2024-04-11 23:04:30 +02:00
|
|
|
// Copyright (C) NHR@FAU, University Erlangen-Nuremberg.
|
2023-04-28 08:49:58 +02:00
|
|
|
// All rights reserved.
|
|
|
|
// Use of this source code is governed by a MIT-style
|
|
|
|
// license that can be found in the LICENSE file.
|
|
|
|
package importer
|
|
|
|
|
|
|
|
import (
|
|
|
|
"encoding/json"
|
|
|
|
"fmt"
|
|
|
|
"strings"
|
|
|
|
"time"
|
|
|
|
|
|
|
|
"github.com/ClusterCockpit/cc-backend/internal/repository"
|
|
|
|
"github.com/ClusterCockpit/cc-backend/pkg/archive"
|
|
|
|
"github.com/ClusterCockpit/cc-backend/pkg/log"
|
|
|
|
"github.com/ClusterCockpit/cc-backend/pkg/schema"
|
|
|
|
)
|
|
|
|
|
|
|
|
// Delete the tables "job", "tag" and "jobtag" from the database and
|
|
|
|
// repopulate them using the jobs found in `archive`.
|
|
|
|
func InitDB() error {
|
|
|
|
r := repository.GetJobRepository()
|
2023-05-04 16:03:04 +02:00
|
|
|
if err := r.Flush(); err != nil {
|
|
|
|
log.Errorf("repository initDB(): %v", err)
|
|
|
|
return err
|
|
|
|
}
|
2023-04-28 08:49:58 +02:00
|
|
|
starttime := time.Now()
|
|
|
|
log.Print("Building job table...")
|
|
|
|
|
|
|
|
t, err := r.TransactionInit()
|
|
|
|
if err != nil {
|
|
|
|
log.Warn("Error while initializing SQL transactions")
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
tags := make(map[string]int64)
|
|
|
|
|
|
|
|
// Not using log.Print because we want the line to end with `\r` and
|
|
|
|
// this function is only ever called when a special command line flag
|
|
|
|
// is passed anyways.
|
|
|
|
fmt.Printf("%d jobs inserted...\r", 0)
|
|
|
|
|
|
|
|
ar := archive.GetHandle()
|
|
|
|
i := 0
|
|
|
|
errorOccured := 0
|
|
|
|
|
|
|
|
for jobContainer := range ar.Iter(false) {
|
|
|
|
|
|
|
|
jobMeta := jobContainer.Meta
|
|
|
|
|
|
|
|
// Bundle 100 inserts into one transaction for better performance
|
|
|
|
if i%100 == 0 {
|
|
|
|
r.TransactionCommit(t)
|
|
|
|
fmt.Printf("%d jobs inserted...\r", i)
|
|
|
|
}
|
|
|
|
|
|
|
|
jobMeta.MonitoringStatus = schema.MonitoringStatusArchivingSuccessful
|
|
|
|
job := schema.Job{
|
|
|
|
BaseJob: jobMeta.BaseJob,
|
|
|
|
StartTime: time.Unix(jobMeta.StartTime, 0),
|
|
|
|
StartTimeUnix: jobMeta.StartTime,
|
|
|
|
}
|
|
|
|
|
|
|
|
// TODO: Other metrics...
|
2023-06-15 11:07:48 +02:00
|
|
|
job.LoadAvg = loadJobStat(jobMeta, "cpu_load")
|
2023-04-28 08:49:58 +02:00
|
|
|
job.FlopsAnyAvg = loadJobStat(jobMeta, "flops_any")
|
2023-06-15 11:07:48 +02:00
|
|
|
job.MemUsedMax = loadJobStat(jobMeta, "mem_used")
|
2023-04-28 08:49:58 +02:00
|
|
|
job.MemBwAvg = loadJobStat(jobMeta, "mem_bw")
|
|
|
|
job.NetBwAvg = loadJobStat(jobMeta, "net_bw")
|
|
|
|
job.FileBwAvg = loadJobStat(jobMeta, "file_bw")
|
|
|
|
|
|
|
|
job.RawResources, err = json.Marshal(job.Resources)
|
|
|
|
if err != nil {
|
|
|
|
log.Errorf("repository initDB(): %v", err)
|
|
|
|
errorOccured++
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
job.RawMetaData, err = json.Marshal(job.MetaData)
|
|
|
|
if err != nil {
|
|
|
|
log.Errorf("repository initDB(): %v", err)
|
|
|
|
errorOccured++
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
if err := SanityChecks(&job.BaseJob); err != nil {
|
|
|
|
log.Errorf("repository initDB(): %v", err)
|
|
|
|
errorOccured++
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
id, err := r.TransactionAdd(t, job)
|
|
|
|
if err != nil {
|
|
|
|
log.Errorf("repository initDB(): %v", err)
|
|
|
|
errorOccured++
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
|
|
|
|
for _, tag := range job.Tags {
|
|
|
|
tagstr := tag.Name + ":" + tag.Type
|
|
|
|
tagId, ok := tags[tagstr]
|
|
|
|
if !ok {
|
|
|
|
tagId, err = r.TransactionAddTag(t, tag)
|
|
|
|
if err != nil {
|
|
|
|
log.Errorf("Error adding tag: %v", err)
|
|
|
|
errorOccured++
|
|
|
|
continue
|
|
|
|
}
|
|
|
|
tags[tagstr] = tagId
|
|
|
|
}
|
|
|
|
|
|
|
|
r.TransactionSetTag(t, id, tagId)
|
|
|
|
}
|
|
|
|
|
|
|
|
if err == nil {
|
|
|
|
i += 1
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if errorOccured > 0 {
|
|
|
|
log.Warnf("Error in import of %d jobs!", errorOccured)
|
|
|
|
}
|
|
|
|
|
|
|
|
r.TransactionEnd(t)
|
|
|
|
log.Printf("A total of %d jobs have been registered in %.3f seconds.\n", i, time.Since(starttime).Seconds())
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
// This function also sets the subcluster if necessary!
