Merge branch 'dev' into metricstore

This commit is contained in:
Jan Eitzinger
2025-09-10 09:14:50 +02:00
committed by GitHub
26 changed files with 2836 additions and 945 deletions

View File

@@ -12,12 +12,13 @@ type Node {
hostname: String!
cluster: String!
subCluster: String!
runningJobs: Int!
nodeState: NodeState!
HealthState: MonitoringState!
healthState: MonitoringState!
metaData: Any
}
type NodeStats {
type NodeStates {
state: String!
count: Int!
}
@@ -236,10 +237,12 @@ enum Aggregate {
USER
PROJECT
CLUSTER
SUBCLUSTER
}
enum SortByAggregate {
TOTALWALLTIME
TOTALJOBS
TOTALUSERS
TOTALNODES
TOTALNODEHOURS
TOTALCORES
@@ -300,9 +303,10 @@ type Query {
user(username: String!): User
allocatedNodes(cluster: String!): [Count!]!
## Node Queries New
node(id: ID!): Node
nodes(filter: [NodeFilter!], order: OrderByInput): NodeStateResultList!
nodeStats(filter: [NodeFilter!]): [NodeStats!]!
nodeStates(filter: [NodeFilter!]): [NodeStates!]!
job(id: ID!): Job
jobMetrics(
@@ -357,6 +361,7 @@ type Query {
from: Time!
to: Time!
): [NodeMetrics!]!
nodeMetricsList(
cluster: String!
subCluster: String!
@@ -393,6 +398,7 @@ type TimeRangeOutput {
input NodeFilter {
hostname: StringInput
cluster: StringInput
subcluster: StringInput
nodeState: NodeState
healthState: MonitoringState
}
@@ -497,11 +503,12 @@ type MetricHistoPoint {
}
type JobsStatistics {
id: ID! # If `groupBy` was used, ID of the user/project/cluster
id: ID! # If `groupBy` was used, ID of the user/project/cluster/subcluster
name: String! # if User-Statistics: Given Name of Account (ID) Owner
totalUsers: Int! # if *not* User-Statistics: Number of active users (based on running jobs)
totalJobs: Int! # Number of jobs
runningJobs: Int! # Number of running jobs
shortJobs: Int! # Number of jobs with a duration of less than duration
shortJobs: Int! # Number of jobs with a duration of less than config'd ShortRunningJobsDuration
totalWalltime: Int! # Sum of the duration of all matched jobs in hours
totalNodes: Int! # Sum of the nodes of all matched jobs
totalNodeHours: Int! # Sum of the node hours of all matched jobs

View File

@@ -113,6 +113,7 @@ func (api *RestApi) getJobs(rw http.ResponseWriter, r *http.Request) {
for key, vals := range r.URL.Query() {
switch key {
// TODO: add project filter
case "state":
for _, s := range vals {
state := schema.JobState(s)
@@ -125,7 +126,7 @@ func (api *RestApi) getJobs(rw http.ResponseWriter, r *http.Request) {
}
case "cluster":
filter.Cluster = &model.StringInput{Eq: &vals[0]}
case "start-time":
case "start-time": // ?startTime=1753707480-1754053139
st := strings.Split(vals[0], "-")
if len(st) != 2 {
handleError(fmt.Errorf("invalid query parameter value: startTime"),

View File

@@ -402,7 +402,7 @@ func (auth *Authentication) AuthUserApi(
return
}
case len(user.Roles) >= 2:
if user.HasRole(schema.RoleApi) && user.HasAnyRole([]schema.Role{schema.RoleUser, schema.RoleManager, schema.RoleAdmin}) {
if user.HasRole(schema.RoleApi) && user.HasAnyRole([]schema.Role{schema.RoleUser, schema.RoleManager, schema.RoleSupport, schema.RoleAdmin}) {
ctx := context.WithValue(r.Context(), repository.ContextUserKey, user)
onsuccess.ServeHTTP(rw, r.WithContext(ctx))
return
@@ -530,6 +530,7 @@ func securedCheck(user *schema.User, r *http.Request) error {
IPAddress = r.RemoteAddr
}
// FIXME: IPV6 not handled
if strings.Contains(IPAddress, ":") {
IPAddress = strings.Split(IPAddress, ":")[0]
}

