moveTo(lowerBorderIntersect.x, flopRateScalarY*1.1667);
u.ctx.lineTo(outerX, flopRateScalarY*1.1667);
u.ctx.stroke();
/* Color Regions */
// MemoryBound
u.ctx.save();
u.ctx.beginPath();
u.ctx.lineTo(x1, y1); // YCut
u.ctx.lineTo(x2, y2); // Upper Knee
u.ctx.lineTo(simdShift, y2); // Upper Helper Knee
u.ctx.lineTo(x1, yhalf); // Half yCut
u.ctx.closePath();
u.ctx.fillStyle = "rgba(255, 200, 0, 0.4)"; // Yellow
u.ctx.fill();
u.ctx.restore();
// Compute Lower
u.ctx.save();
u.ctx.beginPath();
u.ctx.moveTo(lowerBorderIntersect.x, flopRateScalarY*1.1667); // Lower Helper Knee
u.ctx.lineTo(scalarKneeX, flopRateScalarY); // Lower Knee
u.ctx.lineTo(outerX, flopRateScalarY); // Outer Border
u.ctx.lineTo(outerX, flopRateScalarY*1.1667); // Outer Lower Helper Border
u.ctx.closePath();
u.ctx.fillStyle = "rgba(0, 180, 255, 0.4)"; // Cyan Blue
u.ctx.fill();
u.ctx.restore();
// Compute Upper
u.ctx.save();
u.ctx.beginPath();
u.ctx.moveTo(upperBorderIntersect.x, flopRateSimdY*1.667); // Upper Helper Knee
u.ctx.lineTo(simdKneeX, flopRateSimdY); // Upper Knee
u.ctx.lineTo(outerX, flopRateSimdY); // Outer Border
u.ctx.lineTo(outerX, flopRateSimdY*1.667); // Outer Upper Helper Border
u.ctx.closePath();
u.ctx.fillStyle = "rgba(0, 180, 255, 0.4)"; // Cyan Blue
u.ctx.fill();
u.ctx.restore();
// Nomansland Lower
u.ctx.save();
u.ctx.beginPath();
u.ctx.moveTo(originX, originY); // Origin
u.ctx.lineTo(originX, yhalf); // YCut Half
u.ctx.lineTo(helperLowerBorderIntersect.x, flopRateScalarY*1.1667); // Lower Inner Helper Knee
u.ctx.lineTo(outerX, flopRateScalarY*1.1667); // Lower Inner Border
u.ctx.lineTo(outerX, originY); // Lower Right Corner
u.ctx.closePath();
u.ctx.fillStyle = "rgba(255, 50, 50, 0.1)"; // Red Light
u.ctx.fill();
u.ctx.restore();
// Nomansland Upper
u.ctx.save();
u.ctx.beginPath();
u.ctx.moveTo(helperLowerBorderIntersectTop.x, flopRateScalarY); // Lower Knee Top
u.ctx.lineTo(helperUpperBorderIntersect.x, flopRateSimdY*1.667); // Upper Helper Knee
u.ctx.lineTo(outerX, flopRateSimdY*1.667); // Upper Inner Border
u.ctx.lineTo(outerX, flopRateScalarY); // Lower Knee Border
u.ctx.closePath();
u.ctx.fillStyle = "rgba(255, 50, 50, 0.1)"; // Red Light
u.ctx.fill();
u.ctx.restore();
}
}
/* Render Scales */
@@ -852,5 +997,5 @@
{#if roofData != null}
<div bind:this={plotWrapper} class="p-2"></div>
{:else}
<Card class="mx-4" body color="warning">Cannot render roofline: No data!</Card>
<Card class="mx-4 my-2" body color="warning">Cannot render roofline: No data!</Card>
{/if}
@@ -315,10 +315,10 @@
y1,
x2,
y2,
u.valToPos(0.01, "x", true),
u.valToPos(1.0, "y", true), // X-Axis Start Coords
u.valToPos(1000, "x", true),
u.valToPos(1.0, "y", true), // X-Axis End Coords
u.valToPos(0.01, "x", true), // x3; X-Axis Start Coord-X
u.valToPos(1.0, "y", true), // y3; X-Axis Start Coord-Y
u.valToPos(1000, "x", true), // x4; X-Axis End Coord-X
u.valToPos(1.0, "y", true), // y4; X-Axis End Coord-Y
);
if (xAxisIntersect.x > x1) {
+13 -13
View File
@@ -39,72 +39,72 @@
label: "Full",
scale: "y",
width: lineWidth,
fill: cbmode ? "rgba(0, 110, 0, 0.4)" : "rgba(0, 128, 0, 0.4)",
fill: cbmode ? "rgba(0, 110, 0, 0.6)" : "rgba(0, 128, 0, 0.6)",
stroke: cbmode ? "rgb(0, 110, 0)" : "green",
},
partial: {
label: "Partial",
