ClusterCockpit/cc-backend
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-backend synced 2024-12-26 05:19:05 +01:00
Code Issues Packages Releases Activity

Merge branch 'master' into 137-improve-layout-for-smaller-screen-sizes

Browse Source
This commit is contained in:
Jan Eitzinger 2023-08-31 13:05:27 +02:00 committed by GitHub
commit a0b8f36dbb
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
23 changed files with 1355 additions and 661 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
api/schema.graphqls
View File

@ -156,12 +156,18 @@ type MetricFootprints {
}
type Footprints {
nodehours: [NullableFloat!]!
timeWeights: TimeWeights!
metrics: [MetricFootprints!]!
}
type TimeWeights {
nodeHours: [NullableFloat!]!
accHours: [NullableFloat!]!
coreHours: [NullableFloat!]!
}
enum Aggregate { USER, PROJECT, CLUSTER }
enum Weights { NODE_COUNT, NODE_HOURS }
enum SortByAggregate { TOTALWALLTIME, TOTALJOBS, TOTALNODES, TOTALNODEHOURS, TOTALCORES, TOTALCOREHOURS, TOTALACCS, TOTALACCHOURS }
type NodeMetrics {
host: String!
@ -192,8 +198,7 @@ type Query {
jobsFootprints(filter: [JobFilter!], metrics: [String!]!): Footprints
jobs(filter: [JobFilter!], page: PageRequest, order: OrderByInput): JobResultList!
jobsStatistics(filter: [JobFilter!], groupBy: Aggregate): [JobsStatistics!]!
jobsCount(filter: [JobFilter]!, groupBy: Aggregate!, weight: Weights, limit: Int): [Count!]!
jobsStatistics(filter: [JobFilter!], page: PageRequest, sortBy: SortByAggregate, groupBy: Aggregate): [JobsStatistics!]!
rooflineHeatmap(filter: [JobFilter!]!, rows: Int!, cols: Int!, minX: Float!, minY: Float!, maxX: Float!, maxY: Float!): [[Float!]!]!
@ -288,11 +293,16 @@ type JobsStatistics {
runningJobs: Int! # Number of running jobs
shortJobs: Int! # Number of jobs with a duration of less than duration
totalWalltime: Int! # Sum of the duration of all matched jobs in hours
totalNodes: Int! # Sum of the nodes of all matched jobs
totalNodeHours: Int! # Sum of the node hours of all matched jobs
totalCores: Int! # Sum of the cores of all matched jobs
totalCoreHours: Int! # Sum of the core hours of all matched jobs
totalAccs: Int! # Sum of the accs of all matched jobs
totalAccHours: Int! # Sum of the gpu hours of all matched jobs
histDuration: [HistoPoint!]! # value: hour, count: number of jobs with a rounded duration of value
histNumNodes: [HistoPoint!]! # value: number of nodes, count: number of jobs with that number of nodes
histNumCores: [HistoPoint!]! # value: number of cores, count: number of jobs with that number of cores
histNumAccs: [HistoPoint!]! # value: number of accs, count: number of jobs with that number of accs
}
input PageRequest {

1
configs/README.md
View File

@ -32,6 +32,7 @@ It is supported to set these by means of a `.env` file in the project root.
- `search_dn`: Type string. DN for authenticating LDAP admin account with general read rights.
- `user_bind`: Type string. Expression used to authenticate users via LDAP bind. Must contain `uid={username}`.
- `user_filter`: Type string. Filter to extract users for syncing.
- `username_attr`: Type string. Attribute with full user name. Defaults to `gecos` if not provided.
- `sync_interval`: Type string. Interval used for syncing local user table with LDAP directory. Parsed using time.ParseDuration.
- `sync_del_old_users`: Type bool. Delete obsolete users in database.
* `clusters`: Type array of objects

23
internal/auth/ldap.go
View File

@ -21,6 +21,7 @@ import (
type LdapAuthenticator struct {
syncPassword string
UserAttr string
}
var _ Authenticator = (*LdapAuthenticator)(nil)
@ -31,11 +32,13 @@ func (la *LdapAuthenticator) Init() error {
log.Warn("environment variable 'LDAP_ADMIN_PASSWORD' not set (ldap sync will not work)")
}
if config.Keys.LdapConfig.SyncInterval != "" {
interval, err := time.ParseDuration(config.Keys.LdapConfig.SyncInterval)
lc := config.Keys.LdapConfig
if lc.SyncInterval != "" {
interval, err := time.ParseDuration(lc.SyncInterval)
if err != nil {
log.Warnf("Could not parse duration for sync interval: %v",
config.Keys.LdapConfig.SyncInterval)
lc.SyncInterval)
return err
}
@ -58,6 +61,12 @@ func (la *LdapAuthenticator) Init() error {
log.Info("LDAP configuration key sync_interval invalid")
}
if lc.UserAttr != "" {
la.UserAttr = lc.UserAttr
} else {
la.UserAttr = "gecos"
}
return nil
}
@ -86,7 +95,7 @@ func (la *LdapAuthenticator) CanLogin(
lc.UserBase,
ldap.ScopeWholeSubtree, ldap.NeverDerefAliases, 0, 0, false,
fmt.Sprintf("(&%s(uid=%s))", lc.UserFilter, username),
[]string{"dn", "uid", "gecos"}, nil)
[]string{"dn", "uid", la.UserAttr}, nil)
sr, err := l.Search(searchRequest)
if err != nil {
@ -100,7 +109,7 @@ func (la *LdapAuthenticator) CanLogin(
}
entry := sr.Entries[0]
name := entry.GetAttributeValue("gecos")
name := entry.GetAttributeValue(la.UserAttr)
var roles []string
roles = append(roles, schema.GetRoleString(schema.RoleUser))
projects := make([]string, 0)
@ -176,7 +185,7 @@ func (la *LdapAuthenticator) Sync() error {
lc.UserBase,
ldap.ScopeWholeSubtree, ldap.NeverDerefAliases, 0, 0, false,
lc.UserFilter,
[]string{"dn", "uid", "gecos"}, nil))
[]string{"dn", "uid", la.UserAttr}, nil))
if err != nil {
log.Warn("LDAP search error")
return err
@ -192,7 +201,7 @@ func (la *LdapAuthenticator) Sync() error {
_, ok := users[username]
if !ok {
users[username] = IN_LDAP
newnames[username] = entry.GetAttributeValue("gecos")
newnames[username] = entry.GetAttributeValue(la.UserAttr)
} else {
users[username] = IN_BOTH
}

34
internal/config/config.go
View File

@ -26,21 +26,25 @@ var Keys schema.ProgramConfig = schema.ProgramConfig{
StopJobsExceedingWalltime: 0,
ShortRunningJobsDuration: 5 * 60,
UiDefaults: map[string]interface{}{
"analysis_view_histogramMetrics": []string{"flops_any", "mem_bw", "mem_used"},
"analysis_view_scatterPlotMetrics": [][]string{{"flops_any", "mem_bw"}, {"flops_any", "cpu_load"}, {"cpu_load", "mem_bw"}},
"job_view_nodestats_selectedMetrics": []string{"flops_any", "mem_bw", "mem_used"},
"job_view_polarPlotMetrics": []string{"flops_any", "mem_bw", "mem_used"},
"job_view_selectedMetrics": []string{"flops_any", "mem_bw", "mem_used"},
"plot_general_colorBackground": true,
"plot_general_colorscheme": []string{"#00bfff", "#0000ff", "#ff00ff", "#ff0000", "#ff8000", "#ffff00", "#80ff00"},
"plot_general_lineWidth": 3,
"plot_list_jobsPerPage": 50,
"plot_list_selectedMetrics": []string{"cpu_load", "mem_used", "flops_any", "mem_bw"},
"plot_view_plotsPerRow": 3,
"plot_view_showPolarplot": true,
"plot_view_showRoofline": true,
"plot_view_showStatTable": true,
"system_view_selectedMetric": "cpu_load",
"analysis_view_histogramMetrics": []string{"flops_any", "mem_bw", "mem_used"},
"analysis_view_scatterPlotMetrics": [][]string{{"flops_any", "mem_bw"}, {"flops_any", "cpu_load"}, {"cpu_load", "mem_bw"}},
"job_view_nodestats_selectedMetrics": []string{"flops_any", "mem_bw", "mem_used"},
"job_view_polarPlotMetrics": []string{"flops_any", "mem_bw", "mem_used"},
"job_view_selectedMetrics": []string{"flops_any", "mem_bw", "mem_used"},
"plot_general_colorBackground": true,
"plot_general_colorscheme": []string{"#00bfff", "#0000ff", "#ff00ff", "#ff0000", "#ff8000", "#ffff00", "#80ff00"},
"plot_general_lineWidth": 3,
"plot_list_jobsPerPage": 50,
"plot_list_selectedMetrics": []string{"cpu_load", "mem_used", "flops_any", "mem_bw"},
"plot_view_plotsPerRow": 3,
"plot_view_showPolarplot": true,
"plot_view_showRoofline": true,
"plot_view_showStatTable": true,
"system_view_selectedMetric": "cpu_load",
"analysis_view_selectedTopEntity": "user",
"analysis_view_selectedTopCategory": "totalWalltime",
"status_view_selectedTopUserCategory": "totalJobs",
"status_view_selectedTopProjectCategory": "totalJobs",
},
}

