Merge pull request #529 from ClusterCockpit/hotfix

Hotfix
2026-03-27 10:07:34 +01:00 · 2026-03-20 05:50:18 +01:00
parent 676025adfe bf48389aeb
commit a9f335d910
21 changed files with 249 additions and 163 deletions
--- a/api/schema.graphqls
+++ b/api/schema.graphqls
@@ -270,7 +270,8 @@ enum SortByAggregate {

 type NodeMetrics {
  host: String!
-  state: String!
+  nodeState: String!
+  metricHealth: String!
  subCluster: String!
  metrics: [JobMetricWithName!]!
 }
@@ -451,8 +452,6 @@ input JobFilter {
  duration: IntRange
  energy: FloatRange

-  minRunningFor: Int
-
  numNodes: IntRange
  numAccelerators: IntRange
  numHWThreads: IntRange
--- a/internal/graph/generated/generated.go
+++ b/internal/graph/generated/generated.go
@@ -288,10 +288,11 @@ type ComplexityRoot struct {
 	}

 	NodeMetrics struct {
-		Host       func(childComplexity int) int
-		Metrics    func(childComplexity int) int
-		State      func(childComplexity int) int
-		SubCluster func(childComplexity int) int
+		Host         func(childComplexity int) int
+		MetricHealth func(childComplexity int) int
+		Metrics      func(childComplexity int) int
+		NodeState    func(childComplexity int) int
+		SubCluster   func(childComplexity int) int
 	}

 	NodeStateResultList struct {
@@ -1501,18 +1502,24 @@ func (e *executableSchema) Complexity(ctx context.Context, typeName, field strin
 		}

 		return e.ComplexityRoot.NodeMetrics.Host(childComplexity), true
+	case "NodeMetrics.metricHealth":
+		if e.ComplexityRoot.NodeMetrics.MetricHealth == nil {
+			break
+		}
+
+		return e.ComplexityRoot.NodeMetrics.MetricHealth(childComplexity), true
 	case "NodeMetrics.metrics":
 		if e.ComplexityRoot.NodeMetrics.Metrics == nil {
 			break
 		}

 		return e.ComplexityRoot.NodeMetrics.Metrics(childComplexity), true
-	case "NodeMetrics.state":
-		if e.ComplexityRoot.NodeMetrics.State == nil {
+	case "NodeMetrics.nodeState":
+		if e.ComplexityRoot.NodeMetrics.NodeState == nil {
 			break
 		}

-		return e.ComplexityRoot.NodeMetrics.State(childComplexity), true
+		return e.ComplexityRoot.NodeMetrics.NodeState(childComplexity), true
 	case "NodeMetrics.subCluster":
 		if e.ComplexityRoot.NodeMetrics.SubCluster == nil {
 			break
@@ -2537,7 +2544,8 @@ enum SortByAggregate {

 type NodeMetrics {
  host: String!
-  state: String!
+  nodeState: String!
+  metricHealth: String!
  subCluster: String!
  metrics: [JobMetricWithName!]!
 }
@@ -2718,8 +2726,6 @@ input JobFilter {
  duration: IntRange
  energy: FloatRange

-  minRunningFor: Int
-
  numNodes: IntRange
  numAccelerators: IntRange
  numHWThreads: IntRange
@@ -8318,14 +8324,14 @@ func (ec *executionContext) fieldContext_NodeMetrics_host(_ context.Context, fie
 	return fc, nil
 }

-func (ec *executionContext) _NodeMetrics_state(ctx context.Context, field graphql.CollectedField, obj *model.NodeMetrics) (ret graphql.Marshaler) {
+func (ec *executionContext) _NodeMetrics_nodeState(ctx context.Context, field graphql.CollectedField, obj *model.NodeMetrics) (ret graphql.Marshaler) {
 	return graphql.ResolveField(
 		ctx,
 		ec.OperationContext,
 		field,
-		ec.fieldContext_NodeMetrics_state,
+		ec.fieldContext_NodeMetrics_nodeState,
 		func(ctx context.Context) (any, error) {
-			return obj.State, nil
+			return obj.NodeState, nil
 		},
 		nil,
 		ec.marshalNString2string,
@@ -8334,7 +8340,36 @@ func (ec *executionContext) _NodeMetrics_state(ctx context.Context, field graphq
 	)
 }

-func (ec *executionContext) fieldContext_NodeMetrics_state(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
+func (ec *executionContext) fieldContext_NodeMetrics_nodeState(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
+	fc = &graphql.FieldContext{
+		Object:     "NodeMetrics",
+		Field:      field,
+		IsMethod:   false,
+		IsResolver: false,
+		Child: func(ctx context.Context, field graphql.CollectedField) (*graphql.FieldContext, error) {
+			return nil, errors.New("field of type String does not have child fields")
+		},
+	}
+	return fc, nil
+}
+
+func (ec *executionContext) _NodeMetrics_metricHealth(ctx context.Context, field graphql.CollectedField, obj *model.NodeMetrics) (ret graphql.Marshaler) {
+	return graphql.ResolveField(
+		ctx,
+		ec.OperationContext,
+		field,
+		ec.fieldContext_NodeMetrics_metricHealth,
+		func(ctx context.Context) (any, error) {
+			return obj.MetricHealth, nil
+		},
+		nil,
+		ec.marshalNString2string,
+		true,
+		true,
+	)
+}
+
+func (ec *executionContext) fieldContext_NodeMetrics_metricHealth(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
 	fc = &graphql.FieldContext{
 		Object:     "NodeMetrics",
 		Field:      field,
@@ -8668,8 +8703,10 @@ func (ec *executionContext) fieldContext_NodesResultList_items(_ context.Context
 			switch field.Name {
 			case "host":
 				return ec.fieldContext_NodeMetrics_host(ctx, field)
-			case "state":
-				return ec.fieldContext_NodeMetrics_state(ctx, field)
+			case "nodeState":
+				return ec.fieldContext_NodeMetrics_nodeState(ctx, field)
+			case "metricHealth":
+				return ec.fieldContext_NodeMetrics_metricHealth(ctx, field)
 			case "subCluster":
 				return ec.fieldContext_NodeMetrics_subCluster(ctx, field)
 			case "metrics":
@@ -9846,8 +9883,10 @@ func (ec *executionContext) fieldContext_Query_nodeMetrics(ctx context.Context,
 			switch field.Name {
 			case "host":
 				return ec.fieldContext_NodeMetrics_host(ctx, field)
-			case "state":
-				return ec.fieldContext_NodeMetrics_state(ctx, field)
+			case "nodeState":
+				return ec.fieldContext_NodeMetrics_nodeState(ctx, field)
+			case "metricHealth":
+				return ec.fieldContext_NodeMetrics_metricHealth(ctx, field)
 			case "subCluster":
 				return ec.fieldContext_NodeMetrics_subCluster(ctx, field)
 			case "metrics":
@@ -13292,7 +13331,7 @@ func (ec *executionContext) unmarshalInputJobFilter(ctx context.Context, obj any
 		asMap[k] = v
 	}

