Adopt cc-lib hierarchical JobData/Statistics structs

cc-lib changed JobData, ScopedJobStats and Job.Statistics from flat maps
to structs (a .Metrics map plus array-valued Groups) to represent
filesystem (and future interconnect) metric groups. Migrate all
construction, indexing and iteration to the new API across the archive
backends, metricstore query path, metric dispatcher, archiver, importer,
tagger, taskmanager, repository and API layers.

Semantics: DecodeJobStats now projects JobData.Groups into
ScopedJobStats.Groups, and the archiver derives per-filesystem node-scope
statistics into Job.Statistics.Groups. deepCopy, resampling and the
metric/scope filter are group-aware. The metricstore internal storage
(buffers, selector tree, checkpoint/parquet) is unchanged; all conversion
stays at the LoadData/archive-codec seam.

Note: requires the corresponding cc-lib release; bump the cc-lib
dependency version once tagged (a local go.mod replace was used during
development and is intentionally not committed).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-07-01 19:14:12 +02:00
co-authored by Claude Opus 4.8
parent 1bd3f25371
commit 0724b2dc60
25 changed files with 303 additions and 196 deletions
+6 -6
View File
@@ -299,7 +299,7 @@ func LoadAveragesFromArchive(
}
for i, m := range metrics {
if stat, ok := metaFile.Statistics[m]; ok {
if stat, ok := metaFile.Statistics.Metrics[m]; ok {
data[i] = append(data[i], schema.Float(stat.Avg))
} else {
data[i] = append(data[i], schema.NaN)
@@ -323,7 +323,7 @@ func LoadStatsFromArchive(
}
for _, m := range metrics {
stat, ok := metaFile.Statistics[m]
stat, ok := metaFile.Statistics.Metrics[m]
if !ok {
data[m] = schema.MetricStatistics{Min: 0.0, Avg: 0.0, Max: 0.0}
continue
@@ -349,19 +349,19 @@ func LoadScopedStatsFromArchive(
data, err := ar.LoadJobStats(job)
if err != nil {
cclog.Errorf("Error while loading job stats from archiveBackend: %s", err.Error())
return nil, err
return schema.ScopedJobStats{}, err
}
return data, nil
}
// GetStatistics returns all metric statistics for a job.
// Returns a map of metric names to their job-level statistics.
func GetStatistics(job *schema.Job) (map[string]schema.JobStatistics, error) {
// Returns the job-level statistics set.
func GetStatistics(job *schema.Job) (schema.JobStatisticsSet, error) {
metaFile, err := ar.LoadJobMeta(job)
if err != nil {
cclog.Errorf("Error while loading job metadata from archiveBackend: %s", err.Error())
return nil, err
return schema.JobStatisticsSet{}, err
}
return metaFile.Statistics, nil
+6 -6
View File
@@ -93,7 +93,7 @@ func loadJobData(filename string, isCompressed bool) (schema.JobData, error) {
f, err := os.Open(filename)
if err != nil {
cclog.Errorf("fsBackend LoadJobData()- %v", err)
return nil, err
return schema.JobData{}, err
}
defer f.Close()
@@ -101,7 +101,7 @@ func loadJobData(filename string, isCompressed bool) (schema.JobData, error) {
r, err := gzip.NewReader(f)
if err != nil {
cclog.Errorf(" %v", err)
return nil, err
return schema.JobData{}, err
}
defer r.Close()
@@ -126,7 +126,7 @@ func loadJobStats(filename string, isCompressed bool) (schema.ScopedJobStats, er
f, err := os.Open(filename)
if err != nil {
cclog.Errorf("fsBackend LoadJobStats()- %v", err)
return nil, err
return schema.ScopedJobStats{}, err
}
defer f.Close()
@@ -134,13 +134,13 @@ func loadJobStats(filename string, isCompressed bool) (schema.ScopedJobStats, er
r, err := gzip.NewReader(f)
if err != nil {
cclog.Errorf(" %v", err)
return nil, err
return schema.ScopedJobStats{}, err
}
defer r.Close()
if config.Keys.Validate {
if err := schema.Validate(schema.Data, r); err != nil {