825
internal/graph/generated/generated.go
View File

File diff suppressed because it is too large Load Diff

109
internal/graph/model/models_gen.go
View File

@ -22,8 +22,8 @@ type FloatRange struct {
}
type Footprints struct {
Nodehours []schema.Float `json:"nodehours"`
Metrics []*MetricFootprints `json:"metrics"`
TimeWeights *TimeWeights `json:"timeWeights"`
Metrics []*MetricFootprints `json:"metrics"`
}
type HistoPoint struct {
@ -91,11 +91,16 @@ type JobsStatistics struct {
RunningJobs int `json:"runningJobs"`
ShortJobs int `json:"shortJobs"`
TotalWalltime int `json:"totalWalltime"`
TotalNodes int `json:"totalNodes"`
TotalNodeHours int `json:"totalNodeHours"`
TotalCores int `json:"totalCores"`
TotalCoreHours int `json:"totalCoreHours"`
TotalAccs int `json:"totalAccs"`
TotalAccHours int `json:"totalAccHours"`
HistDuration []*HistoPoint `json:"histDuration"`
HistNumNodes []*HistoPoint `json:"histNumNodes"`
HistNumCores []*HistoPoint `json:"histNumCores"`
HistNumAccs []*HistoPoint `json:"histNumAccs"`
}
type MetricFootprints struct {
@ -133,6 +138,12 @@ type TimeRangeOutput struct {
To time.Time `json:"to"`
}
type TimeWeights struct {
NodeHours []schema.Float `json:"nodeHours"`
AccHours []schema.Float `json:"accHours"`
CoreHours []schema.Float `json:"coreHours"`
}
type User struct {
Username string `json:"username"`
Name string `json:"name"`
@ -182,6 +193,59 @@ func (e Aggregate) MarshalGQL(w io.Writer) {
fmt.Fprint(w, strconv.Quote(e.String()))
}
type SortByAggregate string
const (
SortByAggregateTotalwalltime SortByAggregate = "TOTALWALLTIME"
SortByAggregateTotaljobs SortByAggregate = "TOTALJOBS"
SortByAggregateTotalnodes SortByAggregate = "TOTALNODES"
SortByAggregateTotalnodehours SortByAggregate = "TOTALNODEHOURS"
SortByAggregateTotalcores SortByAggregate = "TOTALCORES"
SortByAggregateTotalcorehours SortByAggregate = "TOTALCOREHOURS"
SortByAggregateTotalaccs SortByAggregate = "TOTALACCS"
SortByAggregateTotalacchours SortByAggregate = "TOTALACCHOURS"
)
var AllSortByAggregate = []SortByAggregate{
SortByAggregateTotalwalltime,
SortByAggregateTotaljobs,
SortByAggregateTotalnodes,
SortByAggregateTotalnodehours,
SortByAggregateTotalcores,
SortByAggregateTotalcorehours,
SortByAggregateTotalaccs,
SortByAggregateTotalacchours,
}
func (e SortByAggregate) IsValid() bool {
switch e {
case SortByAggregateTotalwalltime, SortByAggregateTotaljobs, SortByAggregateTotalnodes, SortByAggregateTotalnodehours, SortByAggregateTotalcores, SortByAggregateTotalcorehours, SortByAggregateTotalaccs, SortByAggregateTotalacchours:
return true
}
return false
}
func (e SortByAggregate) String() string {
return string(e)
}
func (e *SortByAggregate) UnmarshalGQL(v interface{}) error {
str, ok := v.(string)
if !ok {
return fmt.Errorf("enums must be strings")
}
*e = SortByAggregate(str)
if !e.IsValid() {
return fmt.Errorf("%s is not a valid SortByAggregate", str)
}
return nil
}
func (e SortByAggregate) MarshalGQL(w io.Writer) {
fmt.Fprint(w, strconv.Quote(e.String()))
}
type SortDirectionEnum string
const (
@ -222,44 +286,3 @@ func (e *SortDirectionEnum) UnmarshalGQL(v interface{}) error {
func (e SortDirectionEnum) MarshalGQL(w io.Writer) {
fmt.Fprint(w, strconv.Quote(e.String()))
}
type Weights string
const (
WeightsNodeCount Weights = "NODE_COUNT"
WeightsNodeHours Weights = "NODE_HOURS"
)
var AllWeights = []Weights{
WeightsNodeCount,
WeightsNodeHours,
}
func (e Weights) IsValid() bool {
switch e {
case WeightsNodeCount, WeightsNodeHours:
return true
}
return false
}
func (e Weights) String() string {
return string(e)
}
func (e *Weights) UnmarshalGQL(v interface{}) error {
str, ok := v.(string)
if !ok {
return fmt.Errorf("enums must be strings")
}
*e = Weights(str)
if !e.IsValid() {
return fmt.Errorf("%s is not a valid Weights", str)
}
return nil
}
func (e Weights) MarshalGQL(w io.Writer) {
fmt.Fprint(w, strconv.Quote(e.String()))
}

34
internal/graph/schema.resolvers.go
View File

@ -244,34 +244,34 @@ func (r *queryResolver) Jobs(ctx context.Context, filter []*model.JobFilter, pag
}
// JobsStatistics is the resolver for the jobsStatistics field.
func (r *queryResolver) JobsStatistics(ctx context.Context, filter []*model.JobFilter, groupBy *model.Aggregate) ([]*model.JobsStatistics, error) {
func (r *queryResolver) JobsStatistics(ctx context.Context, filter []*model.JobFilter, page *model.PageRequest, sortBy *model.SortByAggregate, groupBy *model.Aggregate) ([]*model.JobsStatistics, error) {
var err error
var stats []*model.JobsStatistics
if requireField(ctx, "totalJobs") {
if requireField(ctx, "totalJobs") || requireField(ctx, "totalWalltime") || requireField(ctx, "totalNodes") || requireField(ctx, "totalCores") ||
requireField(ctx, "totalAccs") || requireField(ctx, "totalNodeHours") || requireField(ctx, "totalCoreHours") || requireField(ctx, "totalAccHours") {
if groupBy == nil {
stats, err = r.Repo.JobsStats(ctx, filter)
} else {
stats, err = r.Repo.JobsStatsGrouped(ctx, filter, groupBy)
stats, err = r.Repo.JobsStatsGrouped(ctx, filter, page, sortBy, groupBy)
}
} else {
stats = make([]*model.JobsStatistics, 0, 1)
stats = append(stats,
&model.JobsStatistics{})
stats = append(stats, &model.JobsStatistics{})
}
if groupBy != nil {
if requireField(ctx, "shortJobs") {
stats, err = r.Repo.AddJobCountGrouped(ctx, filter, groupBy, stats, "short")
}
if requireField(ctx, "RunningJobs") {
if requireField(ctx, "runningJobs") {
stats, err = r.Repo.AddJobCountGrouped(ctx, filter, groupBy, stats, "running")
}
} else {
if requireField(ctx, "shortJobs") {
stats, err = r.Repo.AddJobCount(ctx, filter, stats, "short")
}
if requireField(ctx, "RunningJobs") {
if requireField(ctx, "runningJobs") {
stats, err = r.Repo.AddJobCount(ctx, filter, stats, "running")
}
}
@ -280,7 +280,7 @@ func (r *queryResolver) JobsStatistics(ctx context.Context, filter []*model.JobF
return nil, err
}
if requireField(ctx, "histDuration") || requireField(ctx, "histNumNodes") {
if requireField(ctx, "histDuration") || requireField(ctx, "histNumNodes") || requireField(ctx, "histNumCores") || requireField(ctx, "histNumAccs") {
if groupBy == nil {
stats[0], err = r.Repo.AddHistograms(ctx, filter, stats[0])
if err != nil {
@ -294,24 +294,6 @@ func (r *queryResolver) JobsStatistics(ctx context.Context, filter []*model.JobF
return stats, nil
}
// JobsCount is the resolver for the jobsCount field.
func (r *queryResolver) JobsCount(ctx context.Context, filter []*model.JobFilter, groupBy model.Aggregate, weight *model.Weights, limit *int) ([]*model.Count, error) {
counts, err := r.Repo.CountGroupedJobs(ctx, groupBy, filter, weight, limit)
if err != nil {
log.Warn("Error while counting grouped jobs")
return nil, err
}
res := make([]*model.Count, 0, len(counts))
for name, count := range counts {
res = append(res, &model.Count{
Name: name,
Count: count,
})
}
return res, nil
}
// RooflineHeatmap is the resolver for the rooflineHeatmap field.
func (r *queryResolver) RooflineHeatmap(ctx context.Context, filter []*model.JobFilter, rows int, cols int, minX float64, minY float64, maxX float64, maxY float64) ([][]float64, error) {
return r.rooflineHeatmap(ctx, filter, rows, cols, minX, minY, maxX, maxY)