-	fieldsInOrder := [...]string{"tags", "dbId", "jobId", "arrayJobId", "user", "project", "jobName", "cluster", "subCluster", "partition", "duration", "energy", "minRunningFor", "numNodes", "numAccelerators", "numHWThreads", "startTime", "state", "metricStats", "shared", "schedule", "node"}
+	fieldsInOrder := [...]string{"tags", "dbId", "jobId", "arrayJobId", "user", "project", "jobName", "cluster", "subCluster", "partition", "duration", "energy", "numNodes", "numAccelerators", "numHWThreads", "startTime", "state", "metricStats", "shared", "schedule", "node"}
 	for _, k := range fieldsInOrder {
 		v, ok := asMap[k]
 		if !ok {
@@ -13383,13 +13422,6 @@ func (ec *executionContext) unmarshalInputJobFilter(ctx context.Context, obj any
 				return it, err
 			}
 			it.Energy = data
-		case "minRunningFor":
-			ctx := graphql.WithPathContext(ctx, graphql.NewPathWithField("minRunningFor"))
-			data, err := ec.unmarshalOInt2ᚖint(ctx, v)
-			if err != nil {
-				return it, err
-			}
-			it.MinRunningFor = data
 		case "numNodes":
 			ctx := graphql.WithPathContext(ctx, graphql.NewPathWithField("numNodes"))
 			data, err := ec.unmarshalOIntRange2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋconfigᚐIntRange(ctx, v)
@@ -15926,8 +15958,13 @@ func (ec *executionContext) _NodeMetrics(ctx context.Context, sel ast.SelectionS
 			if out.Values[i] == graphql.Null {
 				out.Invalids++
 			}
-		case "state":
-			out.Values[i] = ec._NodeMetrics_state(ctx, field, obj)
+		case "nodeState":
+			out.Values[i] = ec._NodeMetrics_nodeState(ctx, field, obj)
+			if out.Values[i] == graphql.Null {
+				out.Invalids++
+			}
+		case "metricHealth":
+			out.Values[i] = ec._NodeMetrics_metricHealth(ctx, field, obj)
 			if out.Values[i] == graphql.Null {
 				out.Invalids++
 			}
--- a/internal/graph/model/models_gen.go
+++ b/internal/graph/model/models_gen.go
@@ -75,7 +75,6 @@ type JobFilter struct {
 	Partition       *StringInput      `json:"partition,omitempty"`
 	Duration        *config.IntRange  `json:"duration,omitempty"`
 	Energy          *FloatRange       `json:"energy,omitempty"`
-	MinRunningFor   *int              `json:"minRunningFor,omitempty"`
 	NumNodes        *config.IntRange  `json:"numNodes,omitempty"`
 	NumAccelerators *config.IntRange  `json:"numAccelerators,omitempty"`
 	NumHWThreads    *config.IntRange  `json:"numHWThreads,omitempty"`
@@ -194,10 +193,11 @@ type NodeFilter struct {
 }

 type NodeMetrics struct {
-	Host       string               `json:"host"`
-	State      string               `json:"state"`
-	SubCluster string               `json:"subCluster"`
-	Metrics    []*JobMetricWithName `json:"metrics"`
+	Host         string               `json:"host"`
+	NodeState    string               `json:"nodeState"`
+	MetricHealth string               `json:"metricHealth"`
+	SubCluster   string               `json:"subCluster"`
+	Metrics      []*JobMetricWithName `json:"metrics"`
 }

 type NodeStateResultList struct {
--- a/internal/graph/schema.resolvers.go
+++ b/internal/graph/schema.resolvers.go
@@ -676,6 +676,11 @@ func (r *queryResolver) JobsStatistics(ctx context.Context, filter []*model.JobF
 			// Use request-scoped cache: multiple aliases with same (filter, groupBy)
 			// but different sortBy/page hit the DB only once.
 			if cache := getStatsGroupCache(ctx); cache != nil {
+				// Ensure the sort field is computed even if not in the GraphQL selection,
+				// because sortAndPageStats will sort by it in memory.
+				if sortBy != nil {
+					reqFields[sortByFieldName(*sortBy)] = true
+				}
 				key := statsCacheKey(filter, groupBy, reqFields)
 				var allStats []*model.JobsStatistics
 				allStats, err = cache.getOrCompute(key, func() ([]*model.JobsStatistics, error) {
@@ -840,14 +845,15 @@ func (r *queryResolver) NodeMetrics(ctx context.Context, cluster string, nodes [
 	}

 	nodeRepo := repository.GetNodeRepository()
-	stateMap, _ := nodeRepo.MapNodes(cluster)
+	nodeStateMap, metricHealthMap, _ := nodeRepo.MapNodes(cluster)

 	nodeMetrics := make([]*model.NodeMetrics, 0, len(data))
 	for hostname, metrics := range data {
 		host := &model.NodeMetrics{
-			Host:    hostname,
-			State:   stateMap[hostname],
-			Metrics: make([]*model.JobMetricWithName, 0, len(metrics)*len(scopes)),
+			Host:         hostname,
+			NodeState:    nodeStateMap[hostname],
+			MetricHealth: metricHealthMap[hostname],
+			Metrics:      make([]*model.JobMetricWithName, 0, len(metrics)*len(scopes)),
 		}
 		host.SubCluster, err = archive.GetSubClusterByNode(cluster, hostname)
 		if err != nil {
@@ -889,7 +895,7 @@ func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, sub