View File

@@ -202,6 +202,7 @@ type ComplexityRoot struct {
TotalJobs func(childComplexity int) int
TotalNodeHours func(childComplexity int) int
TotalNodes func(childComplexity int) int
TotalUsers func(childComplexity int) int
TotalWalltime func(childComplexity int) int
}
@@ -277,6 +278,7 @@ type ComplexityRoot struct {
ID func(childComplexity int) int
MetaData func(childComplexity int) int
NodeState func(childComplexity int) int
RunningJobs func(childComplexity int) int
SubCluster func(childComplexity int) int
}
@@ -291,7 +293,7 @@ type ComplexityRoot struct {
Items func(childComplexity int) int
}
NodeStats struct {
NodeStates struct {
Count func(childComplexity int) int
State func(childComplexity int) int
}
@@ -319,7 +321,7 @@ type ComplexityRoot struct {
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, nodeFilter string, scopes []schema.MetricScope, metrics []string, from time.Time, to time.Time, page *model.PageRequest, resolution *int) int
NodeStats func(childComplexity int, filter []*model.NodeFilter) int
NodeStates func(childComplexity int, filter []*model.NodeFilter) int
Nodes func(childComplexity int, filter []*model.NodeFilter, order *model.OrderByInput) int
RooflineHeatmap func(childComplexity int, filter []*model.JobFilter, rows int, cols int, minX float64, minY float64, maxX float64, maxY float64) int
ScopedJobStats func(childComplexity int, id string, metrics []string, scopes []schema.MetricScope) int
@@ -445,6 +447,7 @@ type MutationResolver interface {
UpdateConfiguration(ctx context.Context, name string, value string) (*string, error)
}
type NodeResolver interface {
RunningJobs(ctx context.Context, obj *schema.Node) (int, error)
NodeState(ctx context.Context, obj *schema.Node) (string, error)
HealthState(ctx context.Context, obj *schema.Node) (schema.NodeState, error)
MetaData(ctx context.Context, obj *schema.Node) (any, error)
@@ -457,7 +460,7 @@ type QueryResolver interface {
AllocatedNodes(ctx context.Context, cluster string) ([]*model.Count, error)
Node(ctx context.Context, id string) (*schema.Node, error)
Nodes(ctx context.Context, filter []*model.NodeFilter, order *model.OrderByInput) (*model.NodeStateResultList, error)
NodeStats(ctx context.Context, filter []*model.NodeFilter) ([]*model.NodeStats, error)
NodeStates(ctx context.Context, filter []*model.NodeFilter) ([]*model.NodeStates, 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)
JobStats(ctx context.Context, id string, metrics []string) ([]*model.NamedStats, error)
@@ -1165,6 +1168,13 @@ func (e *executableSchema) Complexity(ctx context.Context, typeName, field strin
return e.complexity.JobsStatistics.TotalNodes(childComplexity), true
case "JobsStatistics.totalUsers":
if e.complexity.JobsStatistics.TotalUsers == nil {
break
}
return e.complexity.JobsStatistics.TotalUsers(childComplexity), true
case "JobsStatistics.totalWalltime":
if e.complexity.JobsStatistics.TotalWalltime == nil {
break
@@ -1475,7 +1485,7 @@ func (e *executableSchema) Complexity(ctx context.Context, typeName, field strin
return e.complexity.Node.Cluster(childComplexity), true
case "Node.HealthState":
case "Node.healthState":
if e.complexity.Node.HealthState == nil {
break
}
@@ -1510,6 +1520,13 @@ func (e *executableSchema) Complexity(ctx context.Context, typeName, field strin
return e.complexity.Node.NodeState(childComplexity), true
case "Node.runningJobs":
if e.complexity.Node.RunningJobs == nil {
break
}
return e.complexity.Node.RunningJobs(childComplexity), true
case "Node.subCluster":
if e.complexity.Node.SubCluster == nil {
break
@@ -1552,19 +1569,19 @@ func (e *executableSchema) Complexity(ctx context.Context, typeName, field strin
return e.complexity.NodeStateResultList.Items(childComplexity), true
case "NodeStats.count":
if e.complexity.NodeStats.Count == nil {
case "NodeStates.count":
if e.complexity.NodeStates.Count == nil {
break
}
return e.complexity.NodeStats.Count(childComplexity), true
return e.complexity.NodeStates.Count(childComplexity), true
case "NodeStats.state":
if e.complexity.NodeStats.State == nil {
case "NodeStates.state":
if e.complexity.NodeStates.State == nil {
break
}
return e.complexity.NodeStats.State(childComplexity), true
return e.complexity.NodeStates.State(childComplexity), true
case "NodesResultList.count":
if e.complexity.NodesResultList.Count == nil {
@@ -1754,17 +1771,17 @@ func (e *executableSchema) Complexity(ctx context.Context, typeName, field strin
return e.complexity.Query.NodeMetricsList(childComplexity, args["cluster"].(string), args["subCluster"].(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
case "Query.nodeStats":
if e.complexity.Query.NodeStats == nil {
case "Query.nodeStates":
if e.complexity.Query.NodeStates == nil {
break
}
args, err := ec.field_Query_nodeStats_args(ctx, rawArgs)
args, err := ec.field_Query_nodeStates_args(ctx, rawArgs)
if err != nil {
return 0, false
}
return e.complexity.Query.NodeStats(childComplexity, args["filter"].([]*model.NodeFilter)), true
return e.complexity.Query.NodeStates(childComplexity, args["filter"].([]*model.NodeFilter)), true
case "Query.nodes":
if e.complexity.Query.Nodes == nil {
@@ -2334,12 +2351,13 @@ type Node {
hostname: String!
cluster: String!
subCluster: String!
runningJobs: Int!
nodeState: NodeState!
HealthState: MonitoringState!
healthState: MonitoringState!
metaData: Any
}
type NodeStats {
type NodeStates {
state: String!
count: Int!
}
@@ -2558,10 +2576,12 @@ enum Aggregate {
USER
PROJECT
CLUSTER
SUBCLUSTER
}
enum SortByAggregate {
TOTALWALLTIME
TOTALJOBS
TOTALUSERS
TOTALNODES
TOTALNODEHOURS
TOTALCORES
@@ -2622,9 +2642,10 @@ type Query {
user(username: String!): User
allocatedNodes(cluster: String!): [Count!]!
## Node Queries New
node(id: ID!): Node
nodes(filter: [NodeFilter!], order: OrderByInput): NodeStateResultList!
nodeStats(filter: [NodeFilter!]): [NodeStats!]!
nodeStates(filter: [NodeFilter!]): [NodeStates!]!
job(id: ID!): Job
jobMetrics(
@@ -2679,6 +2700,7 @@ type Query {
from: Time!
to: Time!
): [NodeMetrics!]!
nodeMetricsList(
cluster: String!
subCluster: String!
@@ -2715,6 +2737,7 @@ type TimeRangeOutput {
input NodeFilter {
hostname: StringInput
cluster: StringInput
subcluster: StringInput
nodeState: NodeState
healthState: MonitoringState
}
@@ -2819,11 +2842,12 @@ type MetricHistoPoint {
}
type JobsStatistics {
id: ID! # If ` + "`" + `groupBy` + "`" + ` was used, ID of the user/project/cluster
id: ID! # If ` + "`" + `groupBy` + "`" + ` was used, ID of the user/project/cluster/subcluster
name: String! # if User-Statistics: Given Name of Account (ID) Owner
totalUsers: Int! # if *not* User-Statistics: Number of active users (based on running jobs)
totalJobs: Int! # Number of jobs
runningJobs: Int! # Number of running jobs
shortJobs: Int! # Number of jobs with a duration of less than duration
shortJobs: Int! # Number of jobs with a duration of less than config'd ShortRunningJobsDuration
totalWalltime: Int! # Sum of the duration of all matched jobs in hours
totalNodes: Int! # Sum of the nodes of all matched jobs
totalNodeHours: Int! # Sum of the node hours of all matched jobs
@@ -3197,7 +3221,7 @@ func (ec *executionContext) field_Query_nodeMetrics_args(ctx context.Context, ra
return args, nil
}
func (ec *executionContext) field_Query_nodeStats_args(ctx context.Context, rawArgs map[string]any) (map[string]any, error) {
func (ec *executionContext) field_Query_nodeStates_args(ctx context.Context, rawArgs map[string]any) (map[string]any, error) {
var err error
args := map[string]any{}
arg0, err := graphql.ProcessArgField(ctx, rawArgs, "filter", ec.unmarshalONodeFilter2ᚕᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodeFilterᚄ)
@@ -7125,6 +7149,50 @@ func (ec *executionContext) fieldContext_JobsStatistics_name(_ context.Context,
return fc, nil
}
func (ec *executionContext) _JobsStatistics_totalUsers(ctx context.Context, field graphql.CollectedField, obj *model.JobsStatistics) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_JobsStatistics_totalUsers(ctx, field)
if err != nil {
return graphql.Null
}
ctx = graphql.WithFieldContext(ctx, fc)
defer func() {
if r := recover(); r != nil {
ec.Error(ctx, ec.Recover(ctx, r))
ret = graphql.Null
}
}()
resTmp, err := ec.ResolverMiddleware(ctx, func(rctx context.Context) (any, error) {
ctx = rctx // use context from middleware stack in children
return obj.TotalUsers, nil
})
if err != nil {
ec.Error(ctx, err)
return graphql.Null
}
if resTmp == nil {
if !graphql.HasFieldError(ctx, fc) {
ec.Errorf(ctx, "must not be null")
}
return graphql.Null
}
res := resTmp.(int)
fc.Result = res
return ec.marshalNInt2int(ctx, field.Selections, res)
}
func (ec *executionContext) fieldContext_JobsStatistics_totalUsers(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
fc = &graphql.FieldContext{
Object: "JobsStatistics",
Field: field,
IsMethod: false,
IsResolver: false,
Child: func(ctx context.Context, field graphql.CollectedField) (*graphql.FieldContext, error) {
return nil, errors.New("field of type Int does not have child fields")
},
}
return fc, nil
}
func (ec *executionContext) _JobsStatistics_totalJobs(ctx context.Context, field graphql.CollectedField, obj *model.JobsStatistics) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_JobsStatistics_totalJobs(ctx, field)
if err != nil {
@@ -9788,6 +9856,50 @@ func (ec *executionContext) fieldContext_Node_subCluster(_ context.Context, fiel
return fc, nil
}
func (ec *executionContext) _Node_runningJobs(ctx context.Context, field graphql.CollectedField, obj *schema.Node) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_Node_runningJobs(ctx, field)
if err != nil {
return graphql.Null
}
ctx = graphql.WithFieldContext(ctx, fc)
defer func() {
if r := recover(); r != nil {
ec.Error(ctx, ec.Recover(ctx, r))
ret = graphql.Null
}
}()
resTmp, err := ec.ResolverMiddleware(ctx, func(rctx context.Context) (any, error) {
ctx = rctx // use context from middleware stack in children
return ec.resolvers.Node().RunningJobs(rctx, obj)
})
if err != nil {
ec.Error(ctx, err)
return graphql.Null
}
if resTmp == nil {
if !graphql.HasFieldError(ctx, fc) {
ec.Errorf(ctx, "must not be null")
}
return graphql.Null
}
res := resTmp.(int)
fc.Result = res
return ec.marshalNInt2int(ctx, field.Selections, res)
}
func (ec *executionContext) fieldContext_Node_runningJobs(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
fc = &graphql.FieldContext{
Object: "Node",
Field: field,
IsMethod: true,
IsResolver: true,
Child: func(ctx context.Context, field graphql.CollectedField) (*graphql.FieldContext, error) {
return nil, errors.New("field of type Int does not have child fields")
},
}
return fc, nil
}
func (ec *executionContext) _Node_nodeState(ctx context.Context, field graphql.CollectedField, obj *schema.Node) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_Node_nodeState(ctx, field)
if err != nil {
@@ -9832,8 +9944,8 @@ func (ec *executionContext) fieldContext_Node_nodeState(_ context.Context, field
return fc, nil
}
func (ec *executionContext) _Node_HealthState(ctx context.Context, field graphql.CollectedField, obj *schema.Node) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_Node_HealthState(ctx, field)
func (ec *executionContext) _Node_healthState(ctx context.Context, field graphql.CollectedField, obj *schema.Node) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_Node_healthState(ctx, field)
if err != nil {
return graphql.Null
}
@@ -9863,7 +9975,7 @@ func (ec *executionContext) _Node_HealthState(ctx context.Context, field graphql
return ec.marshalNMonitoringState2githubᚗcomᚋClusterCockpitᚋccᚑlibᚋschemaᚐNodeState(ctx, field.Selections, res)
}
func (ec *executionContext) fieldContext_Node_HealthState(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
func (ec *executionContext) fieldContext_Node_healthState(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
fc = &graphql.FieldContext{
Object: "Node",
Field: field,
@@ -10104,10 +10216,12 @@ func (ec *executionContext) fieldContext_NodeStateResultList_items(_ context.Con
return ec.fieldContext_Node_cluster(ctx, field)
case "subCluster":
return ec.fieldContext_Node_subCluster(ctx, field)
case "runningJobs":
return ec.fieldContext_Node_runningJobs(ctx, field)
case "nodeState":
return ec.fieldContext_Node_nodeState(ctx, field)
case "HealthState":
return ec.fieldContext_Node_HealthState(ctx, field)
case "healthState":
return ec.fieldContext_Node_healthState(ctx, field)
case "metaData":
return ec.fieldContext_Node_metaData(ctx, field)
}
@@ -10158,8 +10272,8 @@ func (ec *executionContext) fieldContext_NodeStateResultList_count(_ context.Con
return fc, nil
}