47
internal/graph/util.go
View File

@ -15,6 +15,7 @@ import (
"github.com/ClusterCockpit/cc-backend/internal/metricdata"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
// "github.com/ClusterCockpit/cc-backend/pkg/archive"
)
const MAX_JOBS_FOR_ANALYSIS = 500
@ -106,7 +107,11 @@ func (r *queryResolver) jobsFootprints(ctx context.Context, filter []*model.JobF
avgs[i] = make([]schema.Float, 0, len(jobs))
}
nodehours := make([]schema.Float, 0, len(jobs))
timeweights := new(model.TimeWeights)
timeweights.NodeHours = make([]schema.Float, 0, len(jobs))
timeweights.AccHours = make([]schema.Float, 0, len(jobs))
timeweights.CoreHours = make([]schema.Float, 0, len(jobs))
for _, job := range jobs {
if job.MonitoringStatus == schema.MonitoringStatusDisabled || job.MonitoringStatus == schema.MonitoringStatusArchivingFailed {
continue
@ -117,7 +122,18 @@ func (r *queryResolver) jobsFootprints(ctx context.Context, filter []*model.JobF
return nil, err
}
nodehours = append(nodehours, schema.Float(float64(job.Duration)/60.0*float64(job.NumNodes)))
// #166 collect arrays: Null values or no null values?
timeweights.NodeHours = append(timeweights.NodeHours, schema.Float(float64(job.Duration)/60.0*float64(job.NumNodes)))
if job.NumAcc > 0 {
timeweights.AccHours = append(timeweights.AccHours, schema.Float(float64(job.Duration)/60.0*float64(job.NumAcc)))
} else {
timeweights.AccHours = append(timeweights.AccHours, schema.Float(1.0))
}
if job.NumHWThreads > 0 {
timeweights.CoreHours = append(timeweights.CoreHours, schema.Float(float64(job.Duration)/60.0*float64(job.NumHWThreads))) // SQLite HWThreads == Cores; numCoresForJob(job)
} else {
timeweights.CoreHours = append(timeweights.CoreHours, schema.Float(1.0))
}
}
res := make([]*model.MetricFootprints, len(avgs))
@ -129,11 +145,34 @@ func (r *queryResolver) jobsFootprints(ctx context.Context, filter []*model.JobF
}
return &model.Footprints{
Nodehours: nodehours,
Metrics: res,
TimeWeights: timeweights,
Metrics: res,
}, nil
}
// func numCoresForJob(job *schema.Job) (numCores int) {
// subcluster, scerr := archive.GetSubCluster(job.Cluster, job.SubCluster)
// if scerr != nil {
// return 1
// }
// totalJobCores := 0
// topology := subcluster.Topology
// for _, host := range job.Resources {
// hwthreads := host.HWThreads
// if hwthreads == nil {
// hwthreads = topology.Node
// }
// hostCores, _ := topology.GetCoresFromHWThreads(hwthreads)
// totalJobCores += len(hostCores)
// }
// return totalJobCores
// }
func requireField(ctx context.Context, name string) bool {
fields := graphql.CollectAllFields(ctx)

2
internal/metricdata/cc-metric-store.go
View File

@ -506,7 +506,7 @@ func (ccms *CCMetricStore) LoadStats(
metrics []string,
ctx context.Context) (map[string]map[string]schema.MetricStatistics, error) {
queries, _, err := ccms.buildQueries(job, metrics, []schema.MetricScope{schema.MetricScopeNode})
queries, _, err := ccms.buildQueries(job, metrics, []schema.MetricScope{schema.MetricScopeNode}) // #166 Add scope shere for analysis view accelerator normalization?
if err != nil {
log.Warn("Error while building query")
return nil, err

4
internal/metricdata/metricdata.go
View File

@ -182,7 +182,7 @@ func LoadAverages(
ctx context.Context) error {
if job.State != schema.JobStateRunning && useArchive {
return archive.LoadAveragesFromArchive(job, metrics, data)
return archive.LoadAveragesFromArchive(job, metrics, data) // #166 change also here?
}
repo, ok := metricDataRepos[job.Cluster]
@ -190,7 +190,7 @@ func LoadAverages(
return fmt.Errorf("METRICDATA/METRICDATA > no metric data repository configured for '%s'", job.Cluster)
}
stats, err := repo.LoadStats(job, metrics, ctx)
stats, err := repo.LoadStats(job, metrics, ctx) // #166 how to handle stats for acc normalizazion?
if err != nil {
log.Errorf("Error while loading statistics for job %v (User %v, Project %v)", job.JobID, job.User, job.Project)
return err

63
internal/repository/job.go
View File

@ -455,69 +455,6 @@ func (r *JobRepository) DeleteJobById(id int64) error {
return err
}
// TODO: Use node hours instead: SELECT job.user, sum(job.num_nodes * (CASE WHEN job.job_state = "running" THEN CAST(strftime('%s', 'now') AS INTEGER) - job.start_time ELSE job.duration END)) as x FROM job GROUP BY user ORDER BY x DESC;
func (r *JobRepository) CountGroupedJobs(
ctx context.Context,
aggreg model.Aggregate,
filters []*model.JobFilter,
weight *model.Weights,
limit *int) (map[string]int, error) {
start := time.Now()
if !aggreg.IsValid() {
return nil, errors.New("invalid aggregate")
}
runner := (sq.BaseRunner)(r.stmtCache)
count := "count(*) as count"
if weight != nil {
switch *weight {
case model.WeightsNodeCount:
count = "sum(job.num_nodes) as count"
case model.WeightsNodeHours:
now := time.Now().Unix()
count = fmt.Sprintf(`sum(job.num_nodes * (CASE WHEN job.job_state = "running" THEN %d - job.start_time ELSE job.duration END)) as count`, now)
runner = r.DB
default:
log.Debugf("CountGroupedJobs() Weight %v unknown.", *weight)
}
}
q, qerr := SecurityCheck(ctx, sq.Select("job."+string(aggreg), count).From("job").GroupBy("job."+string(aggreg)).OrderBy("count DESC"))
if qerr != nil {
return nil, qerr
}
for _, f := range filters {
q = BuildWhereClause(f, q)
}
if limit != nil {
q = q.Limit(uint64(*limit))
}
counts := map[string]int{}
rows, err := q.RunWith(runner).Query()
if err != nil {
log.Error("Error while running query")
return nil, err
}
for rows.Next() {
var group string
var count int
if err := rows.Scan(&group, &count); err != nil {
log.Warn("Error while scanning rows")
return nil, err
}
counts[group] = count
}
log.Debugf("Timer CountGroupedJobs %s", time.Since(start))
return counts, nil
}
func (r *JobRepository) UpdateMonitoringStatus(job int64, monitoringStatus int32) (err error) {
stmt := sq.Update("job").
Set("monitoring_status", monitoringStatus).

62
internal/repository/query.go
View File

@ -18,13 +18,17 @@ import (
sq "github.com/Masterminds/squirrel"
)
// SecurityCheck-less, private: Returns a list of jobs matching the provided filters. page and order are optional-
func (r *JobRepository) queryJobs(
query sq.SelectBuilder,
func (r *JobRepository) QueryJobs(
ctx context.Context,
filters []*model.JobFilter,
page *model.PageRequest,
order *model.OrderByInput) ([]*schema.Job, error) {
query, qerr := SecurityCheck(ctx, sq.Select(jobColumns...).From("job"))
if qerr != nil {
return nil, qerr
}
if order != nil {
field := toSnakeCase(order.Field)
@ -67,34 +71,15 @@ func (r *JobRepository) queryJobs(
return jobs, nil
}
// testFunction for queryJobs
func (r *JobRepository) testQueryJobs(
filters []*model.JobFilter,
page *model.PageRequest,
order *model.OrderByInput) ([]*schema.Job, error) {
return r.queryJobs(sq.Select(jobColumns...).From("job"), filters, page, order)
}
// Public function with added securityCheck, calls private queryJobs function above
func (r *JobRepository) QueryJobs(
func (r *JobRepository) CountJobs(
ctx context.Context,
filters []*model.JobFilter,
page *model.PageRequest,
order *model.OrderByInput) ([]*schema.Job, error) {
query, qerr := SecurityCheck(ctx, sq.Select(jobColumns...).From("job"))
if qerr != nil {
return nil, qerr
}
return r.queryJobs(query, filters, page, order)
}
// SecurityCheck-less, private: Returns the number of jobs matching the filters
func (r *JobRepository) countJobs(query sq.SelectBuilder,
filters []*model.JobFilter) (int, error) {
query, qerr := SecurityCheck(ctx, sq.Select("count(*)").From("job"))
if qerr != nil {
return 0, qerr
}
for _, f := range filters {
query = BuildWhereClause(f, query)
}
@ -107,27 +92,6 @@ func (r *JobRepository) countJobs(query sq.SelectBuilder,
return count, nil
}
// testFunction for countJobs
func (r *JobRepository) testCountJobs(
filters []*model.JobFilter) (int, error) {
return r.countJobs(sq.Select("count(*)").From("job"), filters)
}
// Public function with added securityCheck, calls private countJobs function above
func (r *JobRepository) CountJobs(
ctx context.Context,
filters []*model.JobFilter) (int, error) {
query, qerr := SecurityCheck(ctx, sq.Select("count(*)").From("job"))
if qerr != nil {
return 0, qerr
}
return r.countJobs(query, filters)
}
func SecurityCheck(ctx context.Context, query sq.SelectBuilder) (sq.SelectBuilder, error) {
user := GetUserFromContext(ctx)
if user == nil {