return nil, fmt.Errorf("validate job data: %v", err)
return schema.ScopedJobStats{}, fmt.Errorf("validate job data: %v", err)
}
}
@@ -148,7 +148,7 @@ func loadJobStats(filename string, isCompressed bool) (schema.ScopedJobStats, er
} else {
if config.Keys.Validate {
if err := schema.Validate(schema.Data, bufio.NewReader(f)); err != nil {
return nil, fmt.Errorf("validate job data: %v", err)
return schema.ScopedJobStats{}, fmt.Errorf("validate job data: %v", err)
}
}
return DecodeJobStats(bufio.NewReader(f), filename)
+1 -1
View File
@@ -129,7 +129,7 @@ func TestLoadJobData(t *testing.T) {
t.Fatal(err)
}
for _, scopes := range data {
for _, scopes := range data.Metrics {
// fmt.Printf("Metric name: %s\n", name)
if _, exists := scopes[schema.MetricScopeNode]; !exists {
+55 -29
View File
@@ -27,47 +27,73 @@ func DecodeJobData(r io.Reader, k string) (schema.JobData, error) {
if err, ok := data.(error); ok {
cclog.Warn("Error in decoded job data set")
return nil, err
return schema.JobData{}, err
}
return data.(schema.JobData), nil
}
func DecodeJobStats(r io.Reader, k string) (schema.ScopedJobStats, error) {
jobData, err := DecodeJobData(r, k)
// Convert schema.JobData to schema.ScopedJobStats
if jobData != nil {
scopedJobStats := make(schema.ScopedJobStats)
for metric, metricData := range jobData {
if _, ok := scopedJobStats[metric]; !ok {
scopedJobStats[metric] = make(map[schema.MetricScope][]*schema.ScopedStats)
// scopedStatsFromMetrics converts a flat metric->scope->JobMetric map into the
// corresponding metric->scope->[]*ScopedStats map. It preserves the "remove
// empty" cleanup behaviour: a scope with no series is dropped, and a metric
// with no remaining scopes is dropped.
func scopedStatsFromMetrics(metrics map[string]schema.ScopedMetrics) map[string]schema.ScopedMetricStats {
result := make(map[string]schema.ScopedMetricStats)
for metric, metricData := range metrics {
if _, ok := result[metric]; !ok {
result[metric] = make(schema.ScopedMetricStats)
}
for scope, jobMetric := range metricData {
if _, ok := result[metric][scope]; !ok {
result[metric][scope] = make([]*schema.ScopedStats, 0)
}
for scope, jobMetric := range metricData {
if _, ok := scopedJobStats[metric][scope]; !ok {
scopedJobStats[metric][scope] = make([]*schema.ScopedStats, 0)
}
for _, series := range jobMetric.Series {
result[metric][scope] = append(result[metric][scope], &schema.ScopedStats{
Hostname: series.Hostname,
ID: series.ID,
Data: &series.Statistics,
})
}
for _, series := range jobMetric.Series {
scopedJobStats[metric][scope] = append(scopedJobStats[metric][scope], &schema.ScopedStats{
Hostname: series.Hostname,
ID: series.ID,
Data: &series.Statistics,
})
}
// So that one can later check len(scopedJobStats[metric][scope]): Remove from map if empty
if len(scopedJobStats[metric][scope]) == 0 {
delete(scopedJobStats[metric], scope)
if len(scopedJobStats[metric]) == 0 {
delete(scopedJobStats, metric)
}
// So that one can later check len(result[metric][scope]): Remove from map if empty
if len(result[metric][scope]) == 0 {
delete(result[metric], scope)
if len(result[metric]) == 0 {
delete(result, metric)
}
}
}
return scopedJobStats, nil
}
return nil, err
return result
}
func DecodeJobStats(r io.Reader, k string) (schema.ScopedJobStats, error) {
jobData, err := DecodeJobData(r, k)
if err != nil {
return schema.ScopedJobStats{}, err
}
// Convert schema.JobData to schema.ScopedJobStats
scopedJobStats := schema.ScopedJobStats{
Metrics: scopedStatsFromMetrics(jobData.Metrics),
}
// Project array-valued metric groups (e.g. filesystems) analogously.
