Introduce average resampler support

Fixes #526

Entire-Checkpoint: 893a1de325b5

internal/api/api_test.go

@@ -355,7 +355,7 @@ func TestRestApi(t *testing.T) {
 	}
 
 	t.Run("CheckArchive", func(t *testing.T) {
-		data, err := metricdispatch.LoadData(stoppedJob, []string{"load_one"}, []schema.MetricScope{schema.MetricScopeNode}, context.Background(), 60)
+		data, err := metricdispatch.LoadData(stoppedJob, []string{"load_one"}, []schema.MetricScope{schema.MetricScopeNode}, context.Background(), 60, "")
 		if err != nil {
 			t.Fatal(err)
 		}

internal/api/job.go

@@ -301,7 +301,7 @@ func (api *RestAPI) getCompleteJobByID(rw http.ResponseWriter, r *http.Request)
 	}
 
 	if r.URL.Query().Get("all-metrics") == "true" {
-		data, err = metricdispatch.LoadData(job, nil, scopes, r.Context(), resolution)
+		data, err = metricdispatch.LoadData(job, nil, scopes, r.Context(), resolution, "")
 		if err != nil {
 			cclog.Warnf("REST: error while loading all-metrics job data for JobID %d on %s", job.JobID, job.Cluster)
 			return
@@ -397,7 +397,7 @@ func (api *RestAPI) getJobByID(rw http.ResponseWriter, r *http.Request) {
 		resolution = max(resolution, mc.Timestep)
 	}
 
-	data, err := metricdispatch.LoadData(job, metrics, scopes, r.Context(), resolution)
+	data, err := metricdispatch.LoadData(job, metrics, scopes, r.Context(), resolution, "")
 	if err != nil {
 		cclog.Warnf("REST: error while loading job data for JobID %d on %s", job.JobID, job.Cluster)
 		return
@@ -1078,7 +1078,7 @@ func (api *RestAPI) getJobMetrics(rw http.ResponseWriter, r *http.Request) {
 	}
 
 	resolver := graph.GetResolverInstance()
-	data, err := resolver.Query().JobMetrics(r.Context(), id, metrics, scopes, nil)
+	data, err := resolver.Query().JobMetrics(r.Context(), id, metrics, scopes, nil, nil)
 	if err != nil {
 		if err := json.NewEncoder(rw).Encode(Response{
 			Error: &struct {

internal/archiver/archiver.go

@@ -59,7 +59,7 @@ func ArchiveJob(job *schema.Job, ctx context.Context) (*schema.Job, error) {
 		scopes = append(scopes, schema.MetricScopeAccelerator)
 	}
 
-	jobData, err := metricdispatch.LoadData(job, allMetrics, scopes, ctx, 0) // 0 Resulotion-Value retrieves highest res (60s)
+	jobData, err := metricdispatch.LoadData(job, allMetrics, scopes, ctx, 0, "") // 0 Resulotion-Value retrieves highest res (60s)
 	if err != nil {
 		cclog.Error("Error wile loading job data for archiving")
 		return nil, err