scale: "y",
width: lineWidth,
fill: cbmode ? "rgba(235, 172, 35, 0.4)" : "rgba(255, 215, 0, 0.4)",
fill: cbmode ? "rgba(235, 172, 35, 0.6)" : "rgba(255, 215, 0, 0.6)",
stroke: cbmode ? "rgb(235, 172, 35)" : "gold",
},
failed: {
label: "Failed",
scale: "y",
width: lineWidth,
fill: cbmode ? "rgb(181, 29, 20, 0.4)" : "rgba(255, 0, 0, 0.4)",
fill: cbmode ? "rgb(181, 29, 20, 0.6)" : "rgba(255, 0, 0, 0.6)",
stroke: cbmode ? "rgb(181, 29, 20)" : "red",
},
idle: {
label: "Idle",
scale: "y",
width: lineWidth,
fill: cbmode ? "rgba(0, 140, 249, 0.4)" : "rgba(0, 0, 255, 0.4)",
fill: cbmode ? "rgba(0, 140, 249, 0.6)" : "rgba(0, 0, 255, 0.6)",
stroke: cbmode ? "rgb(0, 140, 249)" : "blue",
},
allocated: {
label: "Allocated",
scale: "y",
width: lineWidth,
fill: cbmode ? "rgba(0, 110, 0, 0.4)" : "rgba(0, 128, 0, 0.4)",
fill: cbmode ? "rgba(0, 110, 0, 0.6)" : "rgba(0, 128, 0, 0.6)",
stroke: cbmode ? "rgb(0, 110, 0)" : "green",
},
reserved: {
label: "Reserved",
scale: "y",
width: lineWidth,
fill: cbmode ? "rgba(209, 99, 230, 0.4)" : "rgba(255, 0, 255, 0.4)",
fill: cbmode ? "rgba(209, 99, 230, 0.6)" : "rgba(255, 0, 255, 0.6)",
stroke: cbmode ? "rgb(209, 99, 230)" : "magenta",
},
mixed: {
label: "Mixed",
scale: "y",
width: lineWidth,
fill: cbmode ? "rgba(235, 172, 35, 0.4)" : "rgba(255, 215, 0, 0.4)",
fill: cbmode ? "rgba(235, 172, 35, 0.6)" : "rgba(255, 215, 0, 0.6)",
stroke: cbmode ? "rgb(235, 172, 35)" : "gold",
},
down: {
label: "Down",
scale: "y",
width: lineWidth,
fill: cbmode ? "rgba(181, 29 ,20, 0.4)" : "rgba(255, 0, 0, 0.4)",
fill: cbmode ? "rgba(181, 29 ,20, 0.6)" : "rgba(255, 0, 0, 0.6)",
stroke: cbmode ? "rgb(181, 29, 20)" : "red",
},
unknown: {
label: "Unknown",
scale: "y",
width: lineWidth,
fill: "rgba(0, 0, 0, 0.4)",
fill: "rgba(0, 0, 0, 0.6)",
stroke: "black",
}
};
// Data Prep For uPlot
const sortedData = data.sort((a, b) => a.state.localeCompare(b.state));
const sortedData = data?.sort((a, b) => a.state.localeCompare(b.state)) || [];
const collectLabel = sortedData.map(d => d.state);
// Align Data to Timesteps, Introduces 'undefied' as placeholder, reiterate and set those to 0
const collectData = uPlot.join(sortedData.map(d => [d.times, d.counts])).map(d => d.map(i => i ? i : 0));
const collectData = (sortedData.length > 0) ? uPlot.join(sortedData.map(d => [d.times, d.counts])).map(d => d.map(i => i ? i : 0)) : [];
// STACKED CHART FUNCTIONS //
function stack(data, omit) {
@@ -156,7 +156,7 @@
{
scale: "y",
grid: { show: true },
labelFont: "sans-serif",
// labelFont: "sans-serif",
label: ylabel + (yunit ? ` (${yunit})` : ''),
// values: (u, vals) => vals.map((v) => formatNumber(v)),
},
@@ -344,7 +344,7 @@
</script>
<!-- Define $width Wrapper and NoData Card -->
{#if data && collectData[0].length > 0}
{#if data && collectData.length > 0}
<div bind:this={plotWrapper} bind:clientWidth={width}
style="background-color: rgba(255, 255, 255, 1.0);" class="rounded"
></div>
@@ -64,6 +64,34 @@
{/each}
</DropdownMenu>
</Dropdown>
{:else if item.title === 'Status'}
<Dropdown nav inNavbar {direction}>
<DropdownToggle nav caret>
<Icon name={item.icon} />
{item.title}
</DropdownToggle>
<DropdownMenu class="dropdown-menu-lg-end">
{#each clusters as cluster}
<Dropdown nav direction="right">