for _, group := range jobData.Groups {
scopedGroup := schema.ScopedStatsGroup{Key: group.Key}
for _, inst := range group.Instances {
scopedGroup.Instances = append(scopedGroup.Instances, schema.ScopedStatsGroupInstance{
Name: inst.Name,
Type: inst.Type,
Metrics: scopedStatsFromMetrics(inst.Metrics),
})
}
scopedJobStats.Groups = append(scopedJobStats.Groups, scopedGroup)
}
return scopedJobStats, nil
}
func DecodeJobMeta(r io.Reader) (*schema.Job, error) {
+1 -1
View File
@@ -23,7 +23,7 @@ func JobToParquetRow(meta *schema.Job, data *schema.JobData) (*ParquetJobRow, er
}
var statisticsJSON []byte
if meta.Statistics != nil {
if meta.Statistics.Metrics != nil {
statisticsJSON, err = json.Marshal(meta.Statistics)
if err != nil {
return nil, fmt.Errorf("marshal statistics: %w", err)
+12 -12
View File
@@ -32,9 +32,9 @@ func TestParquetRowToJob(t *testing.T) {
{Hostname: "node001", HWThreads: []int{0, 1, 2, 3}},
{Hostname: "node002", HWThreads: []int{4, 5, 6, 7}},
},
Statistics: map[string]schema.JobStatistics{
Statistics: schema.JobStatisticsSet{Metrics: map[string]schema.JobStatistics{
"cpu_load": {Avg: 50.0, Min: 10.0, Max: 90.0},
},
}},
Tags: []*schema.Tag{
{Type: "test", Name: "tag1"},
},
@@ -49,7 +49,7 @@ func TestParquetRowToJob(t *testing.T) {
},
}
data := &schema.JobData{
data := &schema.JobData{Metrics: map[string]schema.ScopedMetrics{
"cpu_load": {
schema.MetricScopeNode: &schema.JobMetric{
Unit: schema.Unit{Base: ""},
@@ -62,7 +62,7 @@ func TestParquetRowToJob(t *testing.T) {
},
},
},
}
}}
// Convert to parquet row
row, err := JobToParquetRow(meta, data)
@@ -134,10 +134,10 @@ func TestParquetRowToJob(t *testing.T) {
t.Errorf("Resources[0].HWThreads len = %d, want 4", len(gotMeta.Resources[0].HWThreads))
}
if len(gotMeta.Statistics) != 1 {
t.Fatalf("Statistics len = %d, want 1", len(gotMeta.Statistics))
if len(gotMeta.Statistics.Metrics) != 1 {
t.Fatalf("Statistics len = %d, want 1", len(gotMeta.Statistics.Metrics))
}
if stat, ok := gotMeta.Statistics["cpu_load"]; !ok {
if stat, ok := gotMeta.Statistics.Metrics["cpu_load"]; !ok {
t.Error("Statistics missing cpu_load")
} else if stat.Avg != 50.0 {
t.Errorf("Statistics[cpu_load].Avg = %f, want 50.0", stat.Avg)
@@ -163,7 +163,7 @@ func TestParquetRowToJob(t *testing.T) {
if gotData == nil {
t.Fatal("JobData is nil")
}
cpuLoad, ok := (*gotData)["cpu_load"]
cpuLoad, ok := (*gotData).Metrics["cpu_load"]
if !ok {
t.Fatal("JobData missing cpu_load")
}
@@ -201,7 +201,7 @@ func TestParquetRowToJobNilOptionalFields(t *testing.T) {
},
}
data := &schema.JobData{
data := &schema.JobData{Metrics: map[string]schema.ScopedMetrics{
"cpu_load": {
schema.MetricScopeNode: &schema.JobMetric{
Timestep: 60,
@@ -210,7 +210,7 @@ func TestParquetRowToJobNilOptionalFields(t *testing.T) {
},
},
},
}
}}
row, err := JobToParquetRow(meta, data)
if err != nil {
@@ -228,7 +228,7 @@ func TestParquetRowToJobNilOptionalFields(t *testing.T) {
if gotMeta.Tags != nil {
t.Errorf("Tags should be nil, got %v", gotMeta.Tags)
}
if gotMeta.Statistics != nil {
if gotMeta.Statistics.Metrics != nil {
t.Errorf("Statistics should be nil, got %v", gotMeta.Statistics)
}
if gotMeta.MetaData != nil {
@@ -299,7 +299,7 @@ func TestRoundTripThroughParquetFile(t *testing.T) {