 	nodeRepo := repository.GetNodeRepository()
 	// nodes -> array hostname
-	nodes, stateMap, countNodes, hasNextPage, nerr := nodeRepo.GetNodesForList(ctx, cluster, subCluster, stateFilter, nodeFilter, page)
+	nodes, nodeStateMap, metricHealthMap, countNodes, hasNextPage, nerr := nodeRepo.GetNodesForList(ctx, cluster, subCluster, stateFilter, nodeFilter, page)
 	if nerr != nil {
 		return nil, errors.New("could not retrieve node list required for resolving NodeMetricsList")
 	}
@@ -910,9 +916,10 @@ func (r *queryResolver) NodeMetricsList(ctx context.Context, cluster string, sub
 	nodeMetricsList := make([]*model.NodeMetrics, 0, len(data))
 	for _, hostname := range nodes {
 		host := &model.NodeMetrics{
-			Host:    hostname,
-			State:   stateMap[hostname],
-			Metrics: make([]*model.JobMetricWithName, 0),
+			Host:         hostname,
+			NodeState:    nodeStateMap[hostname],
+			MetricHealth: metricHealthMap[hostname],
+			Metrics:      make([]*model.JobMetricWithName, 0),
 		}
 		host.SubCluster, err = archive.GetSubClusterByNode(cluster, hostname)
 		if err != nil {
--- a/internal/graph/stats_cache.go
+++ b/internal/graph/stats_cache.go
@@ -107,6 +107,33 @@ func sortAndPageStats(allStats []*model.JobsStatistics, sortBy *model.SortByAggr
 	return sorted
 }

+// sortByFieldName maps a SortByAggregate enum to the corresponding reqFields key.
+// This ensures the DB computes the column that sortAndPageStats will sort by.
+func sortByFieldName(sortBy model.SortByAggregate) string {
+	switch sortBy {
+	case model.SortByAggregateTotaljobs:
+		return "totalJobs"
+	case model.SortByAggregateTotalusers:
+		return "totalUsers"
+	case model.SortByAggregateTotalwalltime:
+		return "totalWalltime"
+	case model.SortByAggregateTotalnodes:
+		return "totalNodes"
+	case model.SortByAggregateTotalnodehours:
+		return "totalNodeHours"
+	case model.SortByAggregateTotalcores:
+		return "totalCores"
+	case model.SortByAggregateTotalcorehours:
+		return "totalCoreHours"
+	case model.SortByAggregateTotalaccs:
+		return "totalAccs"
+	case model.SortByAggregateTotalacchours:
+		return "totalAccHours"
+	default:
+		return "totalJobs"
+	}
+}
+
 // statsFieldGetter returns a function that extracts the sortable int field
 // from a JobsStatistics struct for the given sort key.
 func statsFieldGetter(sortBy model.SortByAggregate) func(*model.JobsStatistics) int {
--- a/internal/metricdispatch/dataLoader.go
+++ b/internal/metricdispatch/dataLoader.go
@@ -192,10 +192,10 @@ func LoadData(job *schema.Job,
 		// Generate statistics series for jobs with many nodes to enable min/median/max graphs
 		// instead of overwhelming the UI with individual node lines. Note that newly calculated
 		// statistics use min/median/max, while archived statistics may use min/mean/max.
-		const maxSeriesSize int = 15
+		const maxSeriesSize int = 8
 		for _, scopes := range jd {
 			for _, jm := range scopes {
-				if jm.StatisticsSeries != nil || len(jm.Series) <= maxSeriesSize {
+				if jm.StatisticsSeries != nil || len(jm.Series) < maxSeriesSize {
 					continue
 				}

--- a/internal/repository/jobQuery.go
+++ b/internal/repository/jobQuery.go
@@ -275,13 +275,6 @@ func BuildWhereClause(filter *model.JobFilter, query sq.SelectBuilder) sq.Select
 		}
 	}

-	// Configurable Filter to exclude recently started jobs, see config.go: ShortRunningJobsDuration
-	if filter.MinRunningFor != nil {
-		now := time.Now().Unix()
-		// Only jobs whose start timestamp is more than MinRunningFor seconds in the past
-		// If a job completed within the configured timeframe, it will still show up after the start_time matches the condition!
-		query = query.Where(sq.Lt{"job.start_time": (now - int64(*filter.MinRunningFor))})
-	}
 	return query
 }

--- a/internal/repository/node.go
+++ b/internal/repository/node.go
@@ -593,8 +593,8 @@ func (r *NodeRepository) ListNodes(cluster string) ([]*schema.Node, error) {
 	return nodeList, nil
 }

-func (r *NodeRepository) MapNodes(cluster string) (map[string]string, error) {
-	q := sq.Select("node.hostname", "node_state.node_state").
+func (r *NodeRepository) MapNodes(cluster string) (map[string]string, map[string]string, error) {
+	q := sq.Select("node.hostname", "node_state.node_state", "node_state.health_state").
 		From("node").
 		Join("node_state ON node_state.node_id = node.id").
 		Where(latestStateCondition()).
@@ -604,22 +604,25 @@ func (r *NodeRepository) MapNodes(cluster string) (map[string]string, error) {
 	rows, err := q.RunWith(r.DB).Query()
 	if err != nil {
 		cclog.Warn("Error while querying node list")
-		return nil, err
+		return nil, nil, err
 	}

-	stateMap := make(map[string]string)
+	nodeStateMap := make(map[string]string)
+	metricHealthMap := make(map[string]string)
+
 	defer rows.Close()
 	for rows.Next() {
-		var hostname, nodestate string
-		if err := rows.Scan(&hostname, &nodestate); err != nil {
+		var hostname, nodeState, metricHealth string
+		if err := rows.Scan(&hostname, &nodeState, &metricHealth); err != nil {
 			cclog.Warn("Error while scanning node list (MapNodes)")
-			return nil, err
+			return nil, nil, err
 		}

-		stateMap[hostname] = nodestate
+		nodeStateMap[hostname] = nodeState
+		metricHealthMap[hostname] = metricHealth
 	}

