ClusterCockpit/cc-backend
1
0
Fork 0
mirror of https://github.com/ClusterCockpit/cc-backend synced 2024-12-25 21:09:05 +01:00
Code Issues Packages Releases Activity

Intermediate Save commit

Browse Source
This commit is contained in:
Jan Eitzinger 2022-10-04 10:12:35 +02:00
parent 7251344d4a
commit 29d215fcea
10 changed files with 125 additions and 59 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

35
internal/repository/init.go
View File

@ -17,6 +17,7 @@ import (
"github.com/ClusterCockpit/cc-backend/pkg/archive"
"github.com/ClusterCockpit/cc-backend/pkg/log"
"github.com/ClusterCockpit/cc-backend/pkg/schema"
"github.com/ClusterCockpit/cc-backend/pkg/units"
)
// `AUTO_INCREMENT` is in a comment because of this hack:
@ -103,11 +104,11 @@ func HandleImportFlag(flag string) error {
return err
}
if config.Keys.Validate {
// if config.Keys.Validate {
if err := schema.Validate(schema.Meta, bytes.NewReader(raw)); err != nil {
return fmt.Errorf("validate job meta: %v", err)
}
}
// }
dec := json.NewDecoder(bytes.NewReader(raw))
dec.DisallowUnknownFields()
jobMeta := schema.JobMeta{BaseJob: schema.JobDefaults}
@ -132,6 +133,7 @@ func HandleImportFlag(flag string) error {
return err
}
checkJobData(&jobData)
SanityChecks(&jobMeta.BaseJob)
jobMeta.MonitoringStatus = schema.MonitoringStatusArchivingSuccessful
if job, err := GetJobRepository().Find(&jobMeta.JobID, &jobMeta.Cluster, &jobMeta.StartTime); err != sql.ErrNoRows {
@ -368,14 +370,31 @@ func loadJobStat(job *schema.JobMeta, metric string) float64 {
}
func checkJobData(d *schema.JobData) error {
// for name, scopes := range *d {
for _, scopes := range *d {
var newUnit string
// 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") {
// for scope, metric := range scopes {
// // 1. Unit normalisation
// First get overall avg
sum := 0.0
for _, s := range metric.Series {
sum += s.Statistics.Avg
}
// }
// // 2. Add node scope if missing
avg := sum / float64(len(metric.Series))
// }
for _, s := range metric.Series {
fp := schema.ConvertFloatToFloat64(s.Data)
// Normalize values with new unit prefix
units.NormalizeSeries(fp, avg, metric.Unit, &newUnit)
s.Data = schema.GetFloat64ToFloat(fp)
}
metric.Unit = newUnit
}
}
}
return nil
}

22
internal/repository/job.go
View File

@ -211,7 +211,13 @@ func (r *JobRepository) Stop(
}
// 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) {
func (r *JobRepository) CountGroupedJobs(
ctx context.Context,
aggreg model.Aggregate,
filters []*model.JobFilter,
weight *model.Weights,
limit *int) (map[string]int, error) {
if !aggreg.IsValid() {
return nil, errors.New("invalid aggregate")
}
@ -301,10 +307,14 @@ func (r *JobRepository) Archive(
var ErrNotFound = errors.New("no such job or user")
// FindJobOrUser returns a job database ID or a username if a job or user machtes the search term.
// As 0 is a valid job id, check if username is "" instead in order to check what machted.
// If nothing matches the search, `ErrNotFound` is returned.
func (r *JobRepository) FindJobOrUser(ctx context.Context, searchterm string) (job int64, username string, err error) {
// FindJobOrUser returns a job database ID or a username if a job or user
// machtes the search term. As 0 is a valid job id, check if username is ""
// instead in order to check what matched. If nothing matches the search,
// `ErrNotFound` is returned.
func (r *JobRepository) FindJobOrUser(
ctx context.Context,
searchterm string) (job int64, username string, err error) {
user := auth.GetUser(ctx)
if id, err := strconv.Atoi(searchterm); err == nil {
qb := sq.Select("job.id").From("job").Where("job.job_id = ?", id)
@ -353,6 +363,7 @@ func (r *JobRepository) Partitions(cluster string) ([]string, error) {
// AllocatedNodes returns a map of all subclusters to a map of hostnames to the amount of jobs running on that host.
// Hosts with zero jobs running on them will not show up!
func (r *JobRepository) AllocatedNodes(cluster string) (map[string]map[string]int, error) {
subclusters := make(map[string]map[string]int)
rows, err := sq.Select("resources", "subcluster").From("job").
Where("job.job_state = 'running'").
@ -390,6 +401,7 @@ func (r *JobRepository) AllocatedNodes(cluster string) (map[string]map[string]in
}
func (r *JobRepository) StopJobsExceedingWalltimeBy(seconds int) error {
res, err := sq.Update("job").
Set("monitoring_status", schema.MonitoringStatusArchivingFailed).
Set("duration", 0).

