ClusterCockpit/cc-backend
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-backend synced 2025-07-23 12:51:40 +02:00
Code Issues Packages Releases Activity

all schemas new

Browse Source
This commit is contained in:
Lou Knauer
2021-12-17 15:49:22 +01:00
parent 89333666b3
commit 5403177edc
20 changed files with 3175 additions and 2383 deletions
Download Patch File Download Diff File Expand all files Collapse all files

93
metricdata/archive.go
View File

@@ -13,13 +13,12 @@ import (
"strconv"
"github.com/ClusterCockpit/cc-jobarchive/config"
"github.com/ClusterCockpit/cc-jobarchive/graph/model"
"github.com/ClusterCockpit/cc-jobarchive/schema"
)
// For a given job, return the path of the `data.json`/`meta.json` file.
// TODO: Implement Issue ClusterCockpit/ClusterCockpit#97
func getPath(job *model.Job, file string, checkLegacy bool) (string, error) {
func getPath(job *schema.Job, file string, checkLegacy bool) (string, error) {
lvl1, lvl2 := fmt.Sprintf("%d", job.JobID/1000), fmt.Sprintf("%03d", job.JobID%1000)
if !checkLegacy {
return filepath.Join(JobArchivePath, job.Cluster, lvl1, lvl2, strconv.FormatInt(job.StartTime.Unix(), 10), file), nil
@@ -34,7 +33,7 @@ func getPath(job *model.Job, file string, checkLegacy bool) (string, error) {
}
// Assuming job is completed/archived, return the jobs metric data.
func loadFromArchive(job *model.Job) (schema.JobData, error) {
func loadFromArchive(job *schema.Job) (schema.JobData, error) {
filename, err := getPath(job, "data.json", true)
if err != nil {
return nil, err
@@ -56,8 +55,8 @@ func loadFromArchive(job *model.Job) (schema.JobData, error) {
// If the job is archived, find its `meta.json` file and override the tags list
// in that JSON file. If the job is not archived, nothing is done.
func UpdateTags(job *model.Job, tags []*model.JobTag) error {
if job.State == model.JobStateRunning {
func UpdateTags(job *schema.Job, tags []*schema.Tag) error {
if job.State == schema.JobStateRunning {
return nil
}
@@ -74,23 +73,19 @@ func UpdateTags(job *model.Job, tags []*model.JobTag) error {
return err
}
var metaFile schema.JobMeta
var metaFile schema.JobMeta = schema.JobMeta{
BaseJob: schema.JobDefaults,
}
if err := json.NewDecoder(f).Decode(&metaFile); err != nil {
return err
}
f.Close()
metaFile.Tags = make([]struct {
Name string "json:\"Name\""
Type string "json:\"Type\""
}, 0)
metaFile.Tags = make([]*schema.Tag, 0)
for _, tag := range tags {
metaFile.Tags = append(metaFile.Tags, struct {
Name string "json:\"Name\""
Type string "json:\"Type\""
}{
Name: tag.TagName,
Type: tag.TagType,
metaFile.Tags = append(metaFile.Tags, &schema.Tag{
Name: tag.Name,
Type: tag.Type,
})
}
@@ -103,7 +98,7 @@ func UpdateTags(job *model.Job, tags []*model.JobTag) error {
}
// Helper to metricdata.LoadAverages().
func loadAveragesFromArchive(job *model.Job, metrics []string, data [][]schema.Float) error {
func loadAveragesFromArchive(job *schema.Job, metrics []string, data [][]schema.Float) error {
filename, err := getPath(job, "meta.json", true)
if err != nil {
return err
@@ -131,8 +126,8 @@ func loadAveragesFromArchive(job *model.Job, metrics []string, data [][]schema.F
}
// Writes a running job to the job-archive
func ArchiveJob(job *model.Job, ctx context.Context) (*schema.JobMeta, error) {
if job.State != model.JobStateRunning {
func ArchiveJob(job *schema.Job, ctx context.Context) (*schema.JobMeta, error) {
if job.State != schema.JobStateRunning {
return nil, errors.New("cannot archive job that is not running")
}
@@ -146,51 +141,27 @@ func ArchiveJob(job *model.Job, ctx context.Context) (*schema.JobMeta, error) {
return nil, err
}
tags := []struct {
Name string `json:"Name"`
Type string `json:"Type"`
}{}
for _, tag := range job.Tags {
tags = append(tags, struct {
Name string `json:"Name"`
Type string `json:"Type"`
}{
Name: tag.TagName,
Type: tag.TagType,
})
}
metaData := &schema.JobMeta{
JobId: int64(job.JobID),
User: job.User,
Project: job.Project,
Cluster: job.Cluster,
NumNodes: job.NumNodes,
NumHWThreads: job.NumHWThreads,
NumAcc: job.NumAcc,
Exclusive: int8(job.Exclusive),
MonitoringStatus: int8(job.MonitoringStatus),
SMT: int8(job.Smt),
Partition: job.Partition,
ArrayJobId: job.ArrayJobID,
JobState: string(job.State),
StartTime: job.StartTime.Unix(),
Duration: int64(job.Duration),
Resources: job.Resources,
MetaData: "", // TODO/FIXME: Handle `meta_data`!
Tags: tags,
Statistics: make(map[string]*schema.JobMetaStatistics),
jobMeta := &schema.JobMeta{
BaseJob: job.BaseJob,
StartTime: job.StartTime.Unix(),
Statistics: make(map[string]schema.JobStatistics),
}
for metric, data := range jobData {
avg, min, max := 0.0, math.MaxFloat32, -math.MaxFloat32
for _, nodedata := range data.Series {
avg += nodedata.Statistics.Avg
min = math.Min(min, nodedata.Statistics.Min)
max = math.Max(max, nodedata.Statistics.Max)
nodeData, ok := data["node"]
if !ok {
// TODO/FIXME: Calc average for non-node metrics as well!
continue
}
metaData.Statistics[metric] = &schema.JobMetaStatistics{
for _, series := range nodeData.Series {
avg += series.Statistics.Avg
min = math.Min(min, series.Statistics.Min)
max = math.Max(max, series.Statistics.Max)
}
jobMeta.Statistics[metric] = schema.JobStatistics{
Unit: config.GetMetricConfig(job.Cluster, metric).Unit,
Avg: avg / float64(job.NumNodes),
Min: min,
@@ -202,7 +173,7 @@ func ArchiveJob(job *model.Job, ctx context.Context) (*schema.JobMeta, error) {
// only return the JobMeta structure as the
// statistics in there are needed.
if !useArchive {
return metaData, nil
return jobMeta, nil
}
dirPath, err := getPath(job, "", false)
@@ -220,7 +191,7 @@ func ArchiveJob(job *model.Job, ctx context.Context) (*schema.JobMeta, error) {
}
defer f.Close()
writer := bufio.NewWriter(f)
if err := json.NewEncoder(writer).Encode(metaData); err != nil {
if err := json.NewEncoder(writer).Encode(jobMeta); err != nil {
return nil, err
}
if err := writer.Flush(); err != nil {
@@ -239,5 +210,5 @@ func ArchiveJob(job *model.Job, ctx context.Context) (*schema.JobMeta, error) {
return nil, err
}
return metaData, f.Close()
return jobMeta, f.Close()
}