<DropdownToggle nav caret class="dropdown-item py-1 px-2">
{cluster.name}
</DropdownToggle>
<DropdownMenu>
<DropdownItem class="py-1 px-2"
href={item.href + cluster.name}
>
Status Dashboard
</DropdownItem>
<DropdownItem class="py-1 px-2"
href={item.href + 'detail/' + cluster.name}
>
Status Details
</DropdownItem>
</DropdownMenu>
</Dropdown>
{/each}
</DropdownMenu>
</Dropdown>
{:else}
<Dropdown nav inNavbar {direction}>
<DropdownToggle nav caret>
+1
View File
@@ -6,6 +6,7 @@ mount(Status, {
target: document.getElementById('svelte-app'),
props: {
presetCluster: infos.cluster,
displayType: displayType,
},
context: new Map([
['cc-config', clusterCockpitConfig]
@@ -0,0 +1,82 @@
<!--
@component Main cluster status view component; renders current system-usage information
Properties:
- `presetCluster String`: The cluster to show status information for
-->
<script>
import {
getContext
} from "svelte"
import {
init,
} from "../generic/utils.js";
import {
Row,
Col,
Card,
CardBody,
TabContent,
TabPane,
Spinner
} from "@sveltestrap/sveltestrap";
import StatusDash from "./dashdetails/StatusDash.svelte";
import UsageDash from "./dashdetails/UsageDash.svelte";
import StatisticsDash from "./dashdetails/StatisticsDash.svelte";
/* Svelte 5 Props */
let {
presetCluster,
} = $props();
/*Const Init */
const { query: initq } = init();
const useCbColors = getContext("cc-config")?.plotConfiguration_colorblindMode || false
</script>
<!-- Loading indicator & Refresh -->
<Row cols={1} class="mb-2">
<Col>
<h3 class="mb-0">Current Status of Cluster "{presetCluster.charAt(0).toUpperCase() + presetCluster.slice(1)}"</h3>
</Col>
</Row>
{#if $initq.fetching}
<Row cols={1} class="text-center mt-3">
<Col>
<Spinner />
</Col>
</Row>
{:else if $initq.error}
<Row cols={1} class="text-center mt-3">
<Col>
<Card body color="danger">{$initq.error.message}</Card>
</Col>
</Row>
{:else}
<Card class="overflow-auto" style="height: auto;">
<TabContent>
<TabPane tabId="status-dash" tab="Status" active>
<CardBody>
<StatusDash clusters={$initq.data.clusters} {presetCluster} {useCbColors} useAltColors></StatusDash>
</CardBody>
</TabPane>
<TabPane tabId="usage-dash" tab="Usage">
<CardBody>
<UsageDash {presetCluster} {useCbColors}></UsageDash>
</CardBody>
</TabPane>
<TabPane tabId="metric-dash" tab="Statistics">
<CardBody>
<StatisticsDash {presetCluster} {useCbColors}></StatisticsDash>
</CardBody>
</TabPane>
</TabContent>
</Card>
{/if}
+600
View File
@@ -0,0 +1,600 @@
<!--
@component Main cluster status view component; renders current system-usage information
Properties:
- `presetCluster String`: The cluster to show status information for
-->
<script>
import {
getContext
} from "svelte"
import {
queryStore,
gql,
getContextClient,
} from "@urql/svelte";
import {
init,
scramble,
scrambleNames,
} from "../generic/utils.js";
import {
formatDurationTime,
formatNumber,
} from "../generic/units.js";
import {
Row,
Col,
Card,
CardTitle,
CardHeader,
CardBody,
Spinner,
Table,
Progress,
Icon,
} from "@sveltestrap/sveltestrap";
import Roofline from "../generic/plots/Roofline.svelte";
import Pie, { colors } from "../generic/plots/Pie.svelte";
import Stacked from "../generic/plots/Stacked.svelte";
import DoubleMetric from "../generic/plots/DoubleMetricPlot.svelte";