25
internal/repository/repository_test.go
View File

@ -5,10 +5,12 @@
package repository
import (
"context"
"testing"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
_ "github.com/mattn/go-sqlite3"
)
@ -94,7 +96,7 @@ func BenchmarkDB_CountJobs(b *testing.B) {
b.RunParallel(func(pb *testing.PB) {
for pb.Next() {
_, err := db.testCountJobs([]*model.JobFilter{filter})
_, err := db.CountJobs(getContext(b), []*model.JobFilter{filter})
noErr(b, err)
}
})
@ -118,20 +120,37 @@ func BenchmarkDB_QueryJobs(b *testing.B) {
b.RunParallel(func(pb *testing.PB) {
for pb.Next() {
_, err := db.testQueryJobs([]*model.JobFilter{filter}, page, order)
_, err := db.QueryJobs(getContext(b), []*model.JobFilter{filter}, page, order)
noErr(b, err)
}
})
})
}
func getContext(tb testing.TB) context.Context {
tb.Helper()
var roles []string
roles = append(roles, schema.GetRoleString(schema.RoleAdmin))
projects := make([]string, 0)
user := &schema.User{
Username: "demo",
Name: "The man",
Roles: roles,
Projects: projects,
AuthSource: schema.AuthViaLDAP,
}
ctx := context.Background()
return context.WithValue(ctx, ContextUserKey, user)
}
func setup(tb testing.TB) *JobRepository {
tb.Helper()
log.Init("warn", true)
dbfile := "testdata/job.db"
err := MigrateDB("sqlite3", dbfile)
noErr(tb, err)
Connect("sqlite3", dbfile)
return GetJobRepository()
}

115
internal/repository/stats.go
View File

@ -23,6 +23,17 @@ var groupBy2column = map[model.Aggregate]string{
model.AggregateCluster: "job.cluster",
}
var sortBy2column = map[model.SortByAggregate]string{
model.SortByAggregateTotaljobs: "totalJobs",
model.SortByAggregateTotalwalltime: "totalWalltime",
model.SortByAggregateTotalnodes: "totalNodes",
model.SortByAggregateTotalnodehours: "totalNodeHours",
model.SortByAggregateTotalcores: "totalCores",
model.SortByAggregateTotalcorehours: "totalCoreHours",
model.SortByAggregateTotalaccs: "totalAccs",
model.SortByAggregateTotalacchours: "totalAccHours",
}
func (r *JobRepository) buildCountQuery(
filter []*model.JobFilter,
kind string,
@ -60,19 +71,26 @@ func (r *JobRepository) buildStatsQuery(
castType := r.getCastType()
if col != "" {
// Scan columns: id, totalJobs, totalWalltime, totalNodeHours, totalCoreHours, totalAccHours
query = sq.Select(col, "COUNT(job.id)",
fmt.Sprintf("CAST(ROUND(SUM(job.duration) / 3600) as %s)", castType),
fmt.Sprintf("CAST(ROUND(SUM(job.duration * job.num_nodes) / 3600) as %s)", castType),
fmt.Sprintf("CAST(ROUND(SUM(job.duration * job.num_hwthreads) / 3600) as %s)", castType),
fmt.Sprintf("CAST(ROUND(SUM(job.duration * job.num_acc) / 3600) as %s)", castType),
// Scan columns: id, totalJobs, totalWalltime, totalNodes, totalNodeHours, totalCores, totalCoreHours, totalAccs, totalAccHours
query = sq.Select(col, "COUNT(job.id) as totalJobs",
fmt.Sprintf("CAST(ROUND(SUM(job.duration) / 3600) as %s) as totalWalltime", castType),
fmt.Sprintf("CAST(SUM(job.num_nodes) as %s) as totalNodes", castType),
fmt.Sprintf("CAST(ROUND(SUM(job.duration * job.num_nodes) / 3600) as %s) as totalNodeHours", castType),
fmt.Sprintf("CAST(SUM(job.num_hwthreads) as %s) as totalCores", castType),
fmt.Sprintf("CAST(ROUND(SUM(job.duration * job.num_hwthreads) / 3600) as %s) as totalCoreHours", castType),
fmt.Sprintf("CAST(SUM(job.num_acc) as %s) as totalAccs", castType),
fmt.Sprintf("CAST(ROUND(SUM(job.duration * job.num_acc) / 3600) as %s) as totalAccHours", castType),
).From("job").GroupBy(col)
} else {
// Scan columns: totalJobs, totalWalltime, totalNodeHours, totalCoreHours, totalAccHours
// Scan columns: totalJobs, totalWalltime, totalNodes, totalNodeHours, totalCores, totalCoreHours, totalAccs, totalAccHours
query = sq.Select("COUNT(job.id)",
fmt.Sprintf("CAST(ROUND(SUM(job.duration) / 3600) as %s)", castType),
fmt.Sprintf("CAST(SUM(job.num_nodes) as %s)", castType),
fmt.Sprintf("CAST(ROUND(SUM(job.duration * job.num_nodes) / 3600) as %s)", castType),
fmt.Sprintf("CAST(SUM(job.num_hwthreads) as %s)", castType),
fmt.Sprintf("CAST(ROUND(SUM(job.duration * job.num_hwthreads) / 3600) as %s)", castType),
fmt.Sprintf("CAST(SUM(job.num_acc) as %s)", castType),
fmt.Sprintf("CAST(ROUND(SUM(job.duration * job.num_acc) / 3600) as %s)", castType),
).From("job")
}
@ -112,16 +130,28 @@ func (r *JobRepository) getCastType() string {
func (r *JobRepository) JobsStatsGrouped(
ctx context.Context,
filter []*model.JobFilter,
page *model.PageRequest,
sortBy *model.SortByAggregate,
groupBy *model.Aggregate) ([]*model.JobsStatistics, error) {
start := time.Now()
col := groupBy2column[*groupBy]
query := r.buildStatsQuery(filter, col)
query, err := SecurityCheck(ctx, query)
if err != nil {
return nil, err
}
if sortBy != nil {
sortBy := sortBy2column[*sortBy]
query = query.OrderBy(fmt.Sprintf("%s DESC", sortBy))
}
if page != nil && page.ItemsPerPage != -1 {
limit := uint64(page.ItemsPerPage)
query = query.Offset((uint64(page.Page) - 1) * limit).Limit(limit)
}
rows, err := query.RunWith(r.DB).Query()
if err != nil {
log.Warn("Error while querying DB for job statistics")
@ -132,15 +162,36 @@ func (r *JobRepository) JobsStatsGrouped(
for rows.Next() {
var id sql.NullString
var jobs, walltime, nodeHours, coreHours, accHours sql.NullInt64
if err := rows.Scan(&id, &jobs, &walltime, &nodeHours, &coreHours, &accHours); err != nil {
var jobs, walltime, nodes, nodeHours, cores, coreHours, accs, accHours sql.NullInt64
if err := rows.Scan(&id, &jobs, &walltime, &nodes, &nodeHours, &cores, &coreHours, &accs, &accHours); err != nil {
log.Warn("Error while scanning rows")
return nil, err
}
if id.Valid {
var totalCoreHours, totalAccHours int
var totalJobs, totalWalltime, totalNodes, totalNodeHours, totalCores, totalCoreHours, totalAccs, totalAccHours int
if jobs.Valid {
totalJobs = int(jobs.Int64)
}
if walltime.Valid {
totalWalltime = int(walltime.Int64)
}
if nodes.Valid {
totalNodes = int(nodes.Int64)
}
if cores.Valid {
totalCores = int(cores.Int64)
}
if accs.Valid {
totalAccs = int(accs.Int64)
}
if nodeHours.Valid {
totalNodeHours = int(nodeHours.Int64)
}
if coreHours.Valid {
totalCoreHours = int(coreHours.Int64)
}
@ -154,9 +205,13 @@ func (r *JobRepository) JobsStatsGrouped(
&model.JobsStatistics{
ID: id.String,
Name: name,
TotalJobs: int(jobs.Int64),
TotalWalltime: int(walltime.Int64),
TotalJobs: totalJobs,
TotalWalltime: totalWalltime,
TotalNodes: totalNodes,
TotalNodeHours: totalNodeHours,
TotalCores: totalCores,
TotalCoreHours: totalCoreHours,
TotalAccs: totalAccs,
TotalAccHours: totalAccHours})
} else {
stats = append(stats,
@ -164,7 +219,11 @@ func (r *JobRepository) JobsStatsGrouped(
ID: id.String,
TotalJobs: int(jobs.Int64),
TotalWalltime: int(walltime.Int64),
TotalNodes: totalNodes,
TotalNodeHours: totalNodeHours,
TotalCores: totalCores,
TotalCoreHours: totalCoreHours,
TotalAccs: totalAccs,
TotalAccHours: totalAccHours})
}
}
@ -188,15 +247,18 @@ func (r *JobRepository) JobsStats(
row := query.RunWith(r.DB).QueryRow()
stats := make([]*model.JobsStatistics, 0, 1)
var jobs, walltime, nodeHours, coreHours, accHours sql.NullInt64
if err := row.Scan(&jobs, &walltime, &nodeHours, &coreHours, &accHours); err != nil {
var jobs, walltime, nodes, nodeHours, cores, coreHours, accs, accHours sql.NullInt64
if err := row.Scan(&jobs, &walltime, &nodes, &nodeHours, &cores, &coreHours, &accs, &accHours); err != nil {
log.Warn("Error while scanning rows")
return nil, err
}
if jobs.Valid {
var totalCoreHours, totalAccHours int
var totalNodeHours, totalCoreHours, totalAccHours int
if nodeHours.Valid {
totalNodeHours = int(nodeHours.Int64)
}
if coreHours.Valid {
totalCoreHours = int(coreHours.Int64)
}
@ -207,6 +269,7 @@ func (r *JobRepository) JobsStats(
&model.JobsStatistics{
TotalJobs: int(jobs.Int64),
TotalWalltime: int(walltime.Int64),
TotalNodeHours: totalNodeHours,
TotalCoreHours: totalCoreHours,
TotalAccHours: totalAccHours})
}
@ -321,7 +384,7 @@ func (r *JobRepository) AddJobCount(
return nil, err
}
counts := make(map[string]int)
var count int
for rows.Next() {
var cnt sql.NullInt64
@ -329,20 +392,22 @@ func (r *JobRepository) AddJobCount(
log.Warn("Error while scanning rows")
return nil, err
}
count = int(cnt.Int64)
}
switch kind {
case "running":
for _, s := range stats {
s.RunningJobs = counts[s.ID]
s.RunningJobs = count
}
case "short":
for _, s := range stats {
s.ShortJobs = counts[s.ID]
s.ShortJobs = count
}
}
log.Debugf("Timer JobJobCount %s", time.Since(start))
log.Debugf("Timer AddJobCount %s", time.Since(start))
return stats, nil
}
@ -367,6 +432,18 @@ func (r *JobRepository) AddHistograms(
return nil, err
}
stat.HistNumCores, err = r.jobsStatisticsHistogram(ctx, "job.num_hwthreads as value", filter)
if err != nil {
log.Warn("Error while loading job statistics histogram: num hwthreads")
return nil, err
}
stat.HistNumAccs, err = r.jobsStatisticsHistogram(ctx, "job.num_acc as value", filter)
if err != nil {
log.Warn("Error while loading job statistics histogram: num acc")
return nil, err
}
log.Debugf("Timer AddHistograms %s", time.Since(start))
return stat, nil
}