-	return stateMap, nil
+	return nodeStateMap, metricHealthMap, nil
 }

 func (r *NodeRepository) CountStates(ctx context.Context, filters []*model.NodeFilter, column string) ([]*model.NodeStates, error) {
@@ -741,10 +744,11 @@ func (r *NodeRepository) GetNodesForList(
 	stateFilter string,
 	nodeFilter string,
 	page *model.PageRequest,
-) ([]string, map[string]string, int, bool, error) {
+) ([]string, map[string]string, map[string]string, int, bool, error) {
 	// Init Return Vars
 	nodes := make([]string, 0)
-	stateMap := make(map[string]string)
+	nodeStateMap := make(map[string]string)
+	metricHealthMap := make(map[string]string)
 	countNodes := 0
 	hasNextPage := false

@@ -778,7 +782,7 @@ func (r *NodeRepository) GetNodesForList(
 	rawNodes, serr := r.QueryNodes(ctx, queryFilters, page, nil) // Order not Used
 	if serr != nil {
 		cclog.Warn("error while loading node database data (Resolver.NodeMetricsList)")
-		return nil, nil, 0, false, serr
+		return nil, nil, nil, 0, false, serr
 	}

 	// Intermediate Node Result Info
@@ -787,7 +791,8 @@ func (r *NodeRepository) GetNodesForList(
 			continue
 		}
 		nodes = append(nodes, node.Hostname)
-		stateMap[node.Hostname] = string(node.NodeState)
+		nodeStateMap[node.Hostname] = string(node.NodeState)
+		metricHealthMap[node.Hostname] = string(node.HealthState)
 	}

 	// Special Case: Find Nodes not in DB node table but in metricStore only
@@ -847,7 +852,7 @@ func (r *NodeRepository) GetNodesForList(
 		countNodes, cerr = r.CountNodes(ctx, queryFilters)
 		if cerr != nil {
 			cclog.Warn("error while counting node database data (Resolver.NodeMetricsList)")
-			return nil, nil, 0, false, cerr
+			return nil, nil, nil, 0, false, cerr
 		}
 		hasNextPage = page.Page*page.ItemsPerPage < countNodes
 	}
@@ -857,7 +862,7 @@ func (r *NodeRepository) GetNodesForList(
 		nodes, countNodes, hasNextPage = getNodesFromTopol(cluster, subCluster, nodeFilter, page)
 	}

-	return nodes, stateMap, countNodes, hasNextPage, nil
+	return nodes, nodeStateMap, metricHealthMap, countNodes, hasNextPage, nil
 }

