Add rest-api for starting/stoping jobs

metricdata/archive.go

@@ -2,13 +2,18 @@ package metricdata
 
 import (
 	"bufio"
+	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
+	"math"
 	"os"
+	"path"
 	"path/filepath"
 	"strconv"
 	"strings"
 
+	"github.com/ClusterCockpit/cc-jobarchive/config"
 	"github.com/ClusterCockpit/cc-jobarchive/graph/model"
 	"github.com/ClusterCockpit/cc-jobarchive/schema"
 )
@@ -23,8 +28,13 @@ func getPath(job *model.Job, file string) (string, error) {
 		return "", err
 	}
 
-	lvl1, lvl2 := id/1000, id%1000
-	return filepath.Join(JobArchivePath, job.ClusterID, fmt.Sprintf("%d", lvl1), fmt.Sprintf("%03d", lvl2), file), nil
+	lvl1, lvl2 := fmt.Sprintf("%d", id/1000), fmt.Sprintf("%03d", id%1000)
+	legacyPath := filepath.Join(JobArchivePath, job.ClusterID, lvl1, lvl2, file)
+	if _, err := os.Stat(legacyPath); errors.Is(err, os.ErrNotExist) {
+		return filepath.Join(JobArchivePath, job.ClusterID, lvl1, lvl2, strconv.FormatInt(job.StartTime.Unix(), 10), file), nil
+	}
+
+	return legacyPath, nil
 }
 
 // Assuming job is completed/archived, return the jobs metric data.
@@ -123,3 +133,99 @@ func loadAveragesFromArchive(job *model.Job, metrics []string, data [][]schema.F
 
 	return nil
 }
+
+// Writes a running job to the job-archive
+func ArchiveJob(job *model.Job, ctx context.Context) error {
+	if job.State != model.JobStateRunning {
+		return errors.New("cannot archive job that is not running")
+	}
+
+	allMetrics := make([]string, 0)
+	metricConfigs := config.GetClusterConfig(job.ClusterID).MetricConfig
+	for _, mc := range metricConfigs {
+		allMetrics = append(allMetrics, mc.Name)
+	}
+	jobData, err := LoadData(job, allMetrics, ctx)
+	if err != nil {
+		return 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:      job.JobID,
+		UserId:     job.UserID,
+		ClusterId:  job.ClusterID,
+		NumNodes:   job.NumNodes,
+		JobState:   job.State.String(),
+		StartTime:  job.StartTime.Unix(),
+		Duration:   int64(job.Duration),
+		Nodes:      job.Nodes,
+		Tags:       tags,
+		Statistics: make(map[string]*schema.JobMetaStatistics),
+	}
+
+	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)
+		}
+
+		metaData.Statistics[metric] = &schema.JobMetaStatistics{
+			Unit: config.GetMetricConfig(job.ClusterID, metric).Unit,
+			Avg:  avg / float64(job.NumNodes),
+			Min:  min,
+			Max:  max,
+		}
+	}
+
+	dirPath, err := getPath(job, "")
+	if err != nil {
+		return err
+	}
+
+	if err := os.MkdirAll(dirPath, 0777); err != nil {
+		return err
+	}
+
+	f, err := os.Create(path.Join(dirPath, "meta.json"))
+	if err != nil {
+		return err
+	}
+	defer f.Close()
+	writer := bufio.NewWriter(f)
+	if err := json.NewEncoder(writer).Encode(metaData); err != nil {
+		return err
+	}
+	if err := writer.Flush(); err != nil {
+		return err
+	}
+
+	f, err = os.Create(path.Join(dirPath, "data.json"))
+	if err != nil {
+		return err
+	}
+	writer = bufio.NewWriter(f)
+	if err := json.NewEncoder(writer).Encode(metaData); err != nil {
+		return err
+	}
+	if err := writer.Flush(); err != nil {
+		return err
+	}
+
+	return f.Close()
+}