if gotData == nil {
t.Fatal("JobData is nil")
}
if _, ok := (*gotData)["cpu_load"]; !ok {
if _, ok := (*gotData).Metrics["cpu_load"]; !ok {
t.Error("JobData missing cpu_load")
}
}
+3 -3
View File
@@ -57,7 +57,7 @@ func makeTestJob(jobID int64) (*schema.Job, *schema.JobData) {
},
}
data := schema.JobData{
data := schema.JobData{Metrics: map[string]schema.ScopedMetrics{
"cpu_load": {
schema.MetricScopeNode: &schema.JobMetric{
Unit: schema.Unit{Base: ""},
@@ -70,7 +70,7 @@ func makeTestJob(jobID int64) (*schema.Job, *schema.JobData) {
},
},
},
}
}}
return meta, &data
}
@@ -132,7 +132,7 @@ func TestJobToParquetRowConversion(t *testing.T) {
if err := json.Unmarshal(decompressed, &jobData); err != nil {
t.Fatalf("unmarshal metric data: %v", err)
}
if _, ok := jobData["cpu_load"]; !ok {
if _, ok := jobData.Metrics["cpu_load"]; !ok {
t.Error("metric data missing cpu_load key")
}
}
+6 -6
View File
@@ -330,7 +330,7 @@ func (s3a *S3Archive) LoadJobData(job *schema.Job) (schema.JobData, error) {
})
if err != nil {
cclog.Errorf("S3Archive LoadJobData() > GetObject error: %v", err)
return nil, err
return schema.JobData{}, err
}
defer result.Body.Close()
@@ -349,7 +349,7 @@ func (s3a *S3Archive) LoadJobData(job *schema.Job) (schema.JobData, error) {
r, err := gzip.NewReader(result.Body)
if err != nil {
cclog.Errorf("S3Archive LoadJobData() > gzip error: %v", err)
return nil, err
return schema.JobData{}, err
}
defer r.Close()
@@ -381,14 +381,14 @@ func (s3a *S3Archive) LoadJobStats(job *schema.Job) (schema.ScopedJobStats, erro
})
if err != nil {
cclog.Errorf("S3Archive LoadJobStats() > GetObject error: %v", err)
return nil, err
return schema.ScopedJobStats{}, err
}
defer result.Body.Close()
if config.Keys.Validate {
b, _ := io.ReadAll(result.Body)
if err := schema.Validate(schema.Data, bytes.NewReader(b)); err != nil {
return nil, fmt.Errorf("validate job data: %v", err)
return schema.ScopedJobStats{}, fmt.Errorf("validate job data: %v", err)
}
return DecodeJobStats(bytes.NewReader(b), key)
}
@@ -400,14 +400,14 @@ func (s3a *S3Archive) LoadJobStats(job *schema.Job) (schema.ScopedJobStats, erro
r, err := gzip.NewReader(result.Body)
if err != nil {
cclog.Errorf("S3Archive LoadJobStats() > gzip error: %v", err)
return nil, err
return schema.ScopedJobStats{}, err
}
defer r.Close()
if config.Keys.Validate {
b, _ := io.ReadAll(r)
if err := schema.Validate(schema.Data, bytes.NewReader(b)); err != nil {
return nil, fmt.Errorf("validate job data: %v", err)
return schema.ScopedJobStats{}, fmt.Errorf("validate job data: %v", err)
}
return DecodeJobStats(bytes.NewReader(b), keyGz)
}
+5 -5
View File
@@ -252,7 +252,7 @@ func (sa *SqliteArchive) LoadJobData(job *schema.Job) (schema.JobData, error) {
job.JobID, job.Cluster, job.StartTime).Scan(&dataBlob, &compressed)
if err != nil {
cclog.Errorf("SqliteArchive LoadJobData() > query error: %v", err)
return nil, err
return schema.JobData{}, err
}
key := fmt.Sprintf("%s:%d:%d", job.Cluster, job.JobID, job.StartTime)
@@ -261,7 +261,7 @@ func (sa *SqliteArchive) LoadJobData(job *schema.Job) (schema.JobData, error) {
gzipReader, err := gzip.NewReader(reader)
if err != nil {
cclog.Errorf("SqliteArchive LoadJobData() > gzip error: %v", err)