internal/graph/generated/generated.go

@@ -326,7 +326,7 @@ type ComplexityRoot struct {
 		Clusters        func(childComplexity int) int
 		GlobalMetrics   func(childComplexity int) int
 		Job             func(childComplexity int, id string) int
-		JobMetrics      func(childComplexity int, id string, metrics []string, scopes []schema.MetricScope, resolution *int) int
+		JobMetrics      func(childComplexity int, id string, metrics []string, scopes []schema.MetricScope, resolution *int, resampleAlgo *model.ResampleAlgo) int
 		JobStats        func(childComplexity int, id string, metrics []string) int
 		Jobs            func(childComplexity int, filter []*model.JobFilter, page *model.PageRequest, order *model.OrderByInput) int
 		JobsFootprints  func(childComplexity int, filter []*model.JobFilter, metrics []string) int
@@ -334,7 +334,7 @@ type ComplexityRoot struct {
 		JobsStatistics  func(childComplexity int, filter []*model.JobFilter, metrics []string, page *model.PageRequest, sortBy *model.SortByAggregate, groupBy *model.Aggregate, numDurationBins *string, numMetricBins *int) int
 		Node            func(childComplexity int, id string) int
 		NodeMetrics     func(childComplexity int, cluster string, nodes []string, scopes []schema.MetricScope, metrics []string, from time.Time, to time.Time) int
-		NodeMetricsList func(childComplexity int, cluster string, subCluster string, stateFilter string, nodeFilter string, scopes []schema.MetricScope, metrics []string, from time.Time, to time.Time, page *model.PageRequest, resolution *int) int
+		NodeMetricsList func(childComplexity int, cluster string, subCluster string, stateFilter string, nodeFilter string, scopes []schema.MetricScope, metrics []string, from time.Time, to time.Time, page *model.PageRequest, resolution *int, resampleAlgo *model.ResampleAlgo) int
 		NodeStates      func(childComplexity int, filter []*model.NodeFilter) int
 		NodeStatesTimed func(childComplexity int, filter []*model.NodeFilter, typeArg string) int
 		Nodes           func(childComplexity int, filter []*model.NodeFilter, order *model.OrderByInput) int
@@ -482,7 +482,7 @@ type QueryResolver interface {
 	NodeStates(ctx context.Context, filter []*model.NodeFilter) ([]*model.NodeStates, error)
 	NodeStatesTimed(ctx context.Context, filter []*model.NodeFilter, typeArg string) ([]*model.NodeStatesTimed, error)
 	Job(ctx context.Context, id string) (*schema.Job, error)
-	JobMetrics(ctx context.Context, id string, metrics []string, scopes []schema.MetricScope, resolution *int) ([]*model.JobMetricWithName, error)
+	JobMetrics(ctx context.Context, id string, metrics []string, scopes []schema.MetricScope, resolution *int, resampleAlgo *model.ResampleAlgo) ([]*model.JobMetricWithName, error)
 	JobStats(ctx context.Context, id string, metrics []string) ([]*model.NamedStats, error)
 	ScopedJobStats(ctx context.Context, id string, metrics []string, scopes []schema.MetricScope) ([]*model.NamedStatsWithScope, error)
 	Jobs(ctx context.Context, filter []*model.JobFilter, page *model.PageRequest, order *model.OrderByInput) (*model.JobResultList, error)
@@ -491,7 +491,7 @@ type QueryResolver interface {
 	JobsFootprints(ctx context.Context, filter []*model.JobFilter, metrics []string) (*model.Footprints, error)
 	RooflineHeatmap(ctx context.Context, filter []*model.JobFilter, rows int, cols int, minX float64, minY float64, maxX float64, maxY float64) ([][]float64, error)
 	NodeMetrics(ctx context.Context, cluster string, nodes []string, scopes []schema.MetricScope, metrics []string, from time.Time, to time.Time) ([]*model.NodeMetrics, error)
-	NodeMetricsList(ctx context.Context, cluster string, subCluster string, stateFilter string, nodeFilter string, scopes []schema.MetricScope, metrics []string, from time.Time, to time.Time, page *model.PageRequest, resolution *int) (*model.NodesResultList, error)
+	NodeMetricsList(ctx context.Context, cluster string, subCluster string, stateFilter string, nodeFilter string, scopes []schema.MetricScope, metrics []string, from time.Time, to time.Time, page *model.PageRequest, resolution *int, resampleAlgo *model.ResampleAlgo) (*model.NodesResultList, error)
 	ClusterMetrics(ctx context.Context, cluster string, metrics []string, from time.Time, to time.Time) (*model.ClusterMetrics, error)
 }
 type SubClusterResolver interface {
@@ -1665,7 +1665,7 @@ func (e *executableSchema) Complexity(ctx context.Context, typeName, field strin
 			return 0, false
 		}
 