|
|
|
|
func SanityChecks(job *schema.BaseJob) error {
|
|
|
|
if c := archive.GetCluster(job.Cluster); c == nil {
|
|
|
|
return fmt.Errorf("no such cluster: %v", job.Cluster)
|
|
|
|
}
|
|
|
|
if err := archive.AssignSubCluster(job); err != nil {
|
|
|
|
log.Warn("Error while assigning subcluster to job")
|
|
|
|
return err
|
|
|
|
}
|
|
|
|
if !job.State.Valid() {
|
|
|
|
return fmt.Errorf("not a valid job state: %v", job.State)
|
|
|
|
}
|
|
|
|
if len(job.Resources) == 0 || len(job.User) == 0 {
|
|
|
|
return fmt.Errorf("'resources' and 'user' should not be empty")
|
|
|
|
}
|
2023-05-04 07:00:30 +02:00
|
|
|
if job.NumAcc < 0 || job.NumHWThreads < 0 || job.NumNodes < 1 {
|
2023-04-28 08:49:58 +02:00
|
|
|
return fmt.Errorf("'numNodes', 'numAcc' or 'numHWThreads' invalid")
|
|
|
|
}
|
|
|
|
if len(job.Resources) != int(job.NumNodes) {
|
|
|
|
return fmt.Errorf("len(resources) does not equal numNodes (%d vs %d)", len(job.Resources), job.NumNodes)
|
|
|
|
}
|
|
|
|
|
|
|
|
return nil
|
|
|
|
}
|
|
|
|
|
|
|
|
func loadJobStat(job *schema.JobMeta, metric string) float64 {
|
|
|
|
if stats, ok := job.Statistics[metric]; ok {
|
2023-06-15 11:07:48 +02:00
|
|
|
if metric == "mem_used" {
|
|
|
|
return stats.Max
|
|
|
|
} else {
|
|
|
|
return stats.Avg
|
|
|
|
}
|
2023-04-28 08:49:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0.0
|
|
|
|
}
|
|
|
|
|
|
|
|
func checkJobData(d *schema.JobData) error {
|
|
|
|
for _, scopes := range *d {
|
|
|
|
// var newUnit schema.Unit
|
|
|
|
// TODO Add node scope if missing
|
|
|
|
for _, metric := range scopes {
|
|
|
|
if strings.Contains(metric.Unit.Base, "B/s") ||
|
|
|
|
strings.Contains(metric.Unit.Base, "F/s") ||
|
|
|
|
strings.Contains(metric.Unit.Base, "B") {
|
|
|
|
|
|
|
|
// get overall avg
|
|
|
|
sum := 0.0
|
|
|
|
for _, s := range metric.Series {
|
|
|
|
sum += s.Statistics.Avg
|
|
|
|
}
|
|
|
|
|
|
|
|
avg := sum / float64(len(metric.Series))
|
2023-05-04 16:03:04 +02:00
|
|
|
f, p := Normalize(avg, metric.Unit.Prefix)
|
2023-04-28 08:49:58 +02:00
|
|
|
|
|
|
|
if p != metric.Unit.Prefix {
|
|
|
|
|
|
|
|
fmt.Printf("Convert %e", f)
|
|
|
|
// for _, s := range metric.Series {
|
|
|
|
// fp := schema.ConvertFloatToFloat64(s.Data)
|
|
|
|
//
|
|
|
|
// for i := 0; i < len(fp); i++ {
|
|
|
|
// fp[i] *= f
|
|
|
|
// fp[i] = math.Ceil(fp[i])
|
|
|
|
// }
|
|
|
|
//
|
|
|
|
// s.Data = schema.GetFloat64ToFloat(fp)
|
|
|
|
// }
|
|
|
|
|
|
|
|
metric.Unit.Prefix = p
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return nil
|
|
|
|
}
|