return nil, err
return schema.JobData{}, err
}
defer gzipReader.Close()
reader = gzipReader
@@ -285,7 +285,7 @@ func (sa *SqliteArchive) LoadJobStats(job *schema.Job) (schema.ScopedJobStats, e
job.JobID, job.Cluster, job.StartTime).Scan(&dataBlob, &compressed)
if err != nil {
cclog.Errorf("SqliteArchive LoadJobStats() > query error: %v", err)
return nil, err
return schema.ScopedJobStats{}, err
}
key := fmt.Sprintf("%s:%d:%d", job.Cluster, job.JobID, job.StartTime)
@@ -294,7 +294,7 @@ func (sa *SqliteArchive) LoadJobStats(job *schema.Job) (schema.ScopedJobStats, e
gzipReader, err := gzip.NewReader(reader)
if err != nil {
cclog.Errorf("SqliteArchive LoadJobStats() > gzip error: %v", err)
return nil, err
return schema.ScopedJobStats{}, err
}
defer gzipReader.Close()
reader = gzipReader
@@ -303,7 +303,7 @@ func (sa *SqliteArchive) LoadJobStats(job *schema.Job) (schema.ScopedJobStats, e
if config.Keys.Validate {
data, _ := io.ReadAll(reader)
if err := schema.Validate(schema.Data, bytes.NewReader(data)); err != nil {
return nil, fmt.Errorf("validate job data: %v", err)
return schema.ScopedJobStats{}, fmt.Errorf("validate job data: %v", err)
}
return DecodeJobStats(bytes.NewReader(data), key)
}
+28 -28
View File
@@ -89,7 +89,7 @@ func (ccms *InternalMetricStore) LoadData(
queries, assignedScope, err := buildQueries(job, metrics, scopes, int64(resolution))
if err != nil {
cclog.Errorf("Error while building queries for jobId %d, Metrics %v, Scopes %v: %s", job.JobID, metrics, scopes, err.Error())
return nil, err
return schema.JobData{}, err
}
// Verify assignment is correct - log any inconsistencies for debugging
@@ -110,11 +110,11 @@ func (ccms *InternalMetricStore) LoadData(
resBody, err := FetchData(req)
if err != nil {
cclog.Errorf("Error while fetching data : %s", err.Error())
return nil, err
return schema.JobData{}, err
}
var errors []string
jobData := make(schema.JobData)
jobData := schema.JobData{Metrics: make(map[string]schema.ScopedMetrics)}
// Add safety check for potential index out of range errors
if len(resBody.Results) != len(req.Queries) || len(assignedScope) != len(req.Queries) {
@@ -139,8 +139,8 @@ func (ccms *InternalMetricStore) LoadData(
continue
}
if _, ok := jobData[metric]; !ok {
jobData[metric] = make(map[schema.MetricScope]*schema.JobMetric)
if _, ok := jobData.Metrics[metric]; !ok {
jobData.Metrics[metric] = make(schema.ScopedMetrics)
}
res := mc.Timestep
@@ -148,14 +148,14 @@ func (ccms *InternalMetricStore) LoadData(
res = int(row[0].Resolution)
}
jobMetric, ok := jobData[metric][scope]
jobMetric, ok := jobData.Metrics[metric][scope]
if !ok {
jobMetric = &schema.JobMetric{
Unit: mc.Unit,
Timestep: res,
Series: make([]schema.Series, 0),
}
jobData[metric][scope] = jobMetric
jobData.Metrics[metric][scope] = jobMetric
}
for ndx, res := range row {
@@ -183,9 +183,9 @@ func (ccms *InternalMetricStore) LoadData(
// So that one can later check len(jobData):
if len(jobMetric.Series) == 0 {
delete(jobData[metric], scope)
if len(jobData[metric]) == 0 {
delete(jobData, metric)
delete(jobData.Metrics[metric], scope)
if len(jobData.Metrics[metric]) == 0 {
delete(jobData.Metrics, metric)
}
}
}
@@ -407,7 +407,7 @@ func (ccms *InternalMetricStore) LoadScopedStats(