func (ec *executionContext) _NodeStats_state(ctx context.Context, field graphql.CollectedField, obj *model.NodeStats) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_NodeStats_state(ctx, field)
func (ec *executionContext) _NodeStates_state(ctx context.Context, field graphql.CollectedField, obj *model.NodeStates) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_NodeStates_state(ctx, field)
if err != nil {
return graphql.Null
}
@@ -10189,9 +10303,9 @@ func (ec *executionContext) _NodeStats_state(ctx context.Context, field graphql.
return ec.marshalNString2string(ctx, field.Selections, res)
}
func (ec *executionContext) fieldContext_NodeStats_state(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
func (ec *executionContext) fieldContext_NodeStates_state(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
fc = &graphql.FieldContext{
Object: "NodeStats",
Object: "NodeStates",
Field: field,
IsMethod: false,
IsResolver: false,
@@ -10202,8 +10316,8 @@ func (ec *executionContext) fieldContext_NodeStats_state(_ context.Context, fiel
return fc, nil
}
func (ec *executionContext) _NodeStats_count(ctx context.Context, field graphql.CollectedField, obj *model.NodeStats) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_NodeStats_count(ctx, field)
func (ec *executionContext) _NodeStates_count(ctx context.Context, field graphql.CollectedField, obj *model.NodeStates) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_NodeStates_count(ctx, field)
if err != nil {
return graphql.Null
}
@@ -10233,9 +10347,9 @@ func (ec *executionContext) _NodeStats_count(ctx context.Context, field graphql.
return ec.marshalNInt2int(ctx, field.Selections, res)
}
func (ec *executionContext) fieldContext_NodeStats_count(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
func (ec *executionContext) fieldContext_NodeStates_count(_ context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
fc = &graphql.FieldContext{
Object: "NodeStats",
Object: "NodeStates",
Field: field,
IsMethod: false,
IsResolver: false,
@@ -10830,10 +10944,12 @@ func (ec *executionContext) fieldContext_Query_node(ctx context.Context, field g
return ec.fieldContext_Node_cluster(ctx, field)
case "subCluster":
return ec.fieldContext_Node_subCluster(ctx, field)
case "runningJobs":
return ec.fieldContext_Node_runningJobs(ctx, field)
case "nodeState":
return ec.fieldContext_Node_nodeState(ctx, field)
case "HealthState":
return ec.fieldContext_Node_HealthState(ctx, field)
case "healthState":
return ec.fieldContext_Node_healthState(ctx, field)
case "metaData":
return ec.fieldContext_Node_metaData(ctx, field)
}
@@ -10915,8 +11031,8 @@ func (ec *executionContext) fieldContext_Query_nodes(ctx context.Context, field
return fc, nil
}
func (ec *executionContext) _Query_nodeStats(ctx context.Context, field graphql.CollectedField) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_Query_nodeStats(ctx, field)
func (ec *executionContext) _Query_nodeStates(ctx context.Context, field graphql.CollectedField) (ret graphql.Marshaler) {
fc, err := ec.fieldContext_Query_nodeStates(ctx, field)
if err != nil {
return graphql.Null
}
@@ -10929,7 +11045,7 @@ func (ec *executionContext) _Query_nodeStats(ctx context.Context, field graphql.
}()
resTmp, err := ec.ResolverMiddleware(ctx, func(rctx context.Context) (any, error) {
ctx = rctx // use context from middleware stack in children
return ec.resolvers.Query().NodeStats(rctx, fc.Args["filter"].([]*model.NodeFilter))
return ec.resolvers.Query().NodeStates(rctx, fc.Args["filter"].([]*model.NodeFilter))
})
if err != nil {
ec.Error(ctx, err)
@@ -10941,12 +11057,12 @@ func (ec *executionContext) _Query_nodeStats(ctx context.Context, field graphql.
}
return graphql.Null
}
res := resTmp.([]*model.NodeStats)
res := resTmp.([]*model.NodeStates)
fc.Result = res
return ec.marshalNNodeStats2ᚕᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodeStatsᚄ(ctx, field.Selections, res)
return ec.marshalNNodeStates2ᚕᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodeStatesᚄ(ctx, field.Selections, res)
}
func (ec *executionContext) fieldContext_Query_nodeStats(ctx context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
func (ec *executionContext) fieldContext_Query_nodeStates(ctx context.Context, field graphql.CollectedField) (fc *graphql.FieldContext, err error) {
fc = &graphql.FieldContext{
Object: "Query",
Field: field,
@@ -10955,11 +11071,11 @@ func (ec *executionContext) fieldContext_Query_nodeStats(ctx context.Context, fi
Child: func(ctx context.Context, field graphql.CollectedField) (*graphql.FieldContext, error) {
switch field.Name {
case "state":
return ec.fieldContext_NodeStats_state(ctx, field)
return ec.fieldContext_NodeStates_state(ctx, field)
case "count":
return ec.fieldContext_NodeStats_count(ctx, field)
return ec.fieldContext_NodeStates_count(ctx, field)
}
return nil, fmt.Errorf("no field named %q was found under type NodeStats", field.Name)
return nil, fmt.Errorf("no field named %q was found under type NodeStates", field.Name)
},
}
defer func() {
@@ -10969,7 +11085,7 @@ func (ec *executionContext) fieldContext_Query_nodeStats(ctx context.Context, fi
}
}()
ctx = graphql.WithFieldContext(ctx, fc)
if fc.Args, err = ec.field_Query_nodeStats_args(ctx, field.ArgumentMap(ec.Variables)); err != nil {
if fc.Args, err = ec.field_Query_nodeStates_args(ctx, field.ArgumentMap(ec.Variables)); err != nil {
ec.Error(ctx, err)
return fc, err
}
@@ -11379,6 +11495,8 @@ func (ec *executionContext) fieldContext_Query_jobsStatistics(ctx context.Contex
return ec.fieldContext_JobsStatistics_id(ctx, field)
case "name":
return ec.fieldContext_JobsStatistics_name(ctx, field)
case "totalUsers":
return ec.fieldContext_JobsStatistics_totalUsers(ctx, field)
case "totalJobs":
return ec.fieldContext_JobsStatistics_totalJobs(ctx, field)
case "runningJobs":
@@ -16549,7 +16667,7 @@ func (ec *executionContext) unmarshalInputNodeFilter(ctx context.Context, obj an
asMap[k] = v
}
fieldsInOrder := [...]string{"hostname", "cluster", "nodeState", "healthState"}
fieldsInOrder := [...]string{"hostname", "cluster", "subcluster", "nodeState", "healthState"}
for _, k := range fieldsInOrder {
v, ok := asMap[k]
if !ok {
@@ -16570,6 +16688,13 @@ func (ec *executionContext) unmarshalInputNodeFilter(ctx context.Context, obj an
return it, err
}
it.Cluster = data
case "subcluster":
ctx := graphql.WithPathContext(ctx, graphql.NewPathWithField("subcluster"))
data, err := ec.unmarshalOStringInput2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐStringInput(ctx, v)
if err != nil {
return it, err
}
it.Subcluster = data
case "nodeState":
ctx := graphql.WithPathContext(ctx, graphql.NewPathWithField("nodeState"))
data, err := ec.unmarshalONodeState2ᚖstring(ctx, v)
@@ -17976,6 +18101,11 @@ func (ec *executionContext) _JobsStatistics(ctx context.Context, sel ast.Selecti
if out.Values[i] == graphql.Null {
out.Invalids++
}
case "totalUsers":
out.Values[i] = ec._JobsStatistics_totalUsers(ctx, field, obj)
if out.Values[i] == graphql.Null {
out.Invalids++
}
case "totalJobs":
out.Values[i] = ec._JobsStatistics_totalJobs(ctx, field, obj)
if out.Values[i] == graphql.Null {
@@ -18625,6 +18755,42 @@ func (ec *executionContext) _Node(ctx context.Context, sel ast.SelectionSet, obj
if out.Values[i] == graphql.Null {
atomic.AddUint32(&out.Invalids, 1)
}
case "runningJobs":
field := field
innerFunc := func(ctx context.Context, fs *graphql.FieldSet) (res graphql.Marshaler) {
defer func() {
if r := recover(); r != nil {
ec.Error(ctx, ec.Recover(ctx, r))
}
}()
res = ec._Node_runningJobs(ctx, field, obj)
if res == graphql.Null {
atomic.AddUint32(&fs.Invalids, 1)
}
return res
}
if field.Deferrable != nil {
dfs, ok := deferred[field.Deferrable.Label]
di := 0
if ok {
dfs.AddField(field)
di = len(dfs.Values) - 1
} else {
dfs = graphql.NewFieldSet([]graphql.CollectedField{field})
deferred[field.Deferrable.Label] = dfs
}
dfs.Concurrently(di, func(ctx context.Context) graphql.Marshaler {
return innerFunc(ctx, dfs)
})
// don't run the out.Concurrently() call below
out.Values[i] = graphql.Null
continue
}
out.Concurrently(i, func(ctx context.Context) graphql.Marshaler { return innerFunc(ctx, out) })
case "nodeState":
field := field
@@ -18661,7 +18827,7 @@ func (ec *executionContext) _Node(ctx context.Context, sel ast.SelectionSet, obj
}
out.Concurrently(i, func(ctx context.Context) graphql.Marshaler { return innerFunc(ctx, out) })
case "HealthState":
case "healthState":
field := field
innerFunc := func(ctx context.Context, fs *graphql.FieldSet) (res graphql.Marshaler) {
@@ -18670,7 +18836,7 @@ func (ec *executionContext) _Node(ctx context.Context, sel ast.SelectionSet, obj
ec.Error(ctx, ec.Recover(ctx, r))
}
}()
res = ec._Node_HealthState(ctx, field, obj)
res = ec._Node_healthState(ctx, field, obj)
if res == graphql.Null {
atomic.AddUint32(&fs.Invalids, 1)
}
@@ -18843,24 +19009,24 @@ func (ec *executionContext) _NodeStateResultList(ctx context.Context, sel ast.Se
return out
}
var nodeStatsImplementors = []string{"NodeStats"}
var nodeStatesImplementors = []string{"NodeStates"}
func (ec *executionContext) _NodeStats(ctx context.Context, sel ast.SelectionSet, obj *model.NodeStats) graphql.Marshaler {
fields := graphql.CollectFields(ec.OperationContext, sel, nodeStatsImplementors)
func (ec *executionContext) _NodeStates(ctx context.Context, sel ast.SelectionSet, obj *model.NodeStates) graphql.Marshaler {
fields := graphql.CollectFields(ec.OperationContext, sel, nodeStatesImplementors)
out := graphql.NewFieldSet(fields)
deferred := make(map[string]*graphql.FieldSet)
for i, field := range fields {
switch field.Name {
case "__typename":
out.Values[i] = graphql.MarshalString("NodeStats")
out.Values[i] = graphql.MarshalString("NodeStates")
case "state":
out.Values[i] = ec._NodeStats_state(ctx, field, obj)
out.Values[i] = ec._NodeStates_state(ctx, field, obj)
if out.Values[i] == graphql.Null {
out.Invalids++
}
case "count":
out.Values[i] = ec._NodeStats_count(ctx, field, obj)
out.Values[i] = ec._NodeStates_count(ctx, field, obj)
if out.Values[i] == graphql.Null {
out.Invalids++
}
@@ -19103,7 +19269,7 @@ func (ec *executionContext) _Query(ctx context.Context, sel ast.SelectionSet) gr
}
out.Concurrently(i, func(ctx context.Context) graphql.Marshaler { return rrm(innerCtx) })
case "nodeStats":
case "nodeStates":
field := field
innerFunc := func(ctx context.Context, fs *graphql.FieldSet) (res graphql.Marshaler) {
@@ -19112,7 +19278,7 @@ func (ec *executionContext) _Query(ctx context.Context, sel ast.SelectionSet) gr
ec.Error(ctx, ec.Recover(ctx, r))
}
}()
res = ec._Query_nodeStats(ctx, field)
res = ec._Query_nodeStates(ctx, field)
if res == graphql.Null {
atomic.AddUint32(&fs.Invalids, 1)
}
@@ -21756,7 +21922,7 @@ func (ec *executionContext) marshalNNodeStateResultList2ᚖgithubᚗcomᚋCluste
return ec._NodeStateResultList(ctx, sel, v)
}
func (ec *executionContext) marshalNNodeStats2ᚕᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodeStatsᚄ(ctx context.Context, sel ast.SelectionSet, v []*model.NodeStats) graphql.Marshaler {
func (ec *executionContext) marshalNNodeStates2ᚕᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodeStatesᚄ(ctx context.Context, sel ast.SelectionSet, v []*model.NodeStates) graphql.Marshaler {
ret := make(graphql.Array, len(v))
var wg sync.WaitGroup
isLen1 := len(v) == 1
@@ -21780,7 +21946,7 @@ func (ec *executionContext) marshalNNodeStats2ᚕᚖgithubᚗcomᚋClusterCockpi
if !isLen1 {
defer wg.Done()
}
ret[i] = ec.marshalNNodeStats2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodeStats(ctx, sel, v[i])
ret[i] = ec.marshalNNodeStates2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodeStates(ctx, sel, v[i])
}
if isLen1 {
f(i)
@@ -21800,14 +21966,14 @@ func (ec *executionContext) marshalNNodeStats2ᚕᚖgithubᚗcomᚋClusterCockpi
return ret
}
func (ec *executionContext) marshalNNodeStats2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodeStats(ctx context.Context, sel ast.SelectionSet, v *model.NodeStats) graphql.Marshaler {
func (ec *executionContext) marshalNNodeStates2ᚖgithubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodeStates(ctx context.Context, sel ast.SelectionSet, v *model.NodeStates) graphql.Marshaler {
if v == nil {
if !graphql.HasFieldError(ctx, graphql.GetFieldContext(ctx)) {
ec.Errorf(ctx, "the requested element is null which the schema does not allow")
}
return graphql.Null
}
return ec._NodeStats(ctx, sel, v)
return ec._NodeStates(ctx, sel, v)
}
func (ec *executionContext) marshalNNodesResultList2githubᚗcomᚋClusterCockpitᚋccᚑbackendᚋinternalᚋgraphᚋmodelᚐNodesResultList(ctx context.Context, sel ast.SelectionSet, v model.NodesResultList) graphql.Marshaler {