import Refresher from "../generic/helper/Refresher.svelte";
/* Svelte 5 Props */
let {
presetCluster,
} = $props();
/*Const Init */
const { query: initq } = init();
const client = getContextClient();
const useCbColors = getContext("cc-config")?.plotConfiguration_colorblindMode || false
/* States */
let pagingState = $state({page: 1, itemsPerPage: 10}) // Top 10
let from = $state(new Date(Date.now() - 5 * 60 * 1000));
let clusterFrom = $state(new Date(Date.now() - (8 * 60 * 60 * 1000)));
let to = $state(new Date(Date.now()));
let stackedFrom = $state(Math.floor(Date.now() / 1000) - 14400);
let colWidthJobs = $state(0);
let colWidthRoof = $state(0);
let colWidthTotals =$state(0);
let colWidthStacked1 = $state(0);
let colWidthStacked2 = $state(0);
/* Derived */
// States for Stacked charts
const statesTimed = $derived(queryStore({
client: client,
query: gql`
query ($filter: [NodeFilter!], $typeNode: String!, $typeHealth: String!) {
nodeStates: nodeStatesTimed(filter: $filter, type: $typeNode) {
state
counts
times
}
healthStates: nodeStatesTimed(filter: $filter, type: $typeHealth) {
state
counts
times
}
}
`,
variables: {
filter: { cluster: { eq: presetCluster }, timeStart: stackedFrom},
typeNode: "node",
typeHealth: "health"
},
requestPolicy: "network-only"
}));
// Note: nodeMetrics are requested on configured $timestep resolution
// Result: The latest 5 minutes (datapoints) for each node independent of job
const statusQuery = $derived(queryStore({
client: client,
query: gql`
query (
$cluster: String!
$metrics: [String!]
$from: Time!
$to: Time!
$clusterFrom: Time!
$jobFilter: [JobFilter!]!
$paging: PageRequest!
$sorting: OrderByInput!
) {
# Node 5 Minute Averages for Roofline
nodeMetrics(
cluster: $cluster
metrics: $metrics
from: $from
to: $to
) {
host
subCluster
metrics {
name
metric {
series {
statistics {
avg
}
}
}
}
}
# Running Job Metric Average for Rooflines
jobsMetricStats(filter: $jobFilter, metrics: $metrics) {
id
jobId
duration
numNodes
numAccelerators
subCluster
stats {
name
data {
avg
}
}
}
# Get Jobs for Per-Node Counts
jobs(filter: $jobFilter, order: $sorting, page: $paging) {
items {
jobId
resources {
hostname
}
}
count
}
# Only counts shared nodes once
allocatedNodes(cluster: $cluster) {
name
count
}
# totalNodes includes multiples if shared jobs: Info-Card Data
jobsStatistics(
filter: $jobFilter
page: $paging
sortBy: TOTALJOBS
groupBy: SUBCLUSTER
) {
id
totalJobs
totalUsers
totalCores
totalAccs
}
# ClusterMetrics for doubleMetricPlot
clusterMetrics(
cluster: $cluster
metrics: $metrics
from: $clusterFrom
to: $to
) {
nodeCount
metrics {
name
unit {
prefix
base
}
timestep
data
}
}
}
`,
variables: {
cluster: presetCluster,
metrics: ["flops_any", "mem_bw"], // Fixed names for roofline and status bars
from: from.toISOString(),
to: to.toISOString(),
clusterFrom: clusterFrom.toISOString(),
jobFilter: [{ state: ["running"] }, { cluster: { eq: presetCluster } }],
paging: { itemsPerPage: -1, page: 1 }, // Get all: -1
sorting: { field: "startTime", type: "col", order: "DESC" }
},
requestPolicy: "network-only"
}));
const topJobsQuery = $derived(queryStore({
client: client,
query: gql`
query (
$filter: [JobFilter!]!
$paging: PageRequest!