-		return e.ComplexityRoot.Query.JobMetrics(childComplexity, args["id"].(string), args["metrics"].([]string), args["scopes"].([]schema.MetricScope), args["resolution"].(*int)), true
+		return e.ComplexityRoot.Query.JobMetrics(childComplexity, args["id"].(string), args["metrics"].([]string), args["scopes"].([]schema.MetricScope), args["resolution"].(*int), args["resampleAlgo"].(*model.ResampleAlgo)), true
 	case "Query.jobStats":
 		if e.ComplexityRoot.Query.JobStats == nil {
 			break
@@ -1753,7 +1753,7 @@ func (e *executableSchema) Complexity(ctx context.Context, typeName, field strin
 			return 0, false
 		}
 
-		return e.ComplexityRoot.Query.NodeMetricsList(childComplexity, args["cluster"].(string), args["subCluster"].(string), args["stateFilter"].(string), args["nodeFilter"].(string), args["scopes"].([]schema.MetricScope), args["metrics"].([]string), args["from"].(time.Time), args["to"].(time.Time), args["page"].(*model.PageRequest), args["resolution"].(*int)), true
+		return e.ComplexityRoot.Query.NodeMetricsList(childComplexity, args["cluster"].(string), args["subCluster"].(string), args["stateFilter"].(string), args["nodeFilter"].(string), args["scopes"].([]schema.MetricScope), args["metrics"].([]string), args["from"].(time.Time), args["to"].(time.Time), args["page"].(*model.PageRequest), args["resolution"].(*int), args["resampleAlgo"].(*model.ResampleAlgo)), true
 	case "Query.nodeStates":
 		if e.ComplexityRoot.Query.NodeStates == nil {
 			break
@@ -2524,6 +2524,12 @@ type TimeWeights {
   coreHours: [NullableFloat!]!
 }
 
+enum ResampleAlgo {
+  LTTB
+  AVERAGE
+  SIMPLE
+}
+
 enum Aggregate {
   USER
   PROJECT
@@ -2614,6 +2620,7 @@ type Query {
     metrics: [String!]
     scopes: [MetricScope!]
     resolution: Int
+    resampleAlgo: ResampleAlgo
   ): [JobMetricWithName!]!
 
   jobStats(id: ID!, metrics: [String!]): [NamedStats!]!
@@ -2673,6 +2680,7 @@ type Query {
     to: Time!
     page: PageRequest
     resolution: Int
+    resampleAlgo: ResampleAlgo
   ): NodesResultList!
 
   clusterMetrics(
@@ -3006,6 +3014,11 @@ func (ec *executionContext) field_Query_jobMetrics_args(ctx context.Context, raw
 		return nil, err
 	}
 	args["resolution"] = arg3
+	arg4, err := graphql.ProcessArgField(ctx, rawArgs, "resampleAlgo", ec.unmarshalOResampleAlgo2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐResampleAlgo)
+	if err != nil {
+		return nil, err
+	}
+	args["resampleAlgo"] = arg4
 	return args, nil
 }
 
@@ -3183,6 +3196,11 @@ func (ec *executionContext) field_Query_nodeMetricsList_args(ctx context.Context
 		return nil, err
 	}
 	args["resolution"] = arg9
+	arg10, err := graphql.ProcessArgField(ctx, rawArgs, "resampleAlgo", ec.unmarshalOResampleAlgo2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐResampleAlgo)
+	if err != nil {
+		return nil, err
+	}
+	args["resampleAlgo"] = arg10
 	return args, nil
 }
 