View File

@@ -114,6 +114,7 @@ type JobStats struct {
type JobsStatistics struct {
ID string `json:"id"`
Name string `json:"name"`
TotalUsers int `json:"totalUsers"`
TotalJobs int `json:"totalJobs"`
RunningJobs int `json:"runningJobs"`
ShortJobs int `json:"shortJobs"`
@@ -172,6 +173,7 @@ type NamedStatsWithScope struct {
type NodeFilter struct {
Hostname *StringInput `json:"hostname,omitempty"`
Cluster *StringInput `json:"cluster,omitempty"`
Subcluster *StringInput `json:"subcluster,omitempty"`
NodeState *string `json:"nodeState,omitempty"`
HealthState *schema.NodeState `json:"healthState,omitempty"`
}
@@ -187,7 +189,7 @@ type NodeStateResultList struct {
Count *int `json:"count,omitempty"`
}
type NodeStats struct {
type NodeStates struct {
State string `json:"state"`
Count int `json:"count"`
}
@@ -248,20 +250,22 @@ type User struct {
type Aggregate string
const (
AggregateUser Aggregate = "USER"
AggregateProject Aggregate = "PROJECT"
AggregateCluster Aggregate = "CLUSTER"
AggregateUser Aggregate = "USER"
AggregateProject Aggregate = "PROJECT"
AggregateCluster Aggregate = "CLUSTER"
AggregateSubcluster Aggregate = "SUBCLUSTER"
)
var AllAggregate = []Aggregate{
AggregateUser,
AggregateProject,
AggregateCluster,
AggregateSubcluster,
}
func (e Aggregate) IsValid() bool {
switch e {
case AggregateUser, AggregateProject, AggregateCluster:
case AggregateUser, AggregateProject, AggregateCluster, AggregateSubcluster:
return true
}
return false
@@ -307,6 +311,7 @@ type SortByAggregate string
const (
SortByAggregateTotalwalltime SortByAggregate = "TOTALWALLTIME"
SortByAggregateTotaljobs SortByAggregate = "TOTALJOBS"
SortByAggregateTotalusers SortByAggregate = "TOTALUSERS"
SortByAggregateTotalnodes SortByAggregate = "TOTALNODES"
SortByAggregateTotalnodehours SortByAggregate = "TOTALNODEHOURS"
SortByAggregateTotalcores SortByAggregate = "TOTALCORES"
@@ -318,6 +323,7 @@ const (
var AllSortByAggregate = []SortByAggregate{
SortByAggregateTotalwalltime,
SortByAggregateTotaljobs,
SortByAggregateTotalusers,
SortByAggregateTotalnodes,
SortByAggregateTotalnodehours,
SortByAggregateTotalcores,
@@ -328,7 +334,7 @@ var AllSortByAggregate = []SortByAggregate{
func (e SortByAggregate) IsValid() bool {
switch e {
case SortByAggregateTotalwalltime, SortByAggregateTotaljobs, SortByAggregateTotalnodes, SortByAggregateTotalnodehours, SortByAggregateTotalcores, SortByAggregateTotalcorehours, SortByAggregateTotalaccs, SortByAggregateTotalacchours:
case SortByAggregateTotalwalltime, SortByAggregateTotaljobs, SortByAggregateTotalusers, SortByAggregateTotalnodes, SortByAggregateTotalnodehours, SortByAggregateTotalcores, SortByAggregateTotalcorehours, SortByAggregateTotalaccs, SortByAggregateTotalacchours:
return true
}
return false

View File

@@ -305,14 +305,20 @@ func (r *mutationResolver) UpdateConfiguration(ctx context.Context, name string,
return nil, nil
}
// NodeState is the resolver for the nodeState field.
func (r *nodeResolver) NodeState(ctx context.Context, obj *schema.Node) (string, error) {
panic(fmt.Errorf("not implemented: NodeState - nodeState"))
// RunningJobs is the resolver for the runningJobs field.
func (r *nodeResolver) RunningJobs(ctx context.Context, obj *schema.Node) (int, error) {
panic(fmt.Errorf("not implemented: RunningJobs - runningJobs"))
}
// HealthState is the resolver for the HealthState field.
// NodeState is the resolver for the nodeState field.
func (r *nodeResolver) NodeState(ctx context.Context, obj *schema.Node) (string, error) {
return string(obj.NodeState), nil
}
// HealthState is the resolver for the healthState field.
func (r *nodeResolver) HealthState(ctx context.Context, obj *schema.Node) (schema.NodeState, error) {
panic(fmt.Errorf("not implemented: HealthState - HealthState"))
// FIXME: Why is Output of schema.NodeState Type?
panic(fmt.Errorf("not implemented: HealthState - healthState"))
}
// MetaData is the resolver for the metaData field.
@@ -378,9 +384,26 @@ func (r *queryResolver) Nodes(ctx context.Context, filter []*model.NodeFilter, o
return &model.NodeStateResultList{Items: nodes, Count: &count}, err
}
// NodeStats is the resolver for the nodeStats field.
func (r *queryResolver) NodeStats(ctx context.Context, filter []*model.NodeFilter) ([]*model.NodeStats, error) {
panic(fmt.Errorf("not implemented: NodeStats - nodeStats"))
// NodeStates is the resolver for the nodeStates field.
func (r *queryResolver) NodeStates(ctx context.Context, filter []*model.NodeFilter) ([]*model.NodeStates, error) {
repo := repository.GetNodeRepository()
stateCounts, serr := repo.CountNodeStates(ctx, filter)
if serr != nil {
cclog.Warnf("Error while counting nodeStates: %s", serr.Error())
return nil, serr
}
healthCounts, herr := repo.CountHealthStates(ctx, filter)
if herr != nil {
cclog.Warnf("Error while counting healthStates: %s", herr.Error())
return nil, herr
}
allCounts := make([]*model.NodeStates, 0)
allCounts = append(stateCounts, healthCounts...)
return allCounts, nil
}
// Job is the resolver for the job field.
@@ -558,7 +581,7 @@ func (r *queryResolver) JobsStatistics(ctx context.Context, filter []*model.JobF
defaultDurationBins := "1h"
defaultMetricBins := 10
if requireField(ctx, "totalJobs") || requireField(ctx, "totalWalltime") || requireField(ctx, "totalNodes") || requireField(ctx, "totalCores") ||
if requireField(ctx, "totalJobs") || requireField(ctx, "totalUsers") || requireField(ctx, "totalWalltime") || requireField(ctx, "totalNodes") || requireField(ctx, "totalCores") ||
requireField(ctx, "totalAccs") || requireField(ctx, "totalNodeHours") || requireField(ctx, "totalCoreHours") || requireField(ctx, "totalAccHours") {
if groupBy == nil {
stats, err = r.Repo.JobsStats(ctx, filter)

View File

@@ -40,7 +40,7 @@ func InitDB() error {
}
tags := make(map[string]int64)
// Not using log.Print because we want the line to end with `\r` and
// Not using cclog.Print because we want the line to end with `\r` and
// this function is only ever called when a special command line flag
// is passed anyways.
fmt.Printf("%d jobs inserted...\r", 0)

View File

@@ -337,10 +337,10 @@ func (r *JobRepository) FindColumnValue(user *schema.User, searchterm string, ta
// theSql, args, theErr := theQuery.ToSql()
// if theErr != nil {
// log.Warn("Error while converting query to sql")
// cclog.Warn("Error while converting query to sql")
// return "", err
// }
// log.Debugf("SQL query (FindColumnValue): `%s`, args: %#v", theSql, args)
// cclog.Debugf("SQL query (FindColumnValue): `%s`, args: %#v", theSql, args)
err := theQuery.RunWith(r.stmtCache).QueryRow().Scan(&result)

View File

@@ -4,12 +4,13 @@ CREATE TABLE "node" (
hostname VARCHAR(255) NOT NULL,
cluster VARCHAR(255) NOT NULL,
subcluster VARCHAR(255) NOT NULL,
cpus_allocated INTEGER NOT NULL,
cpus_total INTEGER NOT NULL,
memory_allocated INTEGER NOT NULL,
memory_total INTEGER NOT NULL,
gpus_allocated INTEGER NOT NULL,
gpus_total INTEGER NOT NULL,
jobs_running INTEGER DEFAULT 0 NOT NULL,
cpus_allocated INTEGER DEFAULT 0 NOT NULL,
cpus_total INTEGER DEFAULT 0 NOT NULL,
memory_allocated INTEGER DEFAULT 0 NOT NULL,
memory_total INTEGER DEFAULT 0 NOT NULL,
gpus_allocated INTEGER DEFAULT 0 NOT NULL,
gpus_total INTEGER DEFAULT 0 NOT NULL,
node_state VARCHAR(255) NOT NULL
CHECK (node_state IN (
'allocated', 'reserved', 'idle', 'mixed',