) {
jobsStatistics(
filter: $filter
page: $paging
sortBy: TOTALJOBS
groupBy: PROJECT
) {
id
totalJobs
}
}
`,
variables: {
filter: [{ state: ["running"] }, { cluster: { eq: presetCluster} }],
paging: pagingState // Top 10
},
requestPolicy: "network-only"
}));
const clusterInfo = $derived.by(() => {
if ($initq?.data?.clusters) {
let rawInfos = {};
let subClusters = $initq?.data?.clusters?.find((c) => c.name == presetCluster)?.subClusters || [];
for (let subCluster of subClusters) {
// Allocations
if (!rawInfos['allocatedNodes']) rawInfos['allocatedNodes'] = $statusQuery?.data?.allocatedNodes?.find(({ name }) => name == subCluster.name)?.count || 0;
else rawInfos['allocatedNodes'] += $statusQuery?.data?.allocatedNodes?.find(({ name }) => name == subCluster.name)?.count || 0;
if (!rawInfos['allocatedCores']) rawInfos['allocatedCores'] = $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalCores || 0;
else rawInfos['allocatedCores'] += $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalCores || 0;
if (!rawInfos['allocatedAccs']) rawInfos['allocatedAccs'] = $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalAccs || 0;
else rawInfos['allocatedAccs'] += $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalAccs || 0;
// Infos
if (!rawInfos['processorTypes']) rawInfos['processorTypes'] = subCluster?.processorType ? new Set([subCluster.processorType]) : new Set([]);
else rawInfos['processorTypes'].add(subCluster.processorType);
if (!rawInfos['activeUsers']) rawInfos['activeUsers'] = $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalUsers || 0;
else rawInfos['activeUsers'] += $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalUsers || 0;
if (!rawInfos['runningJobs']) rawInfos['runningJobs'] = $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalJobs || 0;
else rawInfos['runningJobs'] += $statusQuery?.data?.jobsStatistics?.find(({ id }) => id == subCluster.name)?.totalJobs || 0;
if (!rawInfos['totalNodes']) rawInfos['totalNodes'] = subCluster?.numberOfNodes || 0;
else rawInfos['totalNodes'] += subCluster?.numberOfNodes || 0;
if (!rawInfos['totalCores']) rawInfos['totalCores'] = (subCluster?.socketsPerNode * subCluster?.coresPerSocket * subCluster?.numberOfNodes) || 0;
else rawInfos['totalCores'] += (subCluster?.socketsPerNode * subCluster?.coresPerSocket * subCluster?.numberOfNodes) || 0;
if (!rawInfos['totalAccs']) rawInfos['totalAccs'] = (subCluster?.numberOfNodes * subCluster?.topology?.accelerators?.length) || 0;
else rawInfos['totalAccs'] += (subCluster?.numberOfNodes * subCluster?.topology?.accelerators?.length) || 0;
// Units (Set Once)
if (!rawInfos['flopRateUnit']) rawInfos['flopRateUnit'] = subCluster.flopRateSimd.unit.prefix + subCluster.flopRateSimd.unit.base
if (!rawInfos['memBwRateUnit']) rawInfos['memBwRateUnit'] = subCluster.memoryBandwidth.unit.prefix + subCluster.memoryBandwidth.unit.base
// Get Maxima For Roofline Knee Render
if (!rawInfos['roofData']) {
rawInfos['roofData'] = {
flopRateScalar: {value: subCluster.flopRateScalar.value},
flopRateSimd: {value: subCluster.flopRateSimd.value},
memoryBandwidth: {value: subCluster.memoryBandwidth.value}
};
} else {
rawInfos['roofData']['flopRateScalar']['value'] = Math.max(rawInfos['roofData']['flopRateScalar']['value'], subCluster.flopRateScalar.value)
rawInfos['roofData']['flopRateSimd']['value'] = Math.max(rawInfos['roofData']['flopRateSimd']['value'], subCluster.flopRateSimd.value)
rawInfos['roofData']['memoryBandwidth']['value'] = Math.max(rawInfos['roofData']['memoryBandwidth']['value'], subCluster.memoryBandwidth.value)
}
}
// Keymetrics (Data on Cluster-Scope)
let rawFlops = $statusQuery?.data?.nodeMetrics?.reduce((sum, node) =>
sum + (node.metrics.find((m) => m.name == 'flops_any')?.metric?.series[0]?.statistics?.avg || 0),
0, // Initial Value
) || 0;
rawInfos['flopRate'] = Math.floor((rawFlops * 100) / 100)
let rawMemBw = $statusQuery?.data?.nodeMetrics?.reduce((sum, node) =>
sum + (node.metrics.find((m) => m.name == 'mem_bw')?.metric?.series[0]?.statistics?.avg || 0),
0, // Initial Value
) || 0;
rawInfos['memBwRate'] = Math.floor((rawMemBw * 100) / 100)
return rawInfos
} else {
return {};
}
});
/* Functions */
function legendColors(targetIdx) {
// Reuses first color if targetIdx overflows
let c;
if (useCbColors) {
c = [...colors['colorblind']];
// } else if (useAltColors) {
// c = [...colors['alternative']];
} else {
c = [...colors['default']];
}
return c[(c.length + targetIdx) % c.length];
}
function transformJobsStatsToData(clusterData) {
/* c will contain values from 0 to 1 representing the duration */
let data = null
const x = [], y = [], c = [], day = 86400.0
if (clusterData) {
for (let i = 0; i < clusterData.length; i++) {
const flopsData = clusterData[i].stats.find((s) => s.name == "flops_any")
const memBwData = clusterData[i].stats.find((s) => s.name == "mem_bw")
const f = flopsData.data.avg
const m = memBwData.data.avg
const d = clusterData[i].duration / day
const intensity = f / m
if (Number.isNaN(intensity) || !Number.isFinite(intensity))
continue
x.push(intensity)
y.push(f)
// Long Jobs > 1 Day: Use max Color
if (d > 1.0) c.push(1.0)
else c.push(d)
}
} else {
console.warn("transformJobsStatsToData: metrics for 'mem_bw' and/or 'flops_any' missing!")