14
internal/repository/stats_test.go
View File

@ -7,6 +7,8 @@ package repository
import (
"fmt"
"testing"
"github.com/ClusterCockpit/cc-backend/internal/graph/model"
)
func TestBuildJobStatsQuery(t *testing.T) {
@ -19,3 +21,15 @@ func TestBuildJobStatsQuery(t *testing.T) {
fmt.Printf("SQL: %s\n", sql)
}
func TestJobStats(t *testing.T) {
r := setup(t)
filter := &model.JobFilter{}
stats, err := r.JobsStats(getContext(t), []*model.JobFilter{filter})
noErr(t, err)
if stats[0].TotalJobs != 6 {
t.Fatalf("Want 98, Got %d", stats[0].TotalJobs)
}
}

8
internal/repository/user.go
View File

@ -134,8 +134,12 @@ func (r *UserRepository) AddUser(user *schema.User) error {
func (r *UserRepository) DelUser(username string) error {
_, err := r.DB.Exec(`DELETE FROM user WHERE user.username = ?`, username)
log.Errorf("Error while deleting user '%s' from DB", username)
return err
if err != nil {
log.Errorf("Error while deleting user '%s' from DB", username)
return err
}
log.Infof("deleted user '%s' from DB", username)
return nil
}
func (r *UserRepository) ListUsers(specialsOnly bool) ([]*schema.User, error) {

1
pkg/schema/config.go
View File

@ -15,6 +15,7 @@ type LdapConfig struct {
SearchDN string `json:"search_dn"`
UserBind string `json:"user_bind"`
UserFilter string `json:"user_filter"`
UserAttr string `json:"username_attr"`
SyncInterval string `json:"sync_interval"` // Parsed using time.ParseDuration.
SyncDelOldUsers bool `json:"sync_del_old_users"`

4
pkg/schema/schemas/config.schema.json
View File

@ -180,6 +180,10 @@
"description": "Filter to extract users for syncing.",
"type": "string"
},
"username_attr": {
"description": "Attribute with full username. Default: gecos",
"type": "string"
},
"sync_interval": {
"description": "Interval used for syncing local user table with LDAP directory. Parsed using time.ParseDuration.",
"type": "string"