14
metricdata/cc-metric-store.go
View File

@@ -12,7 +12,6 @@ import (
"time"
"github.com/ClusterCockpit/cc-jobarchive/config"
"github.com/ClusterCockpit/cc-jobarchive/graph/model"
"github.com/ClusterCockpit/cc-jobarchive/schema"
)
@@ -57,7 +56,7 @@ func (ccms *CCMetricStore) Init(url string) error {
return nil
}
func (ccms *CCMetricStore) doRequest(job *model.Job, suffix string, metrics []string, ctx context.Context) (*http.Response, error) {
func (ccms *CCMetricStore) doRequest(job *schema.Job, suffix string, metrics []string, ctx context.Context) (*http.Response, error) {
from, to := job.StartTime.Unix(), job.StartTime.Add(time.Duration(job.Duration)*time.Second).Unix()
reqBody := ApiRequestBody{}
reqBody.Metrics = metrics
@@ -85,7 +84,7 @@ func (ccms *CCMetricStore) doRequest(job *model.Job, suffix string, metrics []st
return ccms.client.Do(req)
}
func (ccms *CCMetricStore) LoadData(job *model.Job, metrics []string, ctx context.Context) (schema.JobData, error) {
func (ccms *CCMetricStore) LoadData(job *schema.Job, metrics []string, ctx context.Context) (schema.JobData, error) {
res, err := ccms.doRequest(job, "timeseries?with-stats=true", metrics, ctx)
if err != nil {
return nil, err
@@ -103,8 +102,9 @@ func (ccms *CCMetricStore) LoadData(job *model.Job, metrics []string, ctx contex
Scope: "node", // TODO: FIXME: Whatever...
Unit: mc.Unit,
Timestep: mc.Timestep,
Series: make([]*schema.MetricSeries, 0, len(job.Resources)),
Series: make([]schema.Series, 0, len(job.Resources)),
}
for i, node := range job.Resources {
if node.Accelerators != nil || node.HWThreads != nil {
// TODO/FIXME:
@@ -120,7 +120,7 @@ func (ccms *CCMetricStore) LoadData(job *model.Job, metrics []string, ctx contex
return nil, fmt.Errorf("no data for node '%s' and metric '%s'", node.Hostname, metric)
}
metricData.Series = append(metricData.Series, &schema.MetricSeries{
metricData.Series = append(metricData.Series, schema.Series{
Hostname: node.Hostname,
Data: data.Data,
Statistics: &schema.MetricStatistics{
@@ -130,13 +130,13 @@ func (ccms *CCMetricStore) LoadData(job *model.Job, metrics []string, ctx contex
},
})
}
jobData[metric] = metricData
jobData[metric] = map[string]*schema.JobMetric{"node": metricData}
}
return jobData, nil
}
func (ccms *CCMetricStore) LoadStats(job *model.Job, metrics []string, ctx context.Context) (map[string]map[string]schema.MetricStatistics, error) {
func (ccms *CCMetricStore) LoadStats(job *schema.Job, metrics []string, ctx context.Context) (map[string]map[string]schema.MetricStatistics, error) {
res, err := ccms.doRequest(job, "stats", metrics, ctx)
if err != nil {
return nil, err