}
if (x.length > 0 && y.length > 0 && c.length > 0) {
data = [null, [x, y], c] // for dataformat see roofline.svelte
}
return data
}
function transformJobsStatsToInfo(clusterData) {
if (clusterData) {
return clusterData.map((sc) => { return {id: sc.id, jobId: sc.jobId, numNodes: sc.numNodes, numAcc: sc?.numAccelerators? sc.numAccelerators : 0, duration: formatDurationTime(sc.duration)} })
} else {
console.warn("transformJobsStatsToInfo: jobInfo missing!")
return []
}
}
</script>
<Card style="height: 88vh;">
<CardBody class="align-content-center">
<Row>
<Col>
<Refresher
hideSelector
initially={60}
onRefresh={(interval) => {
from = new Date(Date.now() - 5 * 60 * 1000);
to = new Date(Date.now());
clusterFrom = new Date(Date.now() - (8 * 60 * 60 * 1000))
pagingState = { page:1, itemsPerPage: 10 };
if (interval) stackedFrom += Math.floor(interval / 1000);
else stackedFrom += 1 // Workaround: TimeSelection not linked, just trigger new data on manual refresh
}}
/>
</Col>
</Row>
{#if $statusQuery.fetching || $statesTimed.fetching || $topJobsQuery.fetching}
<Row class="justify-content-center">
<Col xs="auto">
<Spinner />
</Col>
</Row>
{:else if $statusQuery.error || $statesTimed.error || $topJobsQuery.error}
<Row>
{#if $statusQuery.error}
<Col>
<Card color="danger"><CardBody>Error Requesting Status Data: {$statusQuery.error.message}</CardBody></Card>
</Col>
{/if}
{#if $statesTimed.error}
<Col>
<Card color="danger"><CardBody>Error Requesting Node Scheduler States: {$statesTimed.error.message}</CardBody></Card>
</Col>
{/if}
{#if $topJobsQuery.error}
<Col>
<Card color="danger"><CardBody>Error Requesting Jobs By Project: {$topJobsQuery.error.message}</CardBody></Card>
</Col>
{/if}
</Row>
{:else}
<Row cols={{xs:1, md:2, xl: 3}}>
<Col> <!-- Info Card -->
<Card class="h-auto mt-1">
<CardHeader>
<CardTitle class="mb-0">Cluster "{presetCluster.charAt(0).toUpperCase() + presetCluster.slice(1)}"</CardTitle>
<span>{[...clusterInfo?.processorTypes].join(', ')}</span>
</CardHeader>
<CardBody>
<Table borderless>
<tr class="py-2">
<td style="font-size:x-large;">{clusterInfo?.runningJobs} Running Jobs</td>
<td colspan="2" style="font-size:x-large;">{clusterInfo?.activeUsers} Active Users</td>
</tr>
<hr class="my-1"/>
<tr class="pt-2">
<td style="font-size: large;">
Flop Rate (<span style="cursor: help;" title="Flops[Any] = (Flops[Double] x 2) + Flops[Single]">Any</span>)
</td>
<td colspan="2" style="font-size: large;">
Memory BW Rate
</td>
</tr>
<tr class="pb-2">
<td style="font-size:x-large;">
{clusterInfo?.flopRate}
{clusterInfo?.flopRateUnit}
</td>
<td colspan="2" style="font-size:x-large;">
{clusterInfo?.memBwRate}
{clusterInfo?.memBwRateUnit}
</td>
</tr>
<hr class="my-1"/>
<tr class="py-2">
<th scope="col">Allocated Nodes</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={clusterInfo?.allocatedNodes}
max={clusterInfo?.totalNodes}
/>
</div></td
>
<td
>{clusterInfo?.allocatedNodes} / {clusterInfo?.totalNodes}
Nodes</td
>
</tr>
<tr class="py-2">
<th scope="col">Allocated Cores</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={clusterInfo?.allocatedCores}
max={clusterInfo?.totalCores}
/>
</div></td
>
<td
>{formatNumber(clusterInfo?.allocatedCores)} / {formatNumber(clusterInfo?.totalCores)}
Cores</td
>
</tr>
{#if clusterInfo?.totalAccs !== 0}