View File

@@ -49,6 +49,12 @@ func GetNodeRepository() *NodeRepository {
return nodeRepoInstance
}
var nodeColumns []string = []string{
// "node.id,"
"node.hostname", "node.cluster", "node.subcluster",
"node.node_state", "node.health_state", // "node.meta_data",
}
func (r *NodeRepository) FetchMetadata(node *schema.Node) (map[string]string, error) {
start := time.Now()
cachekey := fmt.Sprintf("metadata:%d", node.ID)
@@ -218,9 +224,9 @@ func (r *NodeRepository) DeleteNode(id int64) error {
func (r *NodeRepository) QueryNodes(
ctx context.Context,
filters []*model.NodeFilter,
order *model.OrderByInput,
order *model.OrderByInput, // Currently unused!
) ([]*schema.Node, error) {
query, qerr := SecurityCheck(ctx, sq.Select(jobColumns...).From("node"))
query, qerr := AccessCheck(ctx, sq.Select(nodeColumns...).From("node"))
if qerr != nil {
return nil, qerr
}
@@ -232,6 +238,9 @@ func (r *NodeRepository) QueryNodes(
if f.Cluster != nil {
query = buildStringCondition("node.cluster", f.Cluster, query)
}
if f.Subcluster != nil {
query = buildStringCondition("node.subcluster", f.Subcluster, query)
}
if f.NodeState != nil {
query = query.Where("node.node_state = ?", f.NodeState)
}
@@ -287,3 +296,123 @@ func (r *NodeRepository) ListNodes(cluster string) ([]*schema.Node, error) {
return nodeList, nil
}
func (r *NodeRepository) CountNodeStates(ctx context.Context, filters []*model.NodeFilter) ([]*model.NodeStates, error) {
query, qerr := AccessCheck(ctx, sq.Select("node_state AS state", "count(*) AS count").From("node"))
if qerr != nil {
return nil, qerr
}
for _, f := range filters {
if f.Hostname != nil {
query = buildStringCondition("node.hostname", f.Hostname, query)
}
if f.Cluster != nil {
query = buildStringCondition("node.cluster", f.Cluster, query)
}
if f.Subcluster != nil {
query = buildStringCondition("node.subcluster", f.Subcluster, query)
}
if f.NodeState != nil {
query = query.Where("node.node_state = ?", f.NodeState)
}
if f.HealthState != nil {
query = query.Where("node.health_state = ?", f.HealthState)
}
}
// Add Group and Order
query = query.GroupBy("state").OrderBy("count DESC")
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
queryString, queryVars, _ := query.ToSql()
cclog.Errorf("Error while running query '%s' %v: %v", queryString, queryVars, err)
return nil, err
}
nodes := make([]*model.NodeStates, 0)
for rows.Next() {
node := model.NodeStates{}
if err := rows.Scan(&node.State, &node.Count); err != nil {
rows.Close()
cclog.Warn("Error while scanning rows (NodeStates)")
return nil, err
}
nodes = append(nodes, &node)
}
return nodes, nil
}
func (r *NodeRepository) CountHealthStates(ctx context.Context, filters []*model.NodeFilter) ([]*model.NodeStates, error) {
query, qerr := AccessCheck(ctx, sq.Select("health_state AS state", "count(*) AS count").From("node"))
if qerr != nil {
return nil, qerr
}
for _, f := range filters {
if f.Hostname != nil {
query = buildStringCondition("node.hostname", f.Hostname, query)
}
if f.Cluster != nil {
query = buildStringCondition("node.cluster", f.Cluster, query)
}
if f.Subcluster != nil {
query = buildStringCondition("node.subcluster", f.Subcluster, query)
}
if f.NodeState != nil {
query = query.Where("node.node_state = ?", f.NodeState)
}
if f.HealthState != nil {
query = query.Where("node.health_state = ?", f.HealthState)
}
}
// Add Group and Order
query = query.GroupBy("state").OrderBy("count DESC")
rows, err := query.RunWith(r.stmtCache).Query()
if err != nil {
queryString, queryVars, _ := query.ToSql()
cclog.Errorf("Error while running query '%s' %v: %v", queryString, queryVars, err)
return nil, err
}
nodes := make([]*model.NodeStates, 0)
for rows.Next() {
node := model.NodeStates{}
if err := rows.Scan(&node.State, &node.Count); err != nil {
rows.Close()
cclog.Warn("Error while scanning rows (NodeStates)")
return nil, err
}
nodes = append(nodes, &node)
}
return nodes, nil
}
func AccessCheck(ctx context.Context, query sq.SelectBuilder) (sq.SelectBuilder, error) {
user := GetUserFromContext(ctx)
return AccessCheckWithUser(user, query)
}
func AccessCheckWithUser(user *schema.User, query sq.SelectBuilder) (sq.SelectBuilder, error) {
if user == nil {
var qnil sq.SelectBuilder
return qnil, fmt.Errorf("user context is nil")
}
switch {
// case len(user.Roles) == 1 && user.HasRole(schema.RoleApi): // API-User : Access NodeInfos
// return query, nil
case user.HasAnyRole([]schema.Role{schema.RoleAdmin, schema.RoleSupport}): // Admin & Support : Access NodeInfos
return query, nil
default: // No known Role: No Access, return error
var qnil sq.SelectBuilder
return qnil, fmt.Errorf("user has no or unknown roles")
}
}

View File

@@ -28,6 +28,7 @@ var groupBy2column = map[model.Aggregate]string{
var sortBy2column = map[model.SortByAggregate]string{
model.SortByAggregateTotaljobs: "totalJobs",
model.SortByAggregateTotalusers: "totalUsers",
model.SortByAggregateTotalwalltime: "totalWalltime",
model.SortByAggregateTotalnodes: "totalNodes",
model.SortByAggregateTotalnodehours: "totalNodeHours",
@@ -76,8 +77,12 @@ func (r *JobRepository) buildStatsQuery(
// fmt.Sprintf(`CAST(ROUND((CASE WHEN job.job_state = "running" THEN %d - job.start_time ELSE job.duration END) / 3600) as %s) as value`, time.Now().Unix(), castType)
if col != "" {
// Scan columns: id, totalJobs, name, totalWalltime, totalNodes, totalNodeHours, totalCores, totalCoreHours, totalAccs, totalAccHours
query = sq.Select(col, "COUNT(job.id) as totalJobs", "name",
// Scan columns: id, name, totalJobs, totalUsers, totalWalltime, totalNodes, totalNodeHours, totalCores, totalCoreHours, totalAccs, totalAccHours
query = sq.Select(
col,
"name",
"COUNT(job.id) as totalJobs",
"COUNT(DISTINCT job.hpc_user) AS totalUsers",
fmt.Sprintf(`CAST(ROUND(SUM((CASE WHEN job.job_state = "running" THEN %d - job.start_time ELSE job.duration END)) / 3600) as %s) as totalWalltime`, time.Now().Unix(), castType),
fmt.Sprintf(`CAST(SUM(job.num_nodes) as %s) as totalNodes`, castType),
fmt.Sprintf(`CAST(ROUND(SUM((CASE WHEN job.job_state = "running" THEN %d - job.start_time ELSE job.duration END) * job.num_nodes) / 3600) as %s) as totalNodeHours`, time.Now().Unix(), castType),
@@ -87,8 +92,10 @@ func (r *JobRepository) buildStatsQuery(
fmt.Sprintf(`CAST(ROUND(SUM((CASE WHEN job.job_state = "running" THEN %d - job.start_time ELSE job.duration END) * job.num_acc) / 3600) as %s) as totalAccHours`, time.Now().Unix(), castType),
).From("job").LeftJoin("hpc_user ON hpc_user.username = job.hpc_user").GroupBy(col)
} else {
// Scan columns: totalJobs, name, totalWalltime, totalNodes, totalNodeHours, totalCores, totalCoreHours, totalAccs, totalAccHours
query = sq.Select("COUNT(job.id)",
// Scan columns: totalJobs, totalUsers, totalWalltime, totalNodes, totalNodeHours, totalCores, totalCoreHours, totalAccs, totalAccHours
query = sq.Select(
"COUNT(job.id) as totalJobs",
"COUNT(DISTINCT job.hpc_user) AS totalUsers",
fmt.Sprintf(`CAST(ROUND(SUM((CASE WHEN job.job_state = "running" THEN %d - job.start_time ELSE job.duration END)) / 3600) as %s)`, time.Now().Unix(), castType),
fmt.Sprintf(`CAST(SUM(job.num_nodes) as %s)`, castType),
fmt.Sprintf(`CAST(ROUND(SUM((CASE WHEN job.job_state = "running" THEN %d - job.start_time ELSE job.duration END) * job.num_nodes) / 3600) as %s)`, time.Now().Unix(), castType),
@@ -167,14 +174,14 @@ func (r *JobRepository) JobsStatsGrouped(
for rows.Next() {
var id sql.NullString
var name sql.NullString
var jobs, walltime, nodes, nodeHours, cores, coreHours, accs, accHours sql.NullInt64
if err := rows.Scan(&id, &jobs, &name, &walltime, &nodes, &nodeHours, &cores, &coreHours, &accs, &accHours); err != nil {
var jobs, users, walltime, nodes, nodeHours, cores, coreHours, accs, accHours sql.NullInt64
if err := rows.Scan(&id, &name, &jobs, &users, &walltime, &nodes, &nodeHours, &cores, &coreHours, &accs, &accHours); err != nil {
cclog.Warn("Error while scanning rows")
return nil, err
}
if id.Valid {
var totalJobs, totalWalltime, totalNodes, totalNodeHours, totalCores, totalCoreHours, totalAccs, totalAccHours int
var totalJobs, totalUsers, totalWalltime, totalNodes, totalNodeHours, totalCores, totalCoreHours, totalAccs, totalAccHours int
var personName string
if name.Valid {
@@ -185,6 +192,10 @@ func (r *JobRepository) JobsStatsGrouped(
totalJobs = int(jobs.Int64)
}
if users.Valid {
totalUsers = int(users.Int64)
}
if walltime.Valid {
totalWalltime = int(walltime.Int64)
}
@@ -228,8 +239,9 @@ func (r *JobRepository) JobsStatsGrouped(
stats = append(stats,
&model.JobsStatistics{
ID: id.String,
TotalJobs: int(jobs.Int64),
TotalWalltime: int(walltime.Int64),
TotalJobs: totalJobs,
TotalUsers: totalUsers,
TotalWalltime: totalWalltime,
TotalNodes: totalNodes,
TotalNodeHours: totalNodeHours,
TotalCores: totalCores,
@@ -259,8 +271,8 @@ func (r *JobRepository) JobsStats(
row := query.RunWith(r.DB).QueryRow()
stats := make([]*model.JobsStatistics, 0, 1)
var jobs, walltime, nodes, nodeHours, cores, coreHours, accs, accHours sql.NullInt64
if err := row.Scan(&jobs, &walltime, &nodes, &nodeHours, &cores, &coreHours, &accs, &accHours); err != nil {
var jobs, users, walltime, nodes, nodeHours, cores, coreHours, accs, accHours sql.NullInt64
if err := row.Scan(&jobs, &users, &walltime, &nodes, &nodeHours, &cores, &coreHours, &accs, &accHours); err != nil {
cclog.Warn("Error while scanning rows")
return nil, err
}
@@ -280,6 +292,7 @@ func (r *JobRepository) JobsStats(
stats = append(stats,
&model.JobsStatistics{
TotalJobs: int(jobs.Int64),
TotalUsers: int(users.Int64),
TotalWalltime: int(walltime.Int64),
TotalNodeHours: totalNodeHours,
TotalCoreHours: totalCoreHours,