@@ -9436,7 +9454,7 @@ func (ec *executionContext) _Query_jobMetrics(ctx context.Context, field graphql
 		ec.fieldContext_Query_jobMetrics,
 		func(ctx context.Context) (any, error) {
 			fc := graphql.GetFieldContext(ctx)
-			return ec.Resolvers.Query().JobMetrics(ctx, fc.Args["id"].(string), fc.Args["metrics"].([]string), fc.Args["scopes"].([]schema.MetricScope), fc.Args["resolution"].(*int))
+			return ec.Resolvers.Query().JobMetrics(ctx, fc.Args["id"].(string), fc.Args["metrics"].([]string), fc.Args["scopes"].([]schema.MetricScope), fc.Args["resolution"].(*int), fc.Args["resampleAlgo"].(*model.ResampleAlgo))
 		},
 		nil,
 		ec.marshalNJobMetricWithName2ᚕᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐJobMetricWithNameᚄ,
@@ -9917,7 +9935,7 @@ func (ec *executionContext) _Query_nodeMetricsList(ctx context.Context, field gr
 		ec.fieldContext_Query_nodeMetricsList,
 		func(ctx context.Context) (any, error) {
 			fc := graphql.GetFieldContext(ctx)
-			return ec.Resolvers.Query().NodeMetricsList(ctx, fc.Args["cluster"].(string), fc.Args["subCluster"].(string), fc.Args["stateFilter"].(string), fc.Args["nodeFilter"].(string), fc.Args["scopes"].([]schema.MetricScope), fc.Args["metrics"].([]string), fc.Args["from"].(time.Time), fc.Args["to"].(time.Time), fc.Args["page"].(*model.PageRequest), fc.Args["resolution"].(*int))
+			return ec.Resolvers.Query().NodeMetricsList(ctx, fc.Args["cluster"].(string), fc.Args["subCluster"].(string), fc.Args["stateFilter"].(string), fc.Args["nodeFilter"].(string), fc.Args["scopes"].([]schema.MetricScope), fc.Args["metrics"].([]string), fc.Args["from"].(time.Time), fc.Args["to"].(time.Time), fc.Args["page"].(*model.PageRequest), fc.Args["resolution"].(*int), fc.Args["resampleAlgo"].(*model.ResampleAlgo))
 		},
 		nil,
 		ec.marshalNNodesResultList2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodesResultList,
@@ -19672,6 +19690,22 @@ func (ec *executionContext) unmarshalOPageRequest2ᚖgithubᚗcomᚋClusterCockp
 	return &res, graphql.ErrorOnPath(ctx, err)
 }
 
+func (ec *executionContext) unmarshalOResampleAlgo2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐResampleAlgo(ctx context.Context, v any) (*model.ResampleAlgo, error) {
+	if v == nil {
+		return nil, nil
+	}
+	var res = new(model.ResampleAlgo)
+	err := res.UnmarshalGQL(v)
+	return res, graphql.ErrorOnPath(ctx, err)
+}
+
+func (ec *executionContext) marshalOResampleAlgo2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐResampleAlgo(ctx context.Context, sel ast.SelectionSet, v *model.ResampleAlgo) graphql.Marshaler {
+	if v == nil {
+		return graphql.Null
+	}
+	return v
+}
+
 func (ec *executionContext) unmarshalOSchedulerState2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑlibᚋv2ᚋschemaᚐSchedulerState(ctx context.Context, v any) (*schema.SchedulerState, error) {
 	if v == nil {
 		return nil, nil

internal/graph/model/models_gen.go

@@ -328,6 +328,63 @@ func (e Aggregate) MarshalJSON() ([]byte, error) {
 	return buf.Bytes(), nil
 }
 
+type ResampleAlgo string
+
+const (
+	ResampleAlgoLttb    ResampleAlgo = "LTTB"
+	ResampleAlgoAverage ResampleAlgo = "AVERAGE"
+	ResampleAlgoSimple  ResampleAlgo = "SIMPLE"
+)
+
+var AllResampleAlgo = []ResampleAlgo{
+	ResampleAlgoLttb,
+	ResampleAlgoAverage,
+	ResampleAlgoSimple,
+}
+
+func (e ResampleAlgo) IsValid() bool {
+	switch e {
+	case ResampleAlgoLttb, ResampleAlgoAverage, ResampleAlgoSimple:
+		return true
+	}
+	return false
+}
+
+func (e ResampleAlgo) String() string {
+	return string(e)
+}
+
+func (e *ResampleAlgo) UnmarshalGQL(v any) error {
+	str, ok := v.(string)
+	if !ok {
+		return fmt.Errorf("enums must be strings")
+	}
+
+	*e = ResampleAlgo(str)
+	if !e.IsValid() {
+		return fmt.Errorf("%s is not a valid ResampleAlgo", str)
+	}
+	return nil
+}
+
+func (e ResampleAlgo) MarshalGQL(w io.Writer) {
+	fmt.Fprint(w, strconv.Quote(e.String()))
+}
+
+func (e *ResampleAlgo) UnmarshalJSON(b []byte) error {
+	s, err := strconv.Unquote(string(b))
+	if err != nil {
+		return err
+	}
+	return e.UnmarshalGQL(s)
+}
+
+func (e ResampleAlgo) MarshalJSON() ([]byte, error) {
+	var buf bytes.Buffer
+	e.MarshalGQL(&buf)
+	return buf.Bytes(), nil
+}
+
 type SortByAggregate string
 