195
web/frontend/src/Analysis.root.svelte
View File

@ -1,7 +1,7 @@
<script>
import { init, convert2uplot } from './utils.js'
import { getContext, onMount } from 'svelte'
import { queryStore, gql, getContextClient } from '@urql/svelte'
import { queryStore, gql, getContextClient, mutationStore } from '@urql/svelte'
import { Row, Col, Spinner, Card, Table, Icon } from 'sveltestrap'
import Filters from './filters/Filters.svelte'
import PlotSelection from './PlotSelection.svelte'
@ -42,6 +42,20 @@
$: metrics = [...new Set([...metricsInHistograms, ...metricsInScatterplots.flat()])]
const sortOptions = [
{key: 'totalWalltime', label: 'Walltime'},
{key: 'totalNodeHours', label: 'Node Hours'},
{key: 'totalCoreHours', label: 'Core Hours'},
{key: 'totalAccHours', label: 'Accelerator Hours'}
]
const groupOptions = [
{key: 'user', label: 'User Name'},
{key: 'project', label: 'Project ID'}
]
let sortSelection = sortOptions.find((option) => option.key == ccconfig[`analysis_view_selectedTopCategory:${filterPresets.cluster}`]) || sortOptions.find((option) => option.key == ccconfig.analysis_view_selectedTopCategory)
let groupSelection = groupOptions.find((option) => option.key == ccconfig[`analysis_view_selectedTopEntity:${filterPresets.cluster}`]) || groupOptions.find((option) => option.key == ccconfig.analysis_view_selectedTopEntity)
getContext('on-init')(({ data }) => {
if (data != null) {
cluster = data.clusters.find(c => c.name == filterPresets.cluster)
@ -62,23 +76,39 @@
totalJobs
shortJobs
totalWalltime
totalNodeHours
totalCoreHours
totalAccHours
histDuration { count, value }
histNumNodes { count, value }
histNumCores { count, value }
}
topUsers: jobsCount(filter: $jobFilters, groupBy: USER, weight: NODE_HOURS, limit: 5) { name, count }
}
`,
variables: { jobFilters }
})
$: topQuery = queryStore({
client: client,
query: gql`
query($jobFilters: [JobFilter!]!, $paging: PageRequest!, $sortBy: SortByAggregate!, $groupBy: Aggregate!) {
topList: jobsStatistics(filter: $jobFilters, page: $paging, sortBy: $sortBy, groupBy: $groupBy) {
id
totalWalltime
totalNodeHours
totalCoreHours
totalAccHours
}
}
`,
variables: { jobFilters, paging: { itemsPerPage: 10, page: 1 }, sortBy: sortSelection.key.toUpperCase(), groupBy: groupSelection.key.toUpperCase() }
})
$: footprintsQuery = queryStore({
client: client,
query: gql`
query($jobFilters: [JobFilter!]!, $metrics: [String!]!) {
footprints: jobsFootprints(filter: $jobFilters, metrics: $metrics) {
nodehours,
timeWeights { nodeHours, accHours, coreHours },
metrics { metric, data }
}
}`,
@ -97,6 +127,53 @@
variables: { jobFilters, rows: 50, cols: 50, minX: 0.01, minY: 1., maxX: 1000., maxY }
})
const updateConfigurationMutation = ({ name, value }) => {
return mutationStore({
client: client,
query: gql`
mutation ($name: String!, $value: String!) {
updateConfiguration(name: $name, value: $value)
}
`,
variables: { name, value }
});
}
function updateEntityConfiguration(select) {
if (ccconfig[`analysis_view_selectedTopEntity:${filterPresets.cluster}`] != select) {
updateConfigurationMutation({ name: `analysis_view_selectedTopEntity:${filterPresets.cluster}`, value: JSON.stringify(select) })
.subscribe(res => {
if (res.fetching === false && !res.error) {
// console.log(`analysis_view_selectedTopEntity:${filterPresets.cluster}` + ' -> Updated!')
} else if (res.fetching === false && res.error) {
throw res.error
}
})
} else {
// console.log('No Mutation Required: Entity')
}
};
function updateCategoryConfiguration(select) {
if (ccconfig[`analysis_view_selectedTopCategory:${filterPresets.cluster}`] != select) {
updateConfigurationMutation({ name: `analysis_view_selectedTopCategory:${filterPresets.cluster}`, value: JSON.stringify(select) })
.subscribe(res => {
if (res.fetching === false && !res.error) {
// console.log(`analysis_view_selectedTopCategory:${filterPresets.cluster}` + ' -> Updated!')
} else if (res.fetching === false && res.error) {
throw res.error
}
})
} else {
// console.log('No Mutation Required: Category')
}
};
$: updateEntityConfiguration(groupSelection.key)
$: updateCategoryConfiguration(sortSelection.key)
onMount(() => filterComponent.update())
</script>
@ -151,36 +228,82 @@
<th scope="col">Total Walltime</th>
<td>{$statsQuery.data.stats[0].totalWalltime}</td>
</tr>
<tr>
<th scope="col">Total Node Hours</th>
<td>{$statsQuery.data.stats[0].totalNodeHours}</td>
</tr>
<tr>
<th scope="col">Total Core Hours</th>
<td>{$statsQuery.data.stats[0].totalCoreHours}</td>
</tr>
<tr>
<th scope="col">Total Accelerator Hours</th>
<td>{$statsQuery.data.stats[0].totalAccHours}</td>
</tr>
</Table>
</Col>
<Col>
<div bind:clientWidth={colWidth1}>
<h5>Top Users</h5>
{#key $statsQuery.data.topUsers}
<Pie
size={colWidth1}
sliceLabel='Hours'
quantities={$statsQuery.data.topUsers.sort((a, b) => b.count - a.count).map((tu) => tu.count)}
entities={$statsQuery.data.topUsers.sort((a, b) => b.count - a.count).map((tu) => tu.name)}
/>
<h5>Top
<select class="p-0" bind:value={groupSelection}>
{#each groupOptions as option}
<option value={option}>
{option.key.charAt(0).toUpperCase() + option.key.slice(1)}s
</option>
{/each}
</select>
</h5>
{#key $topQuery.data}
{#if $topQuery.fetching}
<Spinner/>
{:else if $topQuery.error}
<Card body color="danger">{$topQuery.error.message}</Card>
{:else}
<Pie
size={colWidth1}
sliceLabel={sortSelection.label}
quantities={$topQuery.data.topList.map((t) => t[sortSelection.key])}
entities={$topQuery.data.topList.map((t) => t.id)}
/>
{/if}
{/key}
</div>
</Col>
<Col>
<Table>
<tr class="mb-2"><th>Legend</th><th>User Name</th><th>Node Hours</th></tr>
{#each $statsQuery.data.topUsers.sort((a, b) => b.count - a.count) as { name, count }, i}
<tr>
<td><Icon name="circle-fill" style="color: {colors[i]};"/></td>
<th scope="col"><a href="/monitoring/user/{name}?cluster={cluster.name}">{name}</a></th>
<td>{count}</td>
</tr>
{/each}
</Table>
{#key $topQuery.data}
{#if $topQuery.fetching}
<Spinner/>
{:else if $topQuery.error}
<Card body color="danger">{$topQuery.error.message}</Card>
{:else}
<Table>
<tr class="mb-2">
<th>Legend</th>
<th>{groupSelection.label}</th>
<th>
<select class="p-0" bind:value={sortSelection}>
{#each sortOptions as option}
<option value={option}>
{option.label}
</option>
{/each}
</select>
</th>
</tr>
{#each $topQuery.data.topList as te, i}
<tr>
<td><Icon name="circle-fill" style="color: {colors[i]};"/></td>
{#if groupSelection.key == 'User'}
<th scope="col"><a href="/monitoring/user/{te.id}?cluster={cluster.name}">{te.id}</a></th>
{:else}
<th scope="col"><a href="/monitoring/jobs/?cluster={cluster.name}&project={te.id}&projectMatch=eq">{te.id}</a></th>
{/if}
<td>{te[sortSelection.key]}</td>
</tr>
{/each}
</Table>
{/if}
{/key}
</Col>
</Row>
<Row cols={3} class="mb-2">
@ -217,13 +340,13 @@
</Col>
<Col>
<div bind:clientWidth={colWidth4}>
{#key $statsQuery.data.stats[0].histNumNodes}
{#key $statsQuery.data.stats[0].histNumCores}
<Histogram
width={colWidth4} height={300}
data={convert2uplot($statsQuery.data.stats[0].histNumNodes)}
title="Number of Nodes Distribution"
xlabel="Allocated Nodes"
xunit="Nodes"
data={convert2uplot($statsQuery.data.stats[0].histNumCores)}
title="Number of Cores Distribution"
xlabel="Allocated Cores"
xunit="Cores"
ylabel="Number of Jobs"
yunit="Jobs"/>
{/key}
@ -244,8 +367,9 @@
<Row>
<Col>
<Card body>
These histograms show the distribution of the averages of all jobs matching the filters. Each job/average is weighted by its node hours.
Note that some metrics could be disabled for specific subclusters as per metriConfig and thus could affect shown average values.
These histograms show the distribution of the averages of all jobs matching the filters. Each job/average is weighted by its node hours by default
(Accelerator hours for native accelerator scope metrics, coreHours for native core scope metrics).
Note that some metrics could be disabled for specific subclusters as per metricConfig and thus could affect shown average values.
</Card>
<br/>
</Col>
@ -257,7 +381,8 @@
let:width
renderFor="analysis"
items={metricsInHistograms.map(metric => ({ metric, ...binsFromFootprint(
$footprintsQuery.data.footprints.nodehours,
$footprintsQuery.data.footprints.timeWeights,
metricConfig(cluster.name, metric)?.scope,
$footprintsQuery.data.footprints.metrics.find(f => f.metric == metric).data, numBins) }))}
itemsPerRow={ccconfig.plot_view_plotsPerRow}>
@ -265,11 +390,11 @@
data={convert2uplot(item.bins)}
width={width} height={250}
title="Average Distribution of '{item.metric}'"
xlabel={`${item.metric} average [${(metricConfig(cluster.name, item.metric)?.unit?.prefix ? metricConfig(cluster.name, item.metric)?.unit?.prefix : '') +
xlabel={`${item.metric} bin maximum [${(metricConfig(cluster.name, item.metric)?.unit?.prefix ? metricConfig(cluster.name, item.metric)?.unit?.prefix : '') +
(metricConfig(cluster.name, item.metric)?.unit?.base ? metricConfig(cluster.name, item.metric)?.unit?.base : '')}]`}
xunit={`${(metricConfig(cluster.name, item.metric)?.unit?.prefix ? metricConfig(cluster.name, item.metric)?.unit?.prefix : '') +
(metricConfig(cluster.name, item.metric)?.unit?.base ? metricConfig(cluster.name, item.metric)?.unit?.base : '')}`}
ylabel="Node Hours"
ylabel="Normalized Hours"
yunit="Hours"/>
</PlotTable>
</Col>
@ -279,7 +404,7 @@
<Col>
<Card body>
Each circle represents one job. The size of a circle is proportional to its node hours. Darker circles mean multiple jobs have the same averages for the respective metrics.
Note that some metrics could be disabled for specific subclusters as per metriConfig and thus could affect shown average values.
Note that some metrics could be disabled for specific subclusters as per metricConfig and thus could affect shown average values.
</Card>
<br/>
</Col>
@ -301,7 +426,7 @@
(metricConfig(cluster.name, item.m1)?.unit?.base ? metricConfig(cluster.name, item.m1)?.unit?.base : '')}]`}
yLabel={`${item.m2} [${(metricConfig(cluster.name, item.m2)?.unit?.prefix ? metricConfig(cluster.name, item.m2)?.unit?.prefix : '') +
(metricConfig(cluster.name, item.m2)?.unit?.base ? metricConfig(cluster.name, item.m2)?.unit?.base : '')}]`}
X={item.f1} Y={item.f2} S={$footprintsQuery.data.footprints.nodehours} />
X={item.f1} Y={item.f2} S={$footprintsQuery.data.footprints.timeWeights.nodeHours} />
</PlotTable>
</Col>
</Row>