View File

@@ -459,7 +459,7 @@
</tr>
{#each $topQuery.data.topList as te, i}
<tr>
<td><Icon name="circle-fill" style="color: {colors[i]};" /></td>
<td><Icon name="circle-fill" style="color: {colors['colorblind'][i]};" /></td>
{#if groupSelection.key == "user"}
<th scope="col" id="topName-{te.id}"
><a href="/monitoring/user/{te.id}?cluster={clusterName}"

View File

@@ -2,709 +2,62 @@
@component Main cluster status view component; renders current system-usage information
Properties:
- `cluster String`: The cluster to show status information for
- `presetCluster String`: The cluster to show status information for
-->
<script>
import { getContext } from "svelte";
import {
getContext
} from "svelte"
import {
Row,
Col,
Spinner,
Card,
CardHeader,
CardTitle,
CardBody,
Table,
Progress,
Icon,
Button,
Tooltip
TabContent,
TabPane
} from "@sveltestrap/sveltestrap";
import {
queryStore,
gql,
getContextClient,
mutationStore,
} from "@urql/svelte";
import {
init,
convert2uplot,
transformPerNodeDataForRoofline,
scramble,
scrambleNames,
} from "./generic/utils.js";
import { scaleNumbers } from "./generic/units.js";
import PlotGrid from "./generic/PlotGrid.svelte";
import Roofline from "./generic/plots/Roofline.svelte";
import Pie, { colors } from "./generic/plots/Pie.svelte";
import Histogram from "./generic/plots/Histogram.svelte";
import Refresher from "./generic/helper/Refresher.svelte";
import HistogramSelection from "./generic/select/HistogramSelection.svelte";
import StatusDash from "./status/StatusDash.svelte";
import UsageDash from "./status/UsageDash.svelte";
import StatisticsDash from "./status/StatisticsDash.svelte";
/* Svelte 5 Props */
let {
cluster
presetCluster
} = $props();
/* Const Init */
const { query: initq } = init();
const ccconfig = getContext("cc-config");
const client = getContextClient();
const paging = { itemsPerPage: 10, page: 1 }; // Top 10
const topOptions = [
{ key: "totalJobs", label: "Jobs" },
{ key: "totalNodes", label: "Nodes" },
{ key: "totalCores", label: "Cores" },
{ key: "totalAccs", label: "Accelerators" },
];
/*Const Init */
const useCbColors = getContext("cc-config")?.plot_general_colorblindMode || false
/* State Init */
let from = $state(new Date(Date.now() - 5 * 60 * 1000));
let to = $state(new Date(Date.now()));
let colWidth = $state(0);
let plotWidths = $state([]);
// Histrogram
let isHistogramSelectionOpen = $state(false);
let selectedHistograms = $state(cluster
? ccconfig[`user_view_histogramMetrics:${cluster}`] || ( ccconfig['user_view_histogramMetrics'] || [] )
: ccconfig['user_view_histogramMetrics'] || []);
// Bar Gauges
let allocatedNodes = $state({});
let flopRate = $state({});
let flopRateUnitPrefix = $state({});
let flopRateUnitBase = $state({});
let memBwRate = $state({});
let memBwRateUnitPrefix = $state({});
let memBwRateUnitBase = $state({});
// Pie Charts
let topProjectSelection = $state(
topOptions.find(
(option) =>
option.key ==
ccconfig[`status_view_selectedTopProjectCategory:${cluster}`],
) ||
topOptions.find(
(option) => option.key == ccconfig.status_view_selectedTopProjectCategory,
)
);
let topUserSelection = $state(
topOptions.find(
(option) =>
option.key ==
ccconfig[`status_view_selectedTopUserCategory:${cluster}`],
) ||
topOptions.find(
(option) => option.key == ccconfig.status_view_selectedTopUserCategory,
)
);
/* Derived */
// Note: nodeMetrics are requested on configured $timestep resolution
const mainQuery = $derived(queryStore({
client: client,
query: gql`
query (
$cluster: String!
$filter: [JobFilter!]!
$metrics: [String!]
$from: Time!
$to: Time!
$selectedHistograms: [String!]
) {
nodeMetrics(
cluster: $cluster
metrics: $metrics
from: $from
to: $to
) {
host
subCluster
metrics {
name
scope
metric {
timestep
unit {
base
prefix
}
series {
data
}
}
}
}
stats: jobsStatistics(filter: $filter, metrics: $selectedHistograms) {
histDuration {
count
value
}
histNumNodes {
count
value
}
histNumCores {
count
value
}
histNumAccs {
count
value
}
histMetrics {
metric
unit
data {
min
max
count
bin
}
}
}
allocatedNodes(cluster: $cluster) {
name
count
}
}
`,
variables: {
cluster: cluster,
metrics: ["flops_any", "mem_bw"], // Fixed names for roofline and status bars
from: from.toISOString(),
to: to.toISOString(),
filter: [{ state: ["running"] }, { cluster: { eq: cluster } }],
selectedHistograms: selectedHistograms,
},
}));
const topUserQuery = $derived(queryStore({
client: client,
query: gql`
query (
$filter: [JobFilter!]!
$paging: PageRequest!
$sortBy: SortByAggregate!
) {
topUser: jobsStatistics(
filter: $filter
page: $paging
sortBy: $sortBy
groupBy: USER
) {
id
name
totalJobs
totalNodes
totalCores
totalAccs
}
}
`,
variables: {
filter: [{ state: ["running"] }, { cluster: { eq: cluster } }],
paging,
sortBy: topUserSelection.key.toUpperCase(),
},
}));
const topProjectQuery = $derived(queryStore({
client: client,
query: gql`
query (
$filter: [JobFilter!]!
$paging: PageRequest!
$sortBy: SortByAggregate!
) {
topProjects: jobsStatistics(
filter: $filter
page: $paging
sortBy: $sortBy
groupBy: PROJECT
) {
id
totalJobs
totalNodes
totalCores
totalAccs
}
}
`,
variables: {
filter: [{ state: ["running"] }, { cluster: { eq: cluster } }],
paging,
sortBy: topProjectSelection.key.toUpperCase(),
},
}));
/* Effects */
$effect(() => {
if ($initq.data && $mainQuery.data) {
let subClusters = $initq.data.clusters.find(
(c) => c.name == cluster,
).subClusters;
for (let subCluster of subClusters) {
allocatedNodes[subCluster.name] =
$mainQuery.data.allocatedNodes.find(
({ name }) => name == subCluster.name,
)?.count || 0;
flopRate[subCluster.name] =
Math.floor(
sumUp($mainQuery.data.nodeMetrics, subCluster.name, "flops_any") *
100,
) / 100;
flopRateUnitPrefix[subCluster.name] = subCluster.flopRateSimd.unit.prefix;
flopRateUnitBase[subCluster.name] = subCluster.flopRateSimd.unit.base;
memBwRate[subCluster.name] =
Math.floor(
sumUp($mainQuery.data.nodeMetrics, subCluster.name, "mem_bw") * 100,
) / 100;
memBwRateUnitPrefix[subCluster.name] =
subCluster.memoryBandwidth.unit.prefix;
memBwRateUnitBase[subCluster.name] = subCluster.memoryBandwidth.unit.base;
}
}
});
$effect(() => {
updateTopUserConfiguration(topUserSelection.key);
});
$effect(() => {
updateTopProjectConfiguration(topProjectSelection.key);
});
/* Const Functions */
const sumUp = (data, subcluster, metric) =>
data.reduce(
(sum, node) =>
node.subCluster == subcluster
? sum +
(node.metrics
.find((m) => m.name == metric)
?.metric.series.reduce(
(sum, series) => sum + series.data[series.data.length - 1],
0,
) || 0)
: sum,
0,
);
const updateConfigurationMutation = ({ name, value }) => {
return mutationStore({
client: client,
query: gql`
mutation ($name: String!, $value: String!) {
updateConfiguration(name: $name, value: $value)
}
`,
variables: { name, value },
});
};
/* Functions */
function updateTopUserConfiguration(select) {
if (ccconfig[`status_view_selectedTopUserCategory:${cluster}`] != select) {
updateConfigurationMutation({
name: `status_view_selectedTopUserCategory:${cluster}`,
value: JSON.stringify(select),
}).subscribe((res) => {
if (res.fetching === false && res.error) {
throw res.error;
}
});
}
}
function updateTopProjectConfiguration(select) {
if (
ccconfig[`status_view_selectedTopProjectCategory:${cluster}`] != select
) {
updateConfigurationMutation({
name: `status_view_selectedTopProjectCategory:${cluster}`,
value: JSON.stringify(select),
}).subscribe((res) => {
if (res.fetching === false && res.error) {
throw res.error;
}
});
}
}
</script>
<!-- Loading indicator & Refresh -->
<Row cols={{ lg: 3, md: 3, sm: 1 }}>
<Col style="">
<h4 class="mb-0">Current utilization of cluster "{cluster}"</h4>
</Col>
<Col class="mt-2 mt-md-0 text-md-end">
<Button
outline
color="secondary"
onclick={() => (isHistogramSelectionOpen = true)}
>
<Icon name="bar-chart-line" /> Select Histograms
</Button>
</Col>
<Col class="mt-2 mt-md-0">
<Refresher
initially={120}
onRefresh={() => {
from = new Date(Date.now() - 5 * 60 * 1000);
to = new Date(Date.now());
}}
/>
</Col>
</Row>
<Row cols={1} class="text-center mt-3">
<Row cols={1} class="mb-2">
<Col>
{#if $initq.fetching || $mainQuery.fetching}
<Spinner />
{:else if $initq.error}
<Card body color="danger">{$initq.error.message}</Card>
{:else}
<!-- ... -->
{/if}