<tr class="py-2">
<th scope="col">Allocated Accelerators</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={clusterInfo?.allocatedAccs}
max={clusterInfo?.totalAccs}
/>
</div></td
>
<td
>{clusterInfo?.allocatedAccs} / {clusterInfo?.totalAccs}
Accelerators</td
>
</tr>
{/if}
</Table>
</CardBody>
</Card>
</Col>
<Col> <!-- Pie Jobs -->
{#if topJobsQuery?.data?.jobsStatistics?.length > 0}
<Row cols={{xs:1, md:2}}>
<Col class="p-2">
<div bind:clientWidth={colWidthJobs}>
<h4 class="text-center">
Top Projects: Jobs
</h4>
<Pie
{useCbColors}
canvasId="hpcpie-jobs-projects"
size={colWidthJobs * 0.75}
sliceLabel={'Jobs'}
quantities={$topJobsQuery.data.jobsStatistics.map(
(tp) => tp['totalJobs'],
)}
entities={$topJobsQuery.data.jobsStatistics.map((tp) => scrambleNames ? scramble(tp.id) : tp.id)}
/>
</div>
</Col>
<Col class="p-2">
<Table>
<tr class="mb-2">
<th></th>
<th style="padding-left: 0.5rem;">Project</th>
<th>Jobs</th>
</tr>
{#each $topJobsQuery.data.jobsStatistics as tp, i}
<tr>
<td><Icon name="circle-fill" style="color: {legendColors(i)};" /></td>
<td>
<a target="_blank" href="/monitoring/jobs/?cluster={presetCluster}&state=running&project={tp.id}&projectMatch=eq"
>{scrambleNames ? scramble(tp.id) : tp.id}
</a>
</td>
<td>{tp['totalJobs']}</td>
</tr>
{/each}
</Table>
</Col>
</Row>
{:else}
<Card body color="warning" class="mx-4 my-2"
>Cannot render job status: No state data returned for <code>Pie Chart</code></Card
>
{/if}
</Col>
<Col> <!-- Job Roofline -->
<div bind:clientWidth={colWidthRoof}>
{#key $statusQuery?.data?.jobsMetricStats}
<Roofline
useColors={true}
allowSizeChange
width={colWidthRoof - 10}
height={300}
subCluster={clusterInfo?.roofData ? clusterInfo.roofData : null}
roofData={transformJobsStatsToData($statusQuery?.data?.jobsMetricStats)}
jobsData={transformJobsStatsToInfo($statusQuery?.data?.jobsMetricStats)}
/>
{/key}
</div>
</Col>
<Col> <!-- Total Cluster Metric in Time SUMS-->
<div bind:clientWidth={colWidthTotals}>
<DoubleMetric
width={colWidthTotals}
timestep={$statusQuery?.data?.clusterMetrics[0]?.timestep || 60}
numNodes={$statusQuery?.data?.clusterMetrics?.nodeCount || 0}
metricData={$statusQuery?.data?.clusterMetrics?.metrics || []}
cluster={presetCluster}
fixLinewidth={2}
/>
</div>
</Col>
<Col> <!-- Stacked SchedState -->
<div bind:clientWidth={colWidthStacked1}>
{#key $statesTimed?.data?.nodeStates}
<Stacked
data={$statesTimed?.data?.nodeStates}
width={colWidthStacked1 * 0.95}
xlabel="Time"
ylabel="Nodes"
yunit = "#Count"
title = "Node States"
stateType = "Node"
/>
{/key}
</div>
</Col>
<Col> <!-- Stacked Healthstate -->
<div bind:clientWidth={colWidthStacked2}>
{#key $statesTimed?.data?.healthStates}
<Stacked
data={$statesTimed?.data?.healthStates}
width={colWidthStacked2 * 0.95}
xlabel="Time"
ylabel="Nodes"
yunit = "#Count"
title = "Health States"
stateType = "Health"
/>
{/key}
</div>
</Col>
</Row>
{/if}
</CardBody>
</Card>
@@ -22,11 +22,11 @@
} from "@urql/svelte";
import {
convert2uplot,
} from "../generic/utils.js";
import PlotGrid from "../generic/PlotGrid.svelte";
import Histogram from "../generic/plots/Histogram.svelte";
import HistogramSelection from "../generic/select/HistogramSelection.svelte";
import Refresher from "../generic/helper/Refresher.svelte";
} from "../../generic/utils.js";