2
pkg/schema/cluster.go
View File

@ -45,7 +45,7 @@ type SubClusterConfig struct {
type MetricConfig struct {
Name string `json:"name"`
Unit string `json:"unit"`
Unit Unit `json:"unit"`
Scope MetricScope `json:"scope"`
Aggregation *string `json:"aggregation"`
Timestep int `json:"timestep"`

20
pkg/schema/float.go
View File

@ -107,3 +107,23 @@ func (s *Series) MarshalJSON() ([]byte, error) {
buf = append(buf, ']', '}')
return buf, nil
}
func ConvertFloatToFloat64(s []Float) []float64 {
fp := make([]float64, len(s))
for i, val := range s {
fp[i] = float64(val)
}
return fp
}
func GetFloat64ToFloat(s []float64) []Float {
fp := make([]Float, len(s))
for i, val := range s {
fp[i] = Float(val)
}
return fp
}

8
pkg/schema/job.go
View File

@ -89,11 +89,15 @@ var JobDefaults BaseJob = BaseJob{
MonitoringStatus: MonitoringStatusRunningOrArchiving,
}
type Unit struct {
Base string `json:"base"`
Prefix string `json:"prefix"`
}
// JobStatistics model
// @Description Specification for job metric statistics.
type JobStatistics struct {
// Metric unit (see schema/unit.schema.json)
Unit string `json:"unit" example:"GHz"`
Unit Unit `json:"unit" example:"GHz"`
Avg float64 `json:"avg" example:"2500" minimum:"0"` // Job metric average
Min float64 `json:"min" example:"2000" minimum:"0"` // Job metric minimum
Max float64 `json:"max" example:"3000" minimum:"0"` // Job metric maximum

2
pkg/schema/metrics.go
View File

@ -15,7 +15,7 @@ import (
type JobData map[string]map[MetricScope]*JobMetric
type JobMetric struct {
Unit string `json:"unit"`
Unit Unit `json:"unit"`
Scope MetricScope `json:"scope"`
Timestep int `json:"timestep"`
Series []Series `json:"series"`

3
pkg/schema/schemas/cluster.schema.json
View File

@ -21,7 +21,7 @@
},
"unit": {
"description": "Metric unit",
"type": "string"
"$ref": "embedfs://unit.schema.json"
},
"scope": {
"description": "Native measurement resolution",
@ -38,7 +38,6 @@
"sum",
"avg"
]
},
"subClusters": {
"description": "Array of cluster hardware partition metric thresholds",

1
pkg/schema/schemas/job-meta.schema.json
View File

@ -349,7 +349,6 @@
"jobState",
"duration",
"resources",
"tags",
"statistics"
]
}

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

@ -5,7 +5,7 @@
"description": "Format specification for job metric units",
"type": "object",
"properties": {
"base_unit": {
"base": {
"description": "Metric base unit",
"type": "string",
"enum": [
@ -36,6 +36,6 @@
}
},
"required": [
"base_unit"
"base"
]
}

13
test/integration_test.go
View File

@ -602,4 +602,17 @@ func testImportFlag(t *testing.T) {
if len(data) != 8 {
t.Errorf("Job data length: Got %d, want 8", len(data))
}
r := map[string]string{"mem_used": "GB", "net_bw": "KB/s",
"cpu_power": "W", "cpu_used": "cpu_used",
"file_bw": "KB/s", "flops_any": "Flops/s",
"mem_bw": "GB/s", "ipc": "IPC"}
for name, scopes := range data {
for _, metric := range scopes {
if metric.Unit != r[name] {
t.Errorf("Metric %s unit: Got %s, want %s", name, metric.Unit, r[name])
}
}
}
}