<h3 class="mb-0">Current Status of Cluster "{presetCluster.charAt(0).toUpperCase() + presetCluster.slice(1)}"</h3>
</Col>
</Row>
{#if $mainQuery.error}
<Row cols={1}>
<Col>
<Card body color="danger">{$mainQuery.error.message}</Card>
</Col>
</Row>
{/if}
<hr />
<Card class="overflow-auto" style="height: auto;">
<TabContent>
<TabPane tabId="status-dash" tab="Status" active>
<CardBody>
<StatusDash {presetCluster} {useCbColors} useAltColors></StatusDash>
</CardBody>
</TabPane>
<!-- Gauges & Roofline per Subcluster-->
{#if $initq.data && $mainQuery.data}
{#each $initq.data.clusters.find((c) => c.name == cluster).subClusters as subCluster, i}
<Row cols={{ lg: 2, md: 1 , sm: 1}} class="mb-3 justify-content-center">
<Col class="px-3">
<Card class="h-auto mt-1">
<CardHeader>
<CardTitle class="mb-0">SubCluster "{subCluster.name}"</CardTitle>
</CardHeader>
<CardBody>
<Table borderless>
<tr class="py-2">
<th scope="col">Allocated Nodes</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={allocatedNodes[subCluster.name]}
max={subCluster.numberOfNodes}
/>
</div></td
>
<td
>{allocatedNodes[subCluster.name]} / {subCluster.numberOfNodes}
Nodes</td
>
</tr>
<tr class="py-2">
<th scope="col"
>Flop Rate (Any) <Icon
name="info-circle"
class="p-1"
style="cursor: help;"
title="Flops[Any] = (Flops[Double] x 2) + Flops[Single]"
/></th
>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={flopRate[subCluster.name]}
max={subCluster.flopRateSimd.value *
subCluster.numberOfNodes}
/>
</div></td
>
<td>
{scaleNumbers(
flopRate[subCluster.name],
subCluster.flopRateSimd.value * subCluster.numberOfNodes,
flopRateUnitPrefix[subCluster.name],
)}{flopRateUnitBase[subCluster.name]} [Max]
</td>
</tr>
<tr class="py-2">
<th scope="col">MemBw Rate</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={memBwRate[subCluster.name]}
max={subCluster.memoryBandwidth.value *
subCluster.numberOfNodes}
/>
</div></td
>
<td>
{scaleNumbers(
memBwRate[subCluster.name],
subCluster.memoryBandwidth.value * subCluster.numberOfNodes,
memBwRateUnitPrefix[subCluster.name],
)}{memBwRateUnitBase[subCluster.name]} [Max]
</td>
</tr>
</Table>
</CardBody>
</Card>
</Col>
<Col class="px-3 mt-2 mt-lg-0">
<div bind:clientWidth={plotWidths[i]}>
{#key $mainQuery.data.nodeMetrics}
<Roofline
allowSizeChange
width={plotWidths[i] - 10}
height={300}
subCluster={subCluster}
data={transformPerNodeDataForRoofline(
$mainQuery.data.nodeMetrics.filter(
(data) => data.subCluster == subCluster.name,
),
)}
/>
{/key}
</div>
</Col>
</Row>
{/each}
<hr />
<!-- User and Project Stats as Pie-Charts -->
<Row cols={{ lg: 4, md: 2, sm: 1 }}>
<Col class="p-2">
<div bind:clientWidth={colWidth}>
<h4 class="text-center">
Top Users on {cluster.charAt(0).toUpperCase() + cluster.slice(1)}
</h4>
{#key $topUserQuery.data}
{#if $topUserQuery.fetching}
<Spinner />
{:else if $topUserQuery.error}
<Card body color="danger">{$topUserQuery.error.message}</Card>
{:else}
<Pie
canvasId="hpcpie-users"
size={colWidth}
sliceLabel={topUserSelection.label}
quantities={$topUserQuery.data.topUser.map(
(tu) => tu[topUserSelection.key],
)}
entities={$topUserQuery.data.topUser.map((tu) => scrambleNames ? scramble(tu.id) : tu.id)}
/>
{/if}
{/key}
</div>
</Col>
<Col class="px-4 py-2">
{#key $topUserQuery.data}
{#if $topUserQuery.fetching}
<Spinner />
{:else if $topUserQuery.error}
<Card body color="danger">{$topUserQuery.error.message}</Card>
{:else}
<Table>
<tr class="mb-2">
<th>Legend</th>
<th>User Name</th>
<th
>Number of
<select class="p-0" bind:value={topUserSelection}>
{#each topOptions as option}
<option value={option}>
{option.label}
</option>
{/each}
</select>
</th>
</tr>
{#each $topUserQuery.data.topUser as tu, i}
<tr>
<td><Icon name="circle-fill" style="color: {colors[i]};" /></td>
<th scope="col" id="topName-{tu.id}"
><a
href="/monitoring/user/{tu.id}?cluster={cluster}&state=running"
>{scrambleNames ? scramble(tu.id) : tu.id}</a
></th
>
{#if tu?.name}
<Tooltip
target={`topName-${tu.id}`}
placement="left"
>{scrambleNames ? scramble(tu.name) : tu.name}</Tooltip
>
{/if}
<td>{tu[topUserSelection.key]}</td>
</tr>
{/each}
</Table>
{/if}
{/key}
</Col>
<Col class="p-2">
<h4 class="text-center">
Top Projects on {cluster.charAt(0).toUpperCase() + cluster.slice(1)}
</h4>
{#key $topProjectQuery.data}
{#if $topProjectQuery.fetching}
<Spinner />
{:else if $topProjectQuery.error}
<Card body color="danger">{$topProjectQuery.error.message}</Card>
{:else}
<Pie
canvasId="hpcpie-projects"
size={colWidth}
sliceLabel={topProjectSelection.label}
quantities={$topProjectQuery.data.topProjects.map(
(tp) => tp[topProjectSelection.key],
)}
entities={$topProjectQuery.data.topProjects.map((tp) => scrambleNames ? scramble(tp.id) : tp.id)}
/>
{/if}
{/key}
</Col>
<Col class="px-4 py-2">
{#key $topProjectQuery.data}
{#if $topProjectQuery.fetching}
<Spinner />
{:else if $topProjectQuery.error}
<Card body color="danger">{$topProjectQuery.error.message}</Card>
{:else}
<Table>
<tr class="mb-2">
<th>Legend</th>
<th>Project Code</th>
<th
>Number of
<select class="p-0" bind:value={topProjectSelection}>
{#each topOptions as option}
<option value={option}>
{option.label}
</option>
{/each}
</select>
</th>
</tr>
{#each $topProjectQuery.data.topProjects as tp, i}
<tr>
<td><Icon name="circle-fill" style="color: {colors[i]};" /></td>
<th scope="col"
><a
href="/monitoring/jobs/?cluster={cluster}&state=running&project={tp.id}&projectMatch=eq"
>{scrambleNames ? scramble(tp.id) : tp.id}</a
></th
>
<td>{tp[topProjectSelection.key]}</td>
</tr>
{/each}
</Table>
{/if}
{/key}
</Col>
</Row>
<hr class="my-2" />
<!-- Static Stats as Histograms : Running Duration && Allocated Hardware Counts-->
<Row cols={{ lg: 2, md: 1 }}>
<Col class="p-2">
{#key $mainQuery.data.stats}
<Histogram
data={convert2uplot($mainQuery.data.stats[0].histDuration)}
title="Duration Distribution"
xlabel="Current Job Runtimes"
xunit="Runtime"
ylabel="Number of Jobs"
yunit="Jobs"
usesBins
xtime
/>
{/key}
</Col>
<Col class="p-2">
{#key $mainQuery.data.stats}
<Histogram
data={convert2uplot($mainQuery.data.stats[0].histNumNodes)}
title="Number of Nodes Distribution"
xlabel="Allocated Nodes"
xunit="Nodes"
ylabel="Number of Jobs"
yunit="Jobs"
/>
{/key}
</Col>
</Row>
<Row cols={{ lg: 2, md: 1 }}>
<Col class="p-2">
{#key $mainQuery.data.stats}
<Histogram
data={convert2uplot($mainQuery.data.stats[0].histNumCores)}
title="Number of Cores Distribution"
xlabel="Allocated Cores"
xunit="Cores"
ylabel="Number of Jobs"
yunit="Jobs"
/>
{/key}
</Col>
<Col class="p-2">
{#key $mainQuery.data.stats}
<Histogram
data={convert2uplot($mainQuery.data.stats[0].histNumAccs)}
title="Number of Accelerators Distribution"
xlabel="Allocated Accs"
xunit="Accs"
ylabel="Number of Jobs"
yunit="Jobs"
/>
{/key}
</Col>
</Row>
<hr class="my-2" />
<!-- Selectable Stats as Histograms : Average Values of Running Jobs -->
{#if selectedHistograms}
<!-- Note: Ignore '#snippet' Error in IDE -->
{#snippet gridContent(item)}
<Histogram
data={convert2uplot(item.data)}
title="Distribution of '{item.metric}' averages"
xlabel={`${item.metric} bin maximum ${item?.unit ? `[${item.unit}]` : ``}`}
xunit={item.unit}
ylabel="Number of Jobs"
yunit="Jobs"
usesBins
/>
{/snippet}
<TabPane tabId="usage-dash" tab="Usage">
<CardBody>
<UsageDash {presetCluster} {useCbColors}></UsageDash>
</CardBody>
</TabPane>
{#key $mainQuery.data.stats[0].histMetrics}
<PlotGrid
items={$mainQuery.data.stats[0].histMetrics}
itemsPerRow={2}
{gridContent}
/>
{/key}
{/if}
{/if}
<HistogramSelection
{cluster}
bind:isOpen={isHistogramSelectionOpen}
presetSelectedHistograms={selectedHistograms}
applyChange={(newSelection) => {
selectedHistograms = [...newSelection];
}}
/>
<TabPane tabId="metric-dash" tab="Statistics">
<CardBody>
<StatisticsDash {presetCluster} {useCbColors}></StatisticsDash>
</CardBody>
</TabPane>
</TabContent>
</Card>

View File

@@ -404,6 +404,7 @@
cluster={selectedCluster}
bind:isOpen={isHistogramSelectionOpen}
presetSelectedHistograms={selectedHistograms}
configName="user_view_histogramMetrics"
applyChange={(newSelection) => {
selectedHistogramsBuffer[selectedCluster || 'all'] = [...newSelection];
}}

View File

@@ -26,7 +26,7 @@
/* Svelte 5 Props */
let {
matchedCompareJobs = $bindable(0),
metrics = ccconfig?.plot_list_selectedMetrics,
metrics = getContext("cc-config")?.plot_list_selectedMetrics,
filterBuffer = [],
} = $props();