 func AccessCheck(ctx context.Context, query sq.SelectBuilder) (sq.SelectBuilder, error) {
--- a/web/frontend/package-lock.json
+++ b/web/frontend/package-lock.json
@@ -1,12 +1,12 @@
 {
  "name": "cc-frontend",
-  "version": "1.5.0",
-  "lockfileVersion": 3,
+  "version": "1.5.2",
+  "lockfileVersion": 4,
  "requires": true,
  "packages": {
    "": {
      "name": "cc-frontend",
-      "version": "1.5.0",
+      "version": "1.5.2",
      "license": "MIT",
      "dependencies": {
        "@rollup/plugin-replace": "^6.0.3",
@@ -27,7 +27,7 @@
        "rollup": "^4.59.0",
        "rollup-plugin-css-only": "^4.5.5",
        "rollup-plugin-svelte": "^7.2.3",
-        "svelte": "^5.53.9"
+        "svelte": "^5.54.0"
      }
    },
    "node_modules/@0no-co/graphql.web": {
@@ -45,9 +45,9 @@
      }
    },
    "node_modules/@babel/runtime": {
-      "version": "7.28.6",
-      "resolved": "https://registry.npmjs.org/@babel/runtime/-/runtime-7.28.6.tgz",
-      "integrity": "sha512-05WQkdpL9COIMz4LjTxGpPNCdlpyimKppYNoJ5Di5EUObifl8t4tuLuUBBZEpoLYOmfvIWrsp9fCl0HoPRVTdA==",
+      "version": "7.29.2",
+      "resolved": "https://registry.npmjs.org/@babel/runtime/-/runtime-7.29.2.tgz",
+      "integrity": "sha512-JiDShH45zKHWyGe4ZNVRrCjBz8Nh9TMmZG1kh4QTK8hCBTWBi8Da+i7s1fJw7/lYpM4ccepSNfqzZ/QvABBi5g==",
      "license": "MIT",
      "engines": {
        "node": ">=6.9.0"
@@ -654,6 +654,19 @@
      "integrity": "sha512-ScaPdn1dQczgbl0QFTeTOmVHFULt394XJgOQNoyVhZ6r2vLnMLJfBPd53SB52T/3G36VI1/g2MZaX0cwDuXsfw==",
      "license": "MIT"
    },
+    "node_modules/@typescript-eslint/types": {
+      "version": "8.57.1",
+      "resolved": "https://registry.npmjs.org/@typescript-eslint/types/-/types-8.57.1.tgz",
+      "integrity": "sha512-S29BOBPJSFUiblEl6RzPPjJt6w25A6XsBqRVDt53tA/tlL8q7ceQNZHTjPeONt/3S7KRI4quk+yP9jK2WjBiPQ==",
+      "license": "MIT",
+      "engines": {
+        "node": "^18.18.0 || ^20.9.0 || >=21.1.0"
+      },
+      "funding": {
+        "type": "opencollective",
+        "url": "https://opencollective.com/typescript-eslint"
+      }
+    },
    "node_modules/@urql/core": {
      "version": "5.2.0",
      "resolved": "https://registry.npmjs.org/@urql/core/-/core-5.2.0.tgz",
@@ -790,9 +803,9 @@
      }
    },
    "node_modules/devalue": {
-      "version": "5.6.3",
-      "resolved": "https://registry.npmjs.org/devalue/-/devalue-5.6.3.tgz",
-      "integrity": "sha512-nc7XjUU/2Lb+SvEFVGcWLiKkzfw8+qHI7zn8WYXKkLMgfGSHbgCEaR6bJpev8Cm6Rmrb19Gfd/tZvGqx9is3wg==",
+      "version": "5.6.4",
+      "resolved": "https://registry.npmjs.org/devalue/-/devalue-5.6.4.tgz",
+      "integrity": "sha512-Gp6rDldRsFh/7XuouDbxMH3Mx8GMCcgzIb1pDTvNyn8pZGQ22u+Wa+lGV9dQCltFQ7uVw0MhRyb8XDskNFOReA==",
      "license": "MIT"
    },
    "node_modules/escape-latex": {
@@ -808,12 +821,13 @@
      "license": "MIT"
    },
    "node_modules/esrap": {
-      "version": "2.2.3",
-      "resolved": "https://registry.npmjs.org/esrap/-/esrap-2.2.3.tgz",
-      "integrity": "sha512-8fOS+GIGCQZl/ZIlhl59htOlms6U8NvX6ZYgYHpRU/b6tVSh3uHkOHZikl3D4cMbYM0JlpBe+p/BkZEi8J9XIQ==",
+      "version": "2.2.4",
+      "resolved": "https://registry.npmjs.org/esrap/-/esrap-2.2.4.tgz",
+      "integrity": "sha512-suICpxAmZ9A8bzJjEl/+rLJiDKC0X4gYWUxT6URAWBLvlXmtbZd5ySMu/N2ZGEtMCAmflUDPSehrP9BQcsGcSg==",
      "license": "MIT",
      "dependencies": {
-        "@jridgewell/sourcemap-codec": "^1.4.15"
+        "@jridgewell/sourcemap-codec": "^1.4.15",
+        "@typescript-eslint/types": "^8.2.0"
      }
    },
    "node_modules/estree-walker": {
@@ -1193,9 +1207,9 @@
      }
    },
    "node_modules/svelte": {
-      "version": "5.53.9",
-      "resolved": "https://registry.npmjs.org/svelte/-/svelte-5.53.9.tgz",
-      "integrity": "sha512-MwDfWsN8qZzeP0jlQsWF4k/4B3csb3IbzCRggF+L/QqY7T8bbKvnChEo1cPZztF51HJQhilDbevWYl2LvXbquA==",
+      "version": "5.54.0",
+      "resolved": "https://registry.npmjs.org/svelte/-/svelte-5.54.0.tgz",
+      "integrity": "sha512-TTDxwYnHkova6Wsyj1PGt9TByuWqvMoeY1bQiuAf2DM/JeDSMw7FjRKzk8K/5mJ99vGOKhbCqTDpyAKwjp4igg==",
      "license": "MIT",
      "dependencies": {
        "@jridgewell/remapping": "^2.3.4",
@@ -1207,7 +1221,7 @@
        "aria-query": "5.3.1",
        "axobject-query": "^4.1.0",
        "clsx": "^2.1.1",
-        "devalue": "^5.6.3",
+        "devalue": "^5.6.4",
        "esm-env": "^1.2.1",
        "esrap": "^2.2.2",
        "is-reference": "^3.0.3",
@@ -1229,9 +1243,9 @@
      }
    },
    "node_modules/terser": {
-      "version": "5.46.0",
-      "resolved": "https://registry.npmjs.org/terser/-/terser-5.46.0.tgz",
-      "integrity": "sha512-jTwoImyr/QbOWFFso3YoU3ik0jBBDJ6JTOQiy/J2YxVJdZCc+5u7skhNwiOR3FQIygFqVUPHl7qbbxtjW2K3Qg==",
+      "version": "5.46.1",
+      "resolved": "https://registry.npmjs.org/terser/-/terser-5.46.1.tgz",
+      "integrity": "sha512-vzCjQO/rgUuK9sf8VJZvjqiqiHFaZLnOiimmUuOKODxWL8mm/xua7viT7aqX7dgPY60otQjUotzFMmCB4VdmqQ==",
      "dev": true,
      "license": "BSD-2-Clause",
      "dependencies": {
--- a/web/frontend/package.json
+++ b/web/frontend/package.json
@@ -1,6 +1,6 @@
 {
  "name": "cc-frontend",
-  "version": "1.5.0",
+  "version": "1.5.2",
  "license": "MIT",
  "scripts": {
    "build": "rollup -c",
@@ -14,7 +14,7 @@
    "rollup": "^4.59.0",
    "rollup-plugin-css-only": "^4.5.5",
    "rollup-plugin-svelte": "^7.2.3",
-    "svelte": "^5.53.9"
+    "svelte": "^5.54.0"
  },
  "dependencies": {
    "@rollup/plugin-replace": "^6.0.3",
--- a/web/frontend/src/DashPublic.root.svelte
+++ b/web/frontend/src/DashPublic.root.svelte
@@ -130,7 +130,7 @@
          name
          count
        }
-        # Get Current States fir Pie Charts
+        # Get Current States for Pie Charts
        nodeStates(filter: $nodeFilter) {
          state
          count
--- a/web/frontend/src/Node.root.svelte
+++ b/web/frontend/src/Node.root.svelte
@@ -57,7 +57,8 @@
    query ($cluster: String!, $nodes: [String!], $from: Time!, $to: Time!) {
      nodeMetrics(cluster: $cluster, nodes: $nodes, from: $from, to: $to) {
        host
-        state
+        nodeState
+        metricHealth
        subCluster
        metrics {
          name
@@ -92,7 +93,7 @@
      }
    }
  `;
-  // Node State Colors
+  // Node/Metric State Colors
  const stateColors = {
    allocated: 'success',
    reserved: 'info',
@@ -100,7 +101,10 @@
    mixed: 'warning',
    down: 'danger',
    unknown: 'dark',
-    notindb: 'secondary'
+    notindb: 'secondary',
+    full: 'success',
+    partial: 'warning',
+    failed: 'danger'
  }

  /* State Init */
@@ -153,31 +157,46 @@
    })
  );

-  const thisNodeState = $derived($nodeMetricsData?.data?.nodeMetrics[0]?.state ? $nodeMetricsData.data.nodeMetrics[0].state : 'notindb');
+  const thisNodeState = $derived($nodeMetricsData?.data?.nodeMetrics[0]?.nodeState || 'notindb');
+  const thisMetricHealth = $derived($nodeMetricsData?.data?.nodeMetrics[0]?.metricHealth || 'unknown');
 </script>