380
web/frontend/src/Status.root.svelte
View File

@ -1,33 +1,30 @@
<script>
import Refresher from "./joblist/Refresher.svelte";
import Roofline, { transformPerNodeData } from "./plots/Roofline.svelte";
import Pie, { colors } from "./plots/Pie.svelte";
import Histogram from "./plots/Histogram.svelte";
import {
Row,
Col,
Spinner,
Card,
CardHeader,
CardTitle,
CardBody,
Table,
Progress,
Icon,
} from "sveltestrap";
import { init, convert2uplot } from "./utils.js";
import { scaleNumbers } from "./units.js";
import { queryStore, gql, getContextClient } from "@urql/svelte";
import { getContext } from 'svelte'
import Refresher from './joblist/Refresher.svelte'
import Roofline, { transformPerNodeData } from './plots/Roofline.svelte'
import Pie, { colors } from './plots/Pie.svelte'
import Histogram from './plots/Histogram.svelte'
import { Row, Col, Spinner, Card, CardHeader, CardTitle, CardBody, Table, Progress, Icon } from 'sveltestrap'
import { init, convert2uplot } from './utils.js'
import { scaleNumbers } from './units.js'
import { queryStore, gql, getContextClient, mutationStore } from '@urql/svelte'
const { query: initq } = init();
const { query: initq } = init()
const ccconfig = getContext("cc-config")
export let cluster;
let plotWidths = [],
colWidth1 = 0,
colWidth2;
let from = new Date(Date.now() - 5 * 60 * 1000),
to = new Date(Date.now());
let plotWidths = [], colWidth1 = 0, colWidth2
let from = new Date(Date.now() - 5 * 60 * 1000), to = new Date(Date.now())
const topOptions = [
{key: 'totalJobs', label: 'Jobs'},
{key: 'totalNodes', label: 'Nodes'},
{key: 'totalCores', label: 'Cores'},
{key: 'totalAccs', label: 'Accelerators'},
]
let topProjectSelection = topOptions.find((option) => option.key == ccconfig[`status_view_selectedTopProjectCategory:${cluster}`]) || topOptions.find((option) => option.key == ccconfig.status_view_selectedTopProjectCategory)
let topUserSelection = topOptions.find((option) => option.key == ccconfig[`status_view_selectedTopUserCategory:${cluster}`]) || topOptions.find((option) => option.key == ccconfig.status_view_selectedTopUserCategory)
const client = getContextClient();
$: mainQuery = queryStore({
@ -64,65 +61,57 @@
}
}
stats: jobsStatistics(filter: $filter) {
histDuration {
count
value
}
histNumNodes {
count
value
}
}
stats: jobsStatistics(filter: $filter) {
histDuration { count, value }
histNumNodes { count, value }
histNumCores { count, value }
histNumAccs { count, value }
}
allocatedNodes(cluster: $cluster) {
name
count
}
topUsers: jobsCount(
filter: $filter
groupBy: USER
weight: NODE_COUNT
limit: 10
) {
name
count
}
topProjects: jobsCount(
filter: $filter
groupBy: PROJECT
weight: NODE_COUNT
limit: 10
) {
name
count
allocatedNodes(cluster: $cluster) { name, count }
}`,
variables: {
cluster: cluster, metrics: ['flops_any', 'mem_bw'], from: from.toISOString(), to: to.toISOString(),
filter: [{ state: ['running'] }, { cluster: { eq: cluster } }]
}
})
const paging = { itemsPerPage: 10, page: 1 }; // Top 10
$: topUserQuery = queryStore({
client: client,
query: gql`
query($filter: [JobFilter!]!, $paging: PageRequest!, $sortBy: SortByAggregate!) {
topUser: jobsStatistics(filter: $filter, page: $paging, sortBy: $sortBy, groupBy: USER) {
id
totalJobs
totalNodes
totalCores
totalAccs
}
}
`,
variables: {
cluster: cluster,
metrics: ["flops_any", "mem_bw"],
from: from.toISOString(),
to: to.toISOString(),
filter: [{ state: ["running"] }, { cluster: { eq: cluster } }],
},
});
variables: { filter: [{ state: ['running'] }, { cluster: { eq: cluster } }], paging, sortBy: topUserSelection.key.toUpperCase() }
})
const sumUp = (data, subcluster, metric) =>
data.reduce(
(sum, node) =>
node.subCluster == subcluster
? sum +
(node.metrics
.find((m) => m.name == metric)
?.metric.series.reduce(
(sum, series) =>
sum + series.data[series.data.length - 1],
0
) || 0)
: sum,
0
);
$: topProjectQuery = queryStore({
client: client,
query: gql`
query($filter: [JobFilter!]!, $paging: PageRequest!, $sortBy: SortByAggregate!) {
topProjects: jobsStatistics(filter: $filter, page: $paging, sortBy: $sortBy, groupBy: PROJECT) {
id
totalJobs
totalNodes
totalCores
totalAccs
}
}
`,
variables: { filter: [{ state: ['running'] }, { cluster: { eq: cluster } }], paging, sortBy: topProjectSelection.key.toUpperCase() }
})
const sumUp = (data, subcluster, metric) => data.reduce((sum, node) => node.subCluster == subcluster
? sum + (node.metrics.find(m => m.name == metric)?.metric.series.reduce((sum, series) => sum + series.data[series.data.length - 1], 0) || 0)
: sum, 0)
let allocatedNodes = {},
flopRate = {},
@ -166,6 +155,52 @@
subCluster.memoryBandwidth.unit.base;
}
}
const updateConfigurationMutation = ({ name, value }) => {
return mutationStore({
client: client,
query: gql`
mutation ($name: String!, $value: String!) {
updateConfiguration(name: $name, value: $value)
}
`,
variables: { name, value }
});
}
function updateTopUserConfiguration(select) {
if (ccconfig[`status_view_selectedTopUserCategory:${cluster}`] != select) {
updateConfigurationMutation({ name: `status_view_selectedTopUserCategory:${cluster}`, value: JSON.stringify(select) })
.subscribe(res => {
if (res.fetching === false && !res.error) {
// console.log(`status_view_selectedTopUserCategory:${cluster}` + ' -> Updated!')
} else if (res.fetching === false && res.error) {
throw res.error
}
})
} else {
// console.log('No Mutation Required: Top User')
}
}
function updateTopProjectConfiguration(select) {
if (ccconfig[`status_view_selectedTopProjectCategory:${cluster}`] != select) {
updateConfigurationMutation({ name: `status_view_selectedTopProjectCategory:${cluster}`, value: JSON.stringify(select) })
.subscribe(res => {
if (res.fetching === false && !res.error) {
// console.log(`status_view_selectedTopProjectCategory:${cluster}` + ' -> Updated!')
} else if (res.fetching === false && res.error) {
throw res.error
}
})
} else {
// console.log('No Mutation Required: Top Project')
}
};
$: updateTopUserConfiguration(topUserSelection.key)
$: updateTopProjectConfiguration(topProjectSelection.key)
</script>
<!-- Loading indicator & Refresh -->
@ -314,85 +349,103 @@
<Row>
<Col class="p-2">
<div bind:clientWidth={colWidth1}>
<h4 class="text-center">Top Users</h4>
{#key $mainQuery.data}
<Pie
size={colWidth1}
sliceLabel="Jobs"
quantities={$mainQuery.data.topUsers
.sort((a, b) => b.count - a.count)
.map((tu) => tu.count)}
entities={$mainQuery.data.topUsers
.sort((a, b) => b.count - a.count)
.map((tu) => tu.name)}
/>
<h4 class="text-center">Top Users on {cluster.charAt(0).toUpperCase() + cluster.slice(1)}</h4>
{#key $topUserQuery.data}
{#if $topUserQuery.fetching}
<Spinner/>
{:else if $topUserQuery.error}
<Card body color="danger">{$topUserQuery.error.message}</Card>
{:else}
<Pie
size={colWidth1}
sliceLabel={topUserSelection.label}
quantities={$topUserQuery.data.topUser.map((tu) => tu[topUserSelection.key])}
entities={$topUserQuery.data.topUser.map((tu) => tu.id)}
/>
{/if}
{/key}
</div>
</Col>
<Col class="px-4 py-2">
<Table>
<tr class="mb-2"
><th>Legend</th><th>User Name</th><th>Number of Nodes</th
></tr
>
{#each $mainQuery.data.topUsers.sort((a, b) => b.count - a.count) as { name, count }, i}
<tr>
<td
><Icon
name="circle-fill"
style="color: {colors[i]};"
/></td
>
<th scope="col"
><a
href="/monitoring/user/{name}?cluster={cluster}&state=running"
>{name}</a
></th
>
<td>{count}</td>
</tr>
{/each}
</Table>
{#key $topUserQuery.data}
{#if $topUserQuery.fetching}
<Spinner/>
{:else if $topUserQuery.error}
<Card body color="danger">{$topUserQuery.error.message}</Card>
{:else}
<Table>
<tr class="mb-2">
<th>Legend</th>
<th>User Name</th>
<th>Number of
<select class="p-0" bind:value={topUserSelection}>
{#each topOptions as option}
<option value={option}>
{option.label}
</option>
{/each}
</select>
</th>
</tr>
{#each $topUserQuery.data.topUser as tu, i}
<tr>
<td><Icon name="circle-fill" style="color: {colors[i]};"/></td>
<th scope="col"><a href="/monitoring/user/{tu.id}?cluster={cluster}&state=running">{tu.id}</a></th>
<td>{tu[topUserSelection.key]}</td>
</tr>
{/each}
</Table>
{/if}
{/key}
</Col>
<Col class="p-2">
<h4 class="text-center">Top Projects</h4>
{#key $mainQuery.data}
<Pie
size={colWidth1}
sliceLabel="Jobs"
quantities={$mainQuery.data.topProjects
.sort((a, b) => b.count - a.count)
.map((tp) => tp.count)}
entities={$mainQuery.data.topProjects
.sort((a, b) => b.count - a.count)
.map((tp) => tp.name)}
/>
<h4 class="text-center">Top Projects on {cluster.charAt(0).toUpperCase() + cluster.slice(1)}</h4>
{#key $topProjectQuery.data}
{#if $topProjectQuery.fetching}
<Spinner/>
{:else if $topProjectQuery.error}
<Card body color="danger">{$topProjectQuery.error.message}</Card>
{:else}
<Pie
size={colWidth1}
sliceLabel={topProjectSelection.label}
quantities={$topProjectQuery.data.topProjects.map((tp) => tp[topProjectSelection.key])}
entities={$topProjectQuery.data.topProjects.map((tp) => tp.id)}
/>
{/if}
{/key}
</Col>
<Col class="px-4 py-2">
<Table>
<tr class="mb-2"
><th>Legend</th><th>Project Code</th><th>Number of Nodes</th
></tr
>
{#each $mainQuery.data.topProjects.sort((a, b) => b.count - a.count) as { name, count }, i}
<tr>
<td
><Icon
name="circle-fill"
style="color: {colors[i]};"
/></td
>
<th scope="col"
><a
href="/monitoring/jobs/?cluster={cluster}&state=running&project={name}&projectMatch=eq"
>{name}</a
></th
>
<td>{count}</td>
</tr>
{/each}
</Table>
{#key $topProjectQuery.data}
{#if $topProjectQuery.fetching}
<Spinner/>
{:else if $topProjectQuery.error}
<Card body color="danger">{$topProjectQuery.error.message}</Card>
{:else}
<Table>
<tr class="mb-2">
<th>Legend</th>
<th>Project Code</th>
<th>Number of
<select class="p-0" bind:value={topProjectSelection}>
{#each topOptions as option}
<option value={option}>
{option.label}
</option>
{/each}
</select>
</th>
</tr>
{#each $topProjectQuery.data.topProjects as tp, i}
<tr>
<td><Icon name="circle-fill" style="color: {colors[i]};"/></td>
<th scope="col"><a href="/monitoring/jobs/?cluster={cluster}&state=running&project={tp.id}&projectMatch=eq">{tp.id}</a></th>
<td>{tp[topProjectSelection.key]}</td>
</tr>
{/each}
</Table>
{/if}
{/key}
</Col>
</Row>
<hr class="my-2" />
@ -428,5 +481,32 @@
{/key}
</Col>
</Row>
<Row cols={2}>
<Col class="p-2">
<div bind:clientWidth={colWidth2}>
{#key $mainQuery.data.stats}
<Histogram
data={convert2uplot($mainQuery.data.stats[0].histNumCores)}
width={colWidth2 - 25}
title="Number of Cores Distribution"
xlabel="Allocated Cores"
xunit="Cores"
ylabel="Number of Jobs"
yunit="Jobs"/>
{/key}
</div>
</Col>
<Col class="p-2">
{#key $mainQuery.data.stats}
<Histogram
data={convert2uplot($mainQuery.data.stats[0].histNumAccs)}
width={colWidth2 - 25}
title="Number of Accelerators Distribution"
xlabel="Allocated Accs"
xunit="Accs"
ylabel="Number of Jobs"
yunit="Jobs"/>
{/key}
</Col>
</Row>
{/if}