 const (

internal/graph/schema.resolvers.go

@@ -498,7 +498,7 @@ func (r *queryResolver) Job(ctx context.Context, id string) (*schema.Job, error)
 }
 
 // JobMetrics is the resolver for the jobMetrics field.
-func (r *queryResolver) JobMetrics(ctx context.Context, id string, metrics []string, scopes []schema.MetricScope, resolution *int) ([]*model.JobMetricWithName, error) {
+func (r *queryResolver) JobMetrics(ctx context.Context, id string, metrics []string, scopes []schema.MetricScope, resolution *int, resampleAlgo *model.ResampleAlgo) ([]*model.JobMetricWithName, error) {
 	if resolution == nil { // Load from Config
 		if config.Keys.EnableResampling != nil {
 			defaultRes := slices.Max(config.Keys.EnableResampling.Resolutions)
@@ -515,7 +515,12 @@ func (r *queryResolver) JobMetrics(ctx context.Context, id string, metrics []str
 		return nil, err
 	}
 
-	data, err := metricdispatch.LoadData(job, metrics, scopes, ctx, *resolution)
+	algoName := ""
+	if resampleAlgo != nil {
+		algoName = strings.ToLower(resampleAlgo.String())
+	}
+
+	data, err := metricdispatch.LoadData(job, metrics, scopes, ctx, *resolution, algoName)
 	if err != nil {
 		cclog.Warn("Error while loading job data")
 		return nil, err
@@ -872,7 +877,7 @@ func (r *queryResolver) NodeMetrics(ctx context.Context, cluster string, nodes [
 }
 
 // NodeMetricsList is the resolver for the nodeMetricsList field.
-func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, subCluster string, stateFilter string, nodeFilter string, scopes []schema.MetricScope, metrics []string, from time.Time, to time.Time, page *model.PageRequest, resolution *int) (*model.NodesResultList, error) {
+func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, subCluster string, stateFilter string, nodeFilter string, scopes []schema.MetricScope, metrics []string, from time.Time, to time.Time, page *model.PageRequest, resolution *int, resampleAlgo *model.ResampleAlgo) (*model.NodesResultList, error) {
 	if resolution == nil { // Load from Config
 		if config.Keys.EnableResampling != nil {
 			defaultRes := slices.Max(config.Keys.EnableResampling.Resolutions)
@@ -901,8 +906,13 @@ func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, sub
 		}
 	}
 
+	algoName := ""
+	if resampleAlgo != nil {
+		algoName = strings.ToLower(resampleAlgo.String())
+	}
+
 	// data -> map hostname:jobdata
-	data, err := metricdispatch.LoadNodeListData(cluster, subCluster, nodes, metrics, scopes, *resolution, from, to, ctx)
+	data, err := metricdispatch.LoadNodeListData(cluster, subCluster, nodes, metrics, scopes, *resolution, from, to, ctx, algoName)
 	if err != nil {
 		cclog.Warn("error while loading node data (Resolver.NodeMetricsList")
 		return nil, err

internal/graph/util.go

@@ -55,7 +55,7 @@ func (r *queryResolver) rooflineHeatmap(
 		// 	resolution = max(resolution, mc.Timestep)
 		// }
 