-<Row cols={{ xs: 2, lg: 5 }}>
+<Row cols={{ xs: 2, lg: 3}}>
  {#if $initq.error}
    <Card body color="danger">{$initq.error.message}</Card>
  {:else if $initq.fetching}
    <Spinner />
  {:else}
    <!-- Node Col -->
-    <Col>
+    <Col class="mb-2">
      <InputGroup>
        <InputGroupText><Icon name="hdd" /></InputGroupText>
        <InputGroupText>Selected Node</InputGroupText>
        <Input style="background-color: white;" type="text" value="{hostname} [{cluster} {$nodeMetricsData?.data?.nodeMetrics[0] ? `(${$nodeMetricsData.data.nodeMetrics[0].subCluster})` : ''}]" disabled/>
      </InputGroup>
    </Col>
-    <!-- State Col -->
-    <Col>
+    <!-- Node State Col -->
+    <Col class="mb-2">
      <InputGroup>
        <InputGroupText><Icon name="clipboard2-pulse" /></InputGroupText>
        <InputGroupText>Node State</InputGroupText>
        <Button class="flex-grow-1 text-center" color={stateColors[thisNodeState]} disabled>
          {#if $nodeMetricsData?.data}
-            {thisNodeState}
+            {thisNodeState.charAt(0).toUpperCase() + thisNodeState.slice(1)}
+          {:else}
+            <span><Spinner size="sm" secondary/></span>
+          {/if}
+        </Button>
+      </InputGroup>
+    </Col>
+    <!-- Metric Health Col -->
+    <Col class="mb-2">
+      <InputGroup>
+        <InputGroupText><Icon name="clipboard2-pulse" /></InputGroupText>
+        <InputGroupText>Metric Health</InputGroupText>
+        <Button class="flex-grow-1 text-center" color={stateColors[thisMetricHealth]} disabled>
+          {#if $nodeMetricsData?.data}
+            {thisMetricHealth.charAt(0).toUpperCase() + thisMetricHealth.slice(1)}
          {:else}
            <span><Spinner size="sm" secondary/></span>
          {/if}
@@ -185,7 +204,7 @@
      </InputGroup>
    </Col>
    <!-- Concurrent Col -->
-    <Col class="mt-2 mt-lg-0">
+    <Col>
      {#if $nodeJobsData.fetching}
        <Spinner />
      {:else if $nodeJobsData.data}
@@ -217,7 +236,7 @@
      />
    </Col>
    <!-- Refresh Col-->
-    <Col class="mt-2 mt-lg-0">
+    <Col>
      <Refresher
        onRefresh={() => {
          const diff = Date.now() - to;
--- a/web/frontend/src/Systems.root.svelte
+++ b/web/frontend/src/Systems.root.svelte
@@ -59,7 +59,7 @@

  /* State Init */
  let hostnameFilter = $state("");
-  let hoststateFilter = $state("all");
+  let nodeStateFilter = $state("all");
  let pendingHostnameFilter = $state("");
  let isMetricsSelectionOpen = $state(false);

@@ -210,7 +210,7 @@
      <InputGroup>
        <InputGroupText><Icon name="clipboard2-pulse" /></InputGroupText>
        <InputGroupText>State</InputGroupText>
-        <Input type="select" bind:value={hoststateFilter}>
+        <Input type="select" bind:value={nodeStateFilter}>
          {#each stateOptions as so}
            <option value={so}>{so.charAt(0).toUpperCase() + so.slice(1)}</option>
          {/each}
@@ -269,11 +269,11 @@
 {:else}
  {#if displayNodeOverview}
    <!-- ROW2-1: Node Overview (Grid Included)-->
-    <NodeOverview {cluster} {ccconfig} {selectedMetric} {globalMetrics} {from} {to} {hostnameFilter} {hoststateFilter}/>
+    <NodeOverview {cluster} {ccconfig} {selectedMetric} {globalMetrics} {from} {to} {hostnameFilter} {nodeStateFilter}/>
  {:else}
    <!-- ROW2-2: Node List (Grid Included)-->
    <NodeList pendingSelectedMetrics={selectedMetrics} {cluster} {subCluster}
-       {selectedResolution} {hostnameFilter} {hoststateFilter} {from} {to} {systemUnits}/>
+       {selectedResolution} {hostnameFilter} {nodeStateFilter} {from} {to} {systemUnits}/>
  {/if}
 {/if}

--- a/web/frontend/src/generic/joblist/JobListRow.svelte
+++ b/web/frontend/src/generic/joblist/JobListRow.svelte
@@ -229,7 +229,7 @@
            timestep={metric.data.metric.timestep}
            scope={metric.data.scope}
            series={metric.data.metric.series}
-            statisticsSeries={metric.data.metric.statisticsSeries}
+            statisticsSeries={metric.data.metric?.statisticsSeries}
            metric={metric.data.name}
            cluster={clusterInfos.find((c) => c.name == job.cluster)}
            subCluster={job.subCluster}
--- a/web/frontend/src/generic/plots/DoubleMetricPlot.svelte
+++ b/web/frontend/src/generic/plots/DoubleMetricPlot.svelte
@@ -236,7 +236,7 @@

      // conditional hide series color markers:
      if (
-        // useStatsSeries || // Min/Max/Median Self-Explanatory
+        // Min/Max/Median Self-Explanatory
        dataSize === 1 || // Only one Y-Dataseries
        dataSize > 8 // More than 8 Y-Dataseries
      ) {
@@ -273,7 +273,7 @@
      }
    }

-    if (dataSize <= 12 ) { // || useStatsSeries) {
+    if (dataSize <= 12 ) {
      return {
        hooks: {
          init: init,
--- a/web/frontend/src/generic/plots/MetricPlot.svelte
+++ b/web/frontend/src/generic/plots/MetricPlot.svelte
@@ -9,7 +9,6 @@
  - `height Number?`: The plot height [Default: 300]
  - `timestep Number`: The timestep used for X-axis rendering
  - `series [GraphQL.Series]`: The metric data object
-  - `useStatsSeries Bool?`: If this plot uses the statistics Min/Max/Median representation; automatically set to according bool [Default: false]
  - `statisticsSeries [GraphQL.StatisticsSeries]?`: Min/Max/Median representation of metric data [Default: null]
  - `cluster String?`: Cluster name of the parent job / data [Default: ""]
  - `subCluster String`: Name of the subCluster of the parent job
@@ -37,7 +36,6 @@
    height = 300,
    timestep,
    series,
-    useStatsSeries = false,
    statisticsSeries = null,
    cluster = "",
    subCluster,
@@ -78,6 +76,7 @@
  const resampleTrigger = $derived(resampleConfig?.trigger ? Number(resampleConfig.trigger) : null);
  const resampleResolutions = $derived(resampleConfig?.resolutions ? [...resampleConfig.resolutions] : null);
  const resampleMinimum = $derived(resampleConfig?.resolutions ? Math.min(...resampleConfig.resolutions) : null);
+  const useStatsSeries = $derived(!!statisticsSeries); // Display Stats Series By Default if Exists
  const thresholds = $derived(findJobAggregationThresholds(
    subClusterTopology,
    metricConfig,
@@ -243,11 +242,6 @@
    return pendingSeries;
  })