18
web/frontend/src/plots/MetricPlot.svelte
View File

@ -166,7 +166,7 @@
}
}
const plotSeries = [{label: 'Runtime', value: (u, ts, sidx, didx) => didx == null ? null : formatTime(ts)}]
const plotSeries = [{label: 'Runtime', value: (u, ts, sidx, didx) => didx == null ? null : formatTime(ts, forNode)}]
const plotData = [new Array(longestSeries)]
if (forNode === true) {
@ -227,7 +227,7 @@
scale: 'x',
space: 35,
incrs: timeIncrs(timestep, maxX, forNode),
values: (_, vals) => vals.map(v => formatTime(v))
values: (_, vals) => vals.map(v => formatTime(v, forNode))
},
{
scale: 'y',
@ -349,19 +349,21 @@
}
</script>
<script context="module">
export function formatTime(t) {
export function formatTime(t, forNode = false) {
if (t !== null) {
if (isNaN(t)) {
return t
} else {
let h = Math.floor(t / 3600)
let m = Math.floor((t % 3600) / 60)
const tAbs = Math.abs(t)
const h = Math.floor(tAbs / 3600)
const m = Math.floor((tAbs % 3600) / 60)
// Re-Add "negativity" to time ticks only as string, so that if-cases work as intended
if (h == 0)
return `${m}m`
return `${forNode && m != 0 ? '-' : ''}${m}m`
else if (m == 0)
return `${h}h`
return `${forNode?'-':''}${h}h`
else
return `${h}:${m}h`
return `${forNode?'-':''}${h}:${m}h`
}
}
}

4
web/frontend/src/plots/Pie.svelte
View File

@ -43,14 +43,14 @@
export let entities
export let displayLegend = false
const data = {
$: data = {
labels: entities,
datasets: [
{
label: sliceLabel,
data: quantities,
fill: 1,
backgroundColor: colors.slice(0, quantities.length),
backgroundColor: colors.slice(0, quantities.length)
}
]
}

28
web/frontend/src/utils.js
View File

@ -325,7 +325,7 @@ export function convert2uplot(canvasData) {
return uplotData
}
export function binsFromFootprint(weights, values, numBins) {
export function binsFromFootprint(weights, scope, values, numBins) {
let min = 0, max = 0
if (values.length != 0) {
for (let x of values) {
@ -338,10 +338,23 @@ export function binsFromFootprint(weights, values, numBins) {
if (numBins == null || numBins < 3)
numBins = 3
let scopeWeights
switch (scope) {
case 'core':
scopeWeights = weights.coreHours
break
case 'accelerator':
scopeWeights = weights.accHours
break
default: // every other scope: use 'node'
scopeWeights = weights.nodeHours
}
const bins = new Array(numBins).fill(0)
for (let i = 0; i < values.length; i++)
bins[Math.floor(((values[i] - min) / (max - min)) * numBins)] += weights ? weights[i] : 1
bins[Math.floor(((values[i] - min) / (max - min)) * numBins)] += scopeWeights ? scopeWeights[i] : 1
// Manual Canvas Original
// return {
// label: idx => {
// let start = min + (idx / numBins) * (max - min)
@ -355,14 +368,13 @@ export function binsFromFootprint(weights, values, numBins) {
return {
bins: bins.map((count, idx) => ({
value: idx => { // Get rounded down next integer to bins' Start-Stop Mean Value
let start = min + (idx / numBins) * (max - min)
value: idx => { // Use bins' max value instead of mean
// let start = min + (idx / numBins) * (max - min)
let stop = min + ((idx + 1) / numBins) * (max - min)
return `${formatNumber(Math.floor((start+stop)/2))}`
// return `${formatNumber(Math.floor((start+stop)/2))}`
return Math.floor(stop)
},
count: count
})),
min: min,
max: max
}))
}
}