queries, assignedScope, err := buildQueries(job, metrics, scopes, 0)
if err != nil {
cclog.Errorf("Error while building queries for jobId %d, Metrics %v, Scopes %v: %s", job.JobID, metrics, scopes, err.Error())
return nil, err
return schema.ScopedJobStats{}, err
}
req := APIQueryRequest{
@@ -422,11 +422,11 @@ func (ccms *InternalMetricStore) LoadScopedStats(
resBody, err := FetchData(req)
if err != nil {
cclog.Errorf("Error while fetching data : %s", err.Error())
return nil, err
return schema.ScopedJobStats{}, err
}
var errors []string
scopedJobStats := make(schema.ScopedJobStats)
scopedJobStats := schema.ScopedJobStats{Metrics: make(map[string]schema.ScopedMetricStats)}
for i, row := range resBody.Results {
if len(row) == 0 {
@@ -437,12 +437,12 @@ func (ccms *InternalMetricStore) LoadScopedStats(
metric := query.Metric
scope := assignedScope[i]
if _, ok := scopedJobStats[metric]; !ok {
scopedJobStats[metric] = make(map[schema.MetricScope][]*schema.ScopedStats)
if _, ok := scopedJobStats.Metrics[metric]; !ok {
scopedJobStats.Metrics[metric] = make(schema.ScopedMetricStats)
}
if _, ok := scopedJobStats[metric][scope]; !ok {
scopedJobStats[metric][scope] = make([]*schema.ScopedStats, 0)
if _, ok := scopedJobStats.Metrics[metric][scope]; !ok {
scopedJobStats.Metrics[metric][scope] = make([]*schema.ScopedStats, 0)
}
for ndx, res := range row {
@@ -456,7 +456,7 @@ func (ccms *InternalMetricStore) LoadScopedStats(
SanitizeStats(&res.Avg, &res.Min, &res.Max)
scopedJobStats[metric][scope] = append(scopedJobStats[metric][scope], &schema.ScopedStats{
scopedJobStats.Metrics[metric][scope] = append(scopedJobStats.Metrics[metric][scope], &schema.ScopedStats{
Hostname: query.Hostname,
ID: id,
Data: &schema.MetricStatistics{
@@ -467,11 +467,11 @@ func (ccms *InternalMetricStore) LoadScopedStats(
})
}
// So that one can later check len(scopedJobStats[metric][scope]): Remove from map if empty
if len(scopedJobStats[metric][scope]) == 0 {
delete(scopedJobStats[metric], scope)
if len(scopedJobStats[metric]) == 0 {
delete(scopedJobStats, metric)
// So that one can later check len(scopedJobStats.Metrics[metric][scope]): Remove from map if empty
if len(scopedJobStats.Metrics[metric][scope]) == 0 {
delete(scopedJobStats.Metrics[metric], scope)
if len(scopedJobStats.Metrics[metric]) == 0 {
delete(scopedJobStats.Metrics, metric)
}
}
}
@@ -695,14 +695,14 @@ func (ccms *InternalMetricStore) LoadNodeListData(
// Init Nested Map Data Structures If Not Found
hostData, ok := data[query.Hostname]
if !ok {
hostData = make(schema.JobData)
hostData = schema.JobData{Metrics: make(map[string]schema.ScopedMetrics)}
data[query.Hostname] = hostData
}
metricData, ok := hostData[metric]
metricData, ok := hostData.Metrics[metric]
if !ok {
metricData = make(map[schema.MetricScope]*schema.JobMetric)
data[query.Hostname][metric] = metricData
metricData = make(schema.ScopedMetrics)
data[query.Hostname].Metrics[metric] = metricData
}
scopeData, ok := metricData[scope]
@@ -712,7 +712,7 @@ func (ccms *InternalMetricStore) LoadNodeListData(
Timestep: res,
Series: make([]schema.Series, 0),
}
data[query.Hostname][metric][scope] = scopeData
data[query.Hostname].Metrics[metric][scope] = scopeData
}
for ndx, res := range row {