-  /* Effects */
-  $effect(() => {
-    if (!useStatsSeries && statisticsSeries != null) useStatsSeries = true;
-  })
-
  // This updates plot on all size changes if wrapper (== data) exists
  $effect(() => {
    if (plotWrapper) {
--- a/web/frontend/src/job/Metric.svelte
+++ b/web/frontend/src/job/Metric.svelte
@@ -197,7 +197,6 @@
      {zoomState}
      {thresholdState}
      statisticsSeries={statsSeries[selectedScopeIndex]}
-      useStatsSeries={!!statsSeries[selectedScopeIndex]}
      enableFlip
    />
  {/if}
--- a/web/frontend/src/systems/NodeList.svelte
+++ b/web/frontend/src/systems/NodeList.svelte
@@ -6,8 +6,8 @@
  - `subCluster String`: The nodes' subCluster [Default: ""]
  - `pendingSelectedMetrics [String]`: The array of selected metrics [Default []]
  - `selectedResolution Number?`: The selected data resolution [Default: 0]
-  - `hostnameFilter String?`: The active hostnamefilter [Default: ""]
-  - `hoststateFilter String?`: The active hoststatefilter [Default: ""]
+  - `hostnameFilter String?`: The active hostname filter [Default: ""]
+  - `nodeStateFilter String?`: The active nodeState filter [Default: ""]
  - `systemUnits Object`: The object of metric units [Default: null]
  - `from Date?`: The selected "from" date [Default: null]
  - `to Date?`: The selected "to" date [Default: null]
@@ -28,7 +28,7 @@
    pendingSelectedMetrics = [],
    selectedResolution = 0,
    hostnameFilter = "",
-    hoststateFilter = "",
+    nodeStateFilter = "",
    systemUnits = null,
    from = null,
    to = null
@@ -54,7 +54,8 @@
      ) {
        items {
          host
-          state
+          nodeState
+          metricHealth
          subCluster
          metrics {
            name
@@ -110,7 +111,7 @@
    variables: {
      cluster: cluster,
      subCluster: subCluster,
-      stateFilter: hoststateFilter,
+      stateFilter: nodeStateFilter,
      nodeFilter: hostnameFilter,
      scopes: ["core", "socket", "accelerator"],
      metrics: pendingSelectedMetrics,
@@ -164,7 +165,7 @@

  $effect(() => {
    // Update NodeListRows metrics only: Keep ordered nodes on page 1
-    hostnameFilter, hoststateFilter
+    hostnameFilter, nodeStateFilter
    // Continous Scroll: Paging if parameters change: Existing entries will not match new selections
    nodes = [];
    if (!usePaging) {
--- a/web/frontend/src/systems/NodeOverview.svelte
+++ b/web/frontend/src/systems/NodeOverview.svelte
@@ -5,8 +5,8 @@
  - `ccconfig Object?`: The ClusterCockpit Config Context [Default: null]
  - `cluster String`: The cluster to show status information for
  - `selectedMetric String?`: The selectedMetric input [Default: ""]
-  - `hostnameFilter String?`: The active hostnamefilter [Default: ""]
-  - `hostnameFilter String?`: The active hoststatefilter [Default: ""]
+  - `hostnameFilter String?`: The active hostname filter [Default: ""]
+  - `nodeStateFilter String?`: The active nodeState filter [Default: ""]
  - `from Date?`: The selected "from" date [Default: null]
  - `to Date?`: The selected "to" date [Default: null]
  - `globalMetrics [Obj]`: Includes the backend supplied availabilities for cluster and subCluster
@@ -24,7 +24,7 @@
    cluster = "",
    selectedMetric = "",
    hostnameFilter = "",
-    hoststateFilter = "",
+    nodeStateFilter = "",
    from = null,
    to = null,
    globalMetrics
@@ -55,7 +55,7 @@
          to: $to
        ) {
          host
-          state
+          nodeState
          subCluster
          metrics {
            name
@@ -91,11 +91,11 @@
  const mappedData = $derived(handleQueryData($nodesQuery?.data));
  const filteredData = $derived(mappedData.filter((h) => {
    if (hostnameFilter) {
-      if (hoststateFilter == 'all') return h.host.includes(hostnameFilter)
-      else return (h.host.includes(hostnameFilter) && h.state == hoststateFilter)
+      if (nodeStateFilter == 'all') return h.host.includes(hostnameFilter)
+      else return (h.host.includes(hostnameFilter) && h.nodeState == nodeStateFilter)
    } else {
-      if (hoststateFilter == 'all') return true
-      else return h.state == hoststateFilter
+      if (nodeStateFilter == 'all') return true
+      else return h.nodeState == nodeStateFilter
    }
  }));

@@ -116,7 +116,7 @@
    if (rawData.length > 0) {
      pendingMapped = rawData.map((h) => ({
        host: h.host,
-        state: h?.state? h.state : 'notindb',
+        nodeState: h?.nodeState || 'notindb',
        subCluster: h.subCluster,
        data: h.metrics.filter(
          (m) => m?.name == selectedMetric && m.scope == "node",
@@ -157,8 +157,8 @@
              >
            </h4>
            <span style="margin-right: 0.5rem;">
-              <Badge color={stateColors[item?.state? item.state : 'notindb']}>
-                State: {item?.state? item.state.charAt(0).toUpperCase() + item.state.slice(1) : 'Not in DB'}
+              <Badge color={stateColors[item?.nodeState || 'notindb']}>
+                State: {item?.nodeState ? item.nodeState.charAt(0).toUpperCase() + item.nodeState.slice(1) : 'Not in DB'}
              </Badge>
            </span>
          </div>
@@ -202,7 +202,7 @@
      {/each}
    {/key}
  </Row>
-{:else if hostnameFilter || hoststateFilter != 'all'}
+{:else if hostnameFilter || nodeStateFilter != 'all'}
  <Row class="mx-1">
    <Card class="px-0">
      <CardHeader>
--- a/web/frontend/src/systems/nodelist/NodeInfo.svelte
+++ b/web/frontend/src/systems/nodelist/NodeInfo.svelte
@@ -5,7 +5,8 @@
  - `cluster String`: The nodes' cluster
  - `subCluster String`: The nodes' subCluster
  - `hostname String`: The nodes' hostname
-  - `dataHealth [Bool]`: Array of Booleans depicting state of returned data per metric
+  - `nodeState String`: The nodes current state as reported by the scheduler
+  - `metricHealth String`:  The nodes current metric health as reported by the metricstore
  - `nodeJobsData [Object]`: Data returned by GQL for jobs runninig on this node [Default: null] 
 -->