2
metricdata/influxdb-v2.go
View File

@@ -1,5 +1,6 @@
package metricdata
/*
import (
"context"
"errors"
@@ -175,3 +176,4 @@ func (idb *InfluxDBv2DataRepository) LoadStats(job *model.Job, metrics []string,
func (idb *InfluxDBv2DataRepository) LoadNodeData(clusterId string, metrics, nodes []string, from, to int64, ctx context.Context) (map[string]map[string][]schema.Float, error) {
return nil, nil
}
*/

29
metricdata/metricdata.go
View File

@@ -5,7 +5,6 @@ import (
"fmt"
"github.com/ClusterCockpit/cc-jobarchive/config"
"github.com/ClusterCockpit/cc-jobarchive/graph/model"
"github.com/ClusterCockpit/cc-jobarchive/schema"
)
@@ -15,10 +14,10 @@ type MetricDataRepository interface {
Init(url string) error
// Return the JobData for the given job, only with the requested metrics.
LoadData(job *model.Job, metrics []string, ctx context.Context) (schema.JobData, error)
LoadData(job *schema.Job, metrics []string, ctx context.Context) (schema.JobData, error)
// Return a map of metrics to a map of nodes to the metric statistics of the job.
LoadStats(job *model.Job, metrics []string, ctx context.Context) (map[string]map[string]schema.MetricStatistics, error)
LoadStats(job *schema.Job, metrics []string, ctx context.Context) (map[string]map[string]schema.MetricStatistics, error)
// Return a map of nodes to a map of metrics to the data for the requested time.
LoadNodeData(clusterId string, metrics, nodes []string, from, to int64, ctx context.Context) (map[string]map[string][]schema.Float, error)
@@ -41,15 +40,15 @@ func Init(jobArchivePath string, disableArchive bool) error {
if err := ccms.Init(cluster.MetricDataRepository.Url); err != nil {
return err
}
metricDataRepos[cluster.ClusterID] = ccms
case "influxdb-v2":
idb := &InfluxDBv2DataRepository{}
if err := idb.Init(cluster.MetricDataRepository.Url); err != nil {
return err
}
metricDataRepos[cluster.ClusterID] = idb
metricDataRepos[cluster.Name] = ccms
// case "influxdb-v2":
// idb := &InfluxDBv2DataRepository{}
// if err := idb.Init(cluster.MetricDataRepository.Url); err != nil {
// return err
// }
// metricDataRepos[cluster.Name] = idb
default:
return fmt.Errorf("unkown metric data repository '%s' for cluster '%s'", cluster.MetricDataRepository.Kind, cluster.ClusterID)
return fmt.Errorf("unkown metric data repository '%s' for cluster '%s'", cluster.MetricDataRepository.Kind, cluster.Name)
}
}
}
@@ -57,8 +56,8 @@ func Init(jobArchivePath string, disableArchive bool) error {
}
// Fetches the metric data for a job.
func LoadData(job *model.Job, metrics []string, ctx context.Context) (schema.JobData, error) {
if job.State == model.JobStateRunning || !useArchive {
func LoadData(job *schema.Job, metrics []string, ctx context.Context) (schema.JobData, error) {
if job.State == schema.JobStateRunning || !useArchive {
repo, ok := metricDataRepos[job.Cluster]
if !ok {
return nil, fmt.Errorf("no metric data repository configured for '%s'", job.Cluster)
@@ -85,8 +84,8 @@ func LoadData(job *model.Job, metrics []string, ctx context.Context) (schema.Job
}
// Used for the jobsFootprint GraphQL-Query. TODO: Rename/Generalize.
func LoadAverages(job *model.Job, metrics []string, data [][]schema.Float, ctx context.Context) error {
if job.State != model.JobStateRunning && useArchive {
func LoadAverages(job *schema.Job, metrics []string, data [][]schema.Float, ctx context.Context) error {
if job.State != schema.JobStateRunning && useArchive {
return loadAveragesFromArchive(job, metrics, data)
}