import PlotGrid from "../../generic/PlotGrid.svelte";
import Histogram from "../../generic/plots/Histogram.svelte";
import HistogramSelection from "../../generic/select/HistogramSelection.svelte";
import Refresher from "../../generic/helper/Refresher.svelte";
/* Svelte 5 Props */
let {
@@ -22,12 +22,12 @@
gql,
getContextClient,
} from "@urql/svelte";
import { formatDurationTime } from "../generic/units.js";
import Refresher from "../generic/helper/Refresher.svelte";
import TimeSelection from "../generic/select/TimeSelection.svelte";
import Roofline from "../generic/plots/Roofline.svelte";
import Pie, { colors } from "../generic/plots/Pie.svelte";
import Stacked from "../generic/plots/Stacked.svelte";
import { formatDurationTime } from "../../generic/units.js";
import Refresher from "../../generic/helper/Refresher.svelte";
import TimeSelection from "../../generic/select/TimeSelection.svelte";
import Roofline from "../../generic/plots/Roofline.svelte";
import Pie, { colors } from "../../generic/plots/Pie.svelte";
import Stacked from "../../generic/plots/Stacked.svelte";
/* Svelte 5 Props */
let {
@@ -402,7 +402,7 @@
to = new Date(Date.now());
if (interval) stackedFrom += Math.floor(interval / 1000);
else stackedFrom += 1 // Workaround: TineSelection not linked, just trigger new data on manual refresh
else stackedFrom += 1 // Workaround: TimeSelection not linked, just trigger new data on manual refresh
}}
/>
</Col>
@@ -27,10 +27,10 @@
scramble,
scrambleNames,
convert2uplot,
} from "../generic/utils.js";
import Pie, { colors } from "../generic/plots/Pie.svelte";
import Histogram from "../generic/plots/Histogram.svelte";
import Refresher from "../generic/helper/Refresher.svelte";
} from "../../generic/utils.js";
import Pie, { colors } from "../../generic/plots/Pie.svelte";
import Histogram from "../../generic/plots/Histogram.svelte";
import Refresher from "../../generic/helper/Refresher.svelte";
/* Svelte 5 Props */
let {
+16 -1
View File
@@ -23,6 +23,20 @@
</script>
</head>
<body class="site">
{{if eq .Infos.displayType "PUBLIC"}}
<main>
<div class="container">
{{block "content-public" .}}
Whoops, you should not see this... [PUBLIC]
{{end}}
</div>
</main>
{{block "javascript-public" .}}
Whoops, you should not see this... [JS PUBLIC]
{{end}}
{{else}}
{{block "navigation" .}}
<header id="svelte-header"></header>
{{end}}
@@ -30,7 +44,7 @@
<main class="site-content">
<div class="container">
{{block "content" .}}
Whoops, you should not see this...
Whoops, you should not see this... [MAIN]
{{end}}
</div>
</main>
@@ -52,5 +66,6 @@
{{block "javascript" .}}
<script src='/build/header.js'></script>
{{end}}
{{end}}
</body>
</html>
+14
View File
@@ -0,0 +1,14 @@
{{define "content-public"}}
<div id="svelte-app"></div>
{{end}}
{{define "stylesheets"}}
<link rel='stylesheet' href='/build/dashpublic.css'>
{{end}}
{{define "javascript-public"}}
<script>
const presetCluster = {{ .Infos.cluster }};
const clusterCockpitConfig = {{ .Config }};
</script>
<script src='/build/dashpublic.js'></script>
{{end}}
+1
View File
@@ -8,6 +8,7 @@
{{define "javascript"}}
<script>
const infos = {{ .Infos }};
const displayType = {{ .Infos.displayType }};
const clusterCockpitConfig = {{ .Config }};
</script>
<script src='/build/status.js'></script>