@@ -32,8 +33,8 @@
    cluster,
    subCluster,
    hostname,
-    hoststate,
-    dataHealth,
+    nodeState,
+    metricHealth,
    nodeJobsData = null,
  } = $props();

@@ -50,12 +51,6 @@
  }

  /* Derived */
-  // Not at least one returned, selected metric: NodeHealth warning
-  const fetchInfo = $derived(dataHealth.includes('fetching'));
-  // Not at least one returned, selected metric: NodeHealth warning
-  const healthWarn = $derived(!dataHealth.includes(true));
-  // At least one non-returned selected metric: Metric config error?
-  const metricWarn = $derived(dataHealth.includes(false));
  const userList = $derived(nodeJobsData
    ? Array.from(new Set(nodeJobsData.jobs.items.map((j) => scrambleNames ? scramble(j.user) : j.user))).sort((a, b) => a.localeCompare(b))
    : []
@@ -86,14 +81,7 @@
    <Row cols={{xs: 1, lg: 2}}>
      <Col class="mb-2 mb-lg-0">
        <InputGroup size="sm">
-          {#if fetchInfo}
-            <InputGroupText class="flex-grow-1 flex-lg-grow-0">
-              <Icon name="arrow-clockwise" style="padding-right: 0.5rem;"/>
-            </InputGroupText>
-            <Button class="flex-grow-1" color="dark" outline disabled>
-              Fetching
-            </Button>
-          {:else if healthWarn}
+          {#if metricHealth == "failed"}
            <InputGroupText class="flex-grow-1 flex-lg-grow-0">
              <Icon name="exclamation-circle" style="padding-right: 0.5rem;"/>
              <span>Info</span>
@@ -101,13 +89,17 @@
            <Button class="flex-grow-1" color="danger" disabled>
              No Metrics
            </Button>
-          {:else if metricWarn}
+          {:else if metricHealth == "partial" || metricHealth == "unknown"}
            <InputGroupText class="flex-grow-1 flex-lg-grow-0">
              <Icon name="info-circle" style="padding-right: 0.5rem;"/>
              <span>Info</span>
            </InputGroupText>
            <Button class="flex-grow-1" color="warning" disabled>
-              Missing Metric
+              {#if metricHealth == "partial"}
+                Missing Metric(s)
+              {:else if metricHealth == "unknown"}
+                Metric Health Unknown
+              {/if}
            </Button>
          {:else if nodeJobsData.jobs.count == 1 && nodeJobsData?.jobs?.items[0]?.shared == "none"}
            <InputGroupText class="flex-grow-1 flex-lg-grow-0">
@@ -150,8 +142,8 @@
          <InputGroupText class="flex-grow-1 flex-lg-grow-0">
            State
          </InputGroupText>
-          <Button class="flex-grow-1" color={stateColors[hoststate]} disabled>
-            {hoststate.charAt(0).toUpperCase() + hoststate.slice(1)}
+          <Button class="flex-grow-1" color={stateColors[nodeState]} disabled>
+            {nodeState.charAt(0).toUpperCase() + nodeState.slice(1)}
          </Button>
        </InputGroup>
      </Col>
--- a/web/frontend/src/systems/nodelist/NodeListRow.svelte
+++ b/web/frontend/src/systems/nodelist/NodeListRow.svelte
@@ -75,7 +75,6 @@

  const extendedLegendData = $derived($nodeJobsData?.data ? buildExtendedLegend() : null);
  const refinedData = $derived(nodeData?.metrics ? sortAndSelectScope(selectedMetrics, nodeData.metrics) : []);
-  const dataHealth = $derived(refinedData.filter((rd) => rd.availability == "configured").map((enabled) => (nodeDataFetching ? 'fetching' : enabled?.data?.metric?.series?.length > 0)));

  /* Functions */
  function sortAndSelectScope(metricList = [], nodeMetrics = []) {
@@ -145,11 +144,12 @@
    {:else}
      <NodeInfo
        {cluster}
-        {dataHealth}
        nodeJobsData={$nodeJobsData.data}
        subCluster={nodeData.subCluster}
        hostname={nodeData.host}
-        hoststate={nodeData?.state? nodeData.state: 'notindb'}/>
+        nodeState={nodeData?.nodeState || 'notindb'}
+        metricHealth={nodeData?.metricHealth || 'unknown'}
+      />
    {/if}
  </td>
  {#each refinedData as metricData, i (metricData?.data?.name || i)}
@@ -174,7 +174,7 @@
            <p>No dataset(s) returned for <b>{selectedMetrics[i]}</b></p>
            <p class="mb-1">Metric or host was not found in metric store for cluster <b>{cluster}</b>.</p>
          </Card>
-        {:else if !!metricData.data?.metric.statisticsSeries}
+        {:else if !!metricData?.data?.metric?.statisticsSeries}
          <!-- "No Data"-Warning included in MetricPlot-Component -->
          <MetricPlot
            {cluster}
@@ -183,8 +183,7 @@
            scope={metricData.data.scope}
            timestep={metricData.data.metric.timestep}
            series={metricData.data.metric.series}
-            statisticsSeries={metricData.data?.metric.statisticsSeries}
-            useStatsSeries={!!metricData.data?.metric.statisticsSeries}
+            statisticsSeries={metricData.data.metric.statisticsSeries}
            height={175}
            {plotSync}
            forNode