-		jobdata, err := metricdispatch.LoadData(job, []string{"flops_any", "mem_bw"}, []schema.MetricScope{schema.MetricScopeNode}, ctx, 0)
+		jobdata, err := metricdispatch.LoadData(job, []string{"flops_any", "mem_bw"}, []schema.MetricScope{schema.MetricScopeNode}, ctx, 0, "")
 		if err != nil {
 			cclog.Warnf("Error while loading roofline metrics for job %d", *job.ID)
 			return nil, err

internal/metricdispatch/dataLoader.go

@@ -62,9 +62,10 @@ func cacheKey(
 	metrics []string,
 	scopes []schema.MetricScope,
 	resolution int,
+	resampleAlgo string,
 ) string {
-	return fmt.Sprintf("%d(%s):[%v],[%v]-%d",
-		*job.ID, job.State, metrics, scopes, resolution)
+	return fmt.Sprintf("%d(%s):[%v],[%v]-%d-%s",
+		*job.ID, job.State, metrics, scopes, resolution, resampleAlgo)
 }
 
 // LoadData retrieves metric data for a job from the appropriate backend (memory store for running jobs,
@@ -87,8 +88,9 @@ func LoadData(job *schema.Job,
 	scopes []schema.MetricScope,
 	ctx context.Context,
 	resolution int,
+	resampleAlgo string,
 ) (schema.JobData, error) {
-	data := cache.Get(cacheKey(job, metrics, scopes, resolution), func() (_ any, ttl time.Duration, size int) {
+	data := cache.Get(cacheKey(job, metrics, scopes, resolution, resampleAlgo), func() (_ any, ttl time.Duration, size int) {
 		var jd schema.JobData
 		var err error
 
@@ -136,13 +138,17 @@ func LoadData(job *schema.Job,
 
 			jd = deepCopy(jdTemp)
 
-			// Resample archived data using Largest Triangle Three Bucket algorithm to reduce data points
-			// to the requested resolution, improving transfer performance and client-side rendering.
+			// Resample archived data to reduce data points to the requested resolution,
+			// improving transfer performance and client-side rendering.
+			resampleFn, rfErr := resampler.GetResampler(resampleAlgo)
+			if rfErr != nil {
+				return rfErr, 0, 0
+			}
 			for _, v := range jd {
 				for _, v_ := range v {
 					timestep := int64(0)
 					for i := 0; i < len(v_.Series); i += 1 {
-						v_.Series[i].Data, timestep, err = resampler.LargestTriangleThreeBucket(v_.Series[i].Data, int64(v_.Timestep), int64(resolution))
+						v_.Series[i].Data, timestep, err = resampleFn(v_.Series[i].Data, int64(v_.Timestep), int64(resolution))
 						if err != nil {
 							return err, 0, 0
 						}
@@ -414,6 +420,7 @@ func LoadNodeListData(
 	resolution int,
 	from, to time.Time,
 	ctx context.Context,
+	resampleAlgo string,
 ) (map[string]schema.JobData, error) {
 	if metrics == nil {
 		for _, m := range archive.GetCluster(cluster).MetricConfig {
@@ -428,7 +435,7 @@ func LoadNodeListData(
 		return nil, err
 	}
 
-	data, err := ms.LoadNodeListData(cluster, subCluster, nodes, metrics, scopes, resolution, from, to, ctx)
+	data, err := ms.LoadNodeListData(cluster, subCluster, nodes, metrics, scopes, resolution, from, to, ctx, resampleAlgo)
 	if err != nil {
 		if len(data) != 0 {
 			cclog.Warnf("partial error loading node list data from metric store for cluster %s, subcluster %s: %s",

internal/metricdispatch/metricdata.go

@@ -51,7 +51,8 @@ type MetricDataRepository interface {
 		scopes []schema.MetricScope,
 		resolution int,
 		from, to time.Time,
-		ctx context.Context) (map[string]schema.JobData, error)
+		ctx context.Context,
+		resampleAlgo string) (map[string]schema.JobData, error)
 
 	// HealthCheck evaluates the monitoring state for a set of nodes against expected metrics.
 	HealthCheck(cluster string,

internal/metricstoreclient/cc-metric-store.go

@@ -617,6 +617,7 @@ func (ccms *CCMetricStore) LoadNodeListData(
 	resolution int,
 	from, to time.Time,
 	ctx context.Context,
+	resampleAlgo string,
 ) (map[string]schema.JobData, error) {
 	queries, assignedScope, err := ccms.buildNodeQueries(cluster, subCluster, nodes, metrics, scopes, resolution)
 	if err != nil {