View File

@@ -44,7 +44,7 @@
/* Const Init */
const clusterCockpitConfig = getContext("cc-config");
const lineWidth = clusterCockpitConfig.plot_general_lineWidth / window.devicePixelRatio;
const lineWidth = clusterCockpitConfig?.plot_general_lineWidth / window.devicePixelRatio || 2;
const cbmode = clusterCockpitConfig?.plot_general_colorblindMode || false;
// UPLOT SERIES INIT //

View File

@@ -14,28 +14,59 @@
-->
<script module>
// http://tsitsul.in/blog/coloropt/ : 12 colors normal
export const colors = [
'rgb(235,172,35)',
'rgb(184,0,88)',
'rgb(0,140,249)',
'rgb(0,110,0)',
'rgb(0,187,173)',
'rgb(209,99,230)',
'rgb(178,69,2)',
'rgb(255,146,135)',
'rgb(89,84,214)',
'rgb(0,198,248)',
'rgb(135,133,0)',
'rgb(0,167,108)',
'rgb(189,189,189)'
];
export const colors = {
// https://www.learnui.design/tools/data-color-picker.html#divergent: 11, Shallow Green-Red
default: [
"#00876c",
"#449c6e",
"#70af6f",
"#9bc271",
"#c8d377",
"#f7e382",
"#f6c468",
"#f3a457",
"#ed834e",
"#e3614d",
"#d43d51",
],
// https://www.learnui.design/tools/data-color-picker.html#palette: 12, Colorwheel-Like
alternative: [
"#0022bb",
"#ba0098",
"#fa0066",
"#ff6234",
"#ffae00",
"#b1af00",
"#67a630",
"#009753",
"#00836c",
"#006d77",
"#005671",
"#003f5c",
],
// http://tsitsul.in/blog/coloropt/ : 12 colors normal
colorblind: [
'rgb(235,172,35)',
'rgb(184,0,88)',
'rgb(0,140,249)',
'rgb(0,110,0)',
'rgb(0,187,173)',
'rgb(209,99,230)',
'rgb(178,69,2)',
'rgb(255,146,135)',
'rgb(89,84,214)',
'rgb(0,198,248)',
'rgb(135,133,0)',
'rgb(0,167,108)',
'rgb(189,189,189)',
]
}
</script>
<script>
/* Ignore Double Script Section Error in IDE */
// Ignore VSC IDE "One Instance Level Script" Error
import { onMount, getContext } from "svelte";
import Chart from 'chart.js/auto';
import { onMount } from 'svelte';
/* Svelte 5 Props */
let {
@@ -45,21 +76,11 @@
quantities,
entities,
displayLegend = false,
useAltColors = false,
} = $props();
/* Const Init */
const data = {
labels: entities,
datasets: [
{
label: sliceLabel,
data: quantities,
fill: 1,
backgroundColor: colors.slice(0, quantities.length)
}
]
};
const useCbColors = getContext("cc-config")?.plot_general_colorblindMode || false
const options = {
maintainAspectRatio: false,
animation: false,
@@ -70,6 +91,31 @@
}
};
/* Derived */
const colorPalette = $derived.by(() => {
let c;
if (useCbColors) {
c = [...colors['colorblind']];
} else if (useAltColors) {
c = [...colors['alternative']];
} else {
c = [...colors['default']];
}
return c.slice(0, quantities.length);
})
const data = $derived({
labels: entities,
datasets: [
{
label: sliceLabel,
data: quantities,
fill: 1,
backgroundColor: colorPalette,
}
]
});
/* On Mount */
onMount(() => {
new Chart(
@@ -84,7 +130,7 @@
</script>
<!-- <div style="width: 500px;"><canvas id="dimensions"></canvas></div><br/> -->
<div class="chart-container" style="--container-width: {size}; --container-height: {size}">
<div class="chart-container" style="--container-width: {size}px; --container-height: {size}px">
<canvas id={canvasId}></canvas>
</div>

View File

@@ -3,7 +3,6 @@
Properties:
- `data [null, [], []]`: Roofline Data Structure, see below for details [Default: null]
- `renderTime Bool?`: If time information should be rendered as colored dots [Default: false]
- `allowSizeChange Bool?`: If dimensions of rendered plot can change [Default: false]
- `subCluster GraphQL.SubCluster?`: SubCluster Object; contains required topology information [Default: null]
- `width Number?`: Plot width (reactively adaptive) [Default: 600]
@@ -21,19 +20,22 @@
- `data[2] = [0.1, 0.15, 0.2, ...]`
- Color Code: Time Information (Floats from 0 to 1) (Optional)
-->
<script>
import uPlot from "uplot";
import { formatNumber } from "../units.js";
import { onMount, onDestroy } from "svelte";
import { Card } from "@sveltestrap/sveltestrap";
import { roundTwoDigits } from "../units.js";
/* Svelte 5 Props */
let {
data = null,
renderTime = false,
allowSizeChange = false,
roofData = null,
jobsData = null,
nodesData = null,
cluster = null,
subCluster = null,
allowSizeChange = false,
useColors = true,
width = 600,
height = 380,
} = $props();
@@ -54,8 +56,27 @@
if (allowSizeChange) sizeChanged(width, height);
});
// Copied Example Vars for Uplot Bubble
// https://developer.mozilla.org/en-US/docs/Web/API/CanvasRenderingContext2D/isPointInPath
let qt;
let hRect;
let pxRatio;
function setPxRatio() {
pxRatio = uPlot.pxRatio;
}
setPxRatio();
window.addEventListener('dppxchange', setPxRatio);
// let minSize = 6;
let maxSize = 60;
// let maxArea = Math.PI * (maxSize / 2) ** 2;
// let minArea = Math.PI * (minSize / 2) ** 2;
/* Functions */
// Helper
function pointWithin(px, py, rlft, rtop, rrgt, rbtm) {
return px >= rlft && px <= rrgt && py >= rtop && py <= rbtm;
}
function getGradientR(x) {
if (x < 0.5) return 0;
if (x > 0.75) return 255;
@@ -74,8 +95,9 @@
x = 1.0 - (x - 0.25) * 4.0;
return Math.floor(x * 255.0);
}
function getRGB(c) {
return `rgb(${cbmode ? '0' : getGradientR(c)}, ${getGradientG(c)}, ${getGradientB(c)})`;
function getRGB(c, transparent = false) {
if (transparent) return `rgba(${cbmode ? '0' : getGradientR(c)}, ${getGradientG(c)}, ${getGradientB(c)}, 0.5)`;
else return `rgb(${cbmode ? '0' : getGradientR(c)}, ${getGradientG(c)}, ${getGradientB(c)})`;
}
function nearestThousand(num) {
return Math.ceil(num / 1000) * 1000;
@@ -89,126 +111,492 @@
};
}
// Dot Renderers
const drawColorPoints = (u, seriesIdx, idx0, idx1) => {
const size = 5 * devicePixelRatio;
uPlot.orient(
u,
seriesIdx,
(
series,
dataX,
dataY,
scaleX,
scaleY,
valToPosX,
valToPosY,
xOff,
yOff,
xDim,
yDim,
moveTo,
lineTo,
rect,
arc,
) => {
// quadratic scaling (px area)
// function getSize(value, minValue, maxValue) {
// let pct = value / maxValue;
// // clamp to min area
// //let area = Math.max(maxArea * pct, minArea);
// let area = maxArea * pct;
// return Math.sqrt(area / Math.PI) * 2;
// }
// function getSizeMinMax(u) {
// let minValue = Infinity;
// let maxValue = -Infinity;
// for (let i = 1; i < u.series.length; i++) {
// let sizeData = u.data[i][2];
// for (let j = 0; j < sizeData.length; j++) {
// minValue = Math.min(minValue, sizeData[j]);
// maxValue = Math.max(maxValue, sizeData[j]);
// }
// }
// return [minValue, maxValue];
// }
// Quadtree Object (TODO: Split and Import)
class Quadtree {
constructor (x, y, w, h, l) {
let t = this;
t.x = x;
t.y = y;
t.w = w;
t.h = h;
t.l = l || 0;
t.o = [];
t.q = null;
t.MAX_OBJECTS = 10;
t.MAX_LEVELS = 4;
};
get quadtree() {
return "Implement me!";
}
split() {
let t = this,
x = t.x,
y = t.y,
w = t.w / 2,
h = t.h / 2,
l = t.l + 1;
t.q = [
// top right
new Quadtree(x + w, y, w, h, l),
// top left
new Quadtree(x, y, w, h, l),
// bottom left
new Quadtree(x, y + h, w, h, l),
// bottom right
new Quadtree(x + w, y + h, w, h, l),
];
};
quads(x, y, w, h, cb) {
let t = this,
q = t.q,
hzMid = t.x + t.w / 2,
vtMid = t.y + t.h / 2,
startIsNorth = y < vtMid,
startIsWest = x < hzMid,
endIsEast = x + w > hzMid,
endIsSouth = y + h > vtMid;
// top-right quad
startIsNorth && endIsEast && cb(q[0]);
// top-left quad
startIsWest && startIsNorth && cb(q[1]);
// bottom-left quad
startIsWest && endIsSouth && cb(q[2]);
// bottom-right quad
endIsEast && endIsSouth && cb(q[3]);
};
add(o) {
let t = this;
if (t.q != null) {
t.quads(o.x, o.y, o.w, o.h, q => {
q.add(o);
});
}
else {
let os = t.o;
os.push(o);
if (os.length > t.MAX_OBJECTS && t.l < t.MAX_LEVELS) {
t.split();
for (let i = 0; i < os.length; i++) {
let oi = os[i];
t.quads(oi.x, oi.y, oi.w, oi.h, q => {
q.add(oi);
});
}
t.o.length = 0;
}
}
};
get(x, y, w, h, cb) {
let t = this;
let os = t.o;
for (let i = 0; i < os.length; i++)
cb(os[i]);
if (t.q != null) {
t.quads(x, y, w, h, q => {
q.get(x, y, w, h, cb);
});
}
}
clear() {
this.o.length = 0;
this.q = null;
}
}
// Dot Renderer
const makeDrawPoints = (opts) => {
let {/*size, disp,*/ transparentFill, each = () => {}} = opts;
const sizeBase = 6 * pxRatio;
return (u, seriesIdx, idx0, idx1) => {
uPlot.orient(u, seriesIdx, (series, dataX, dataY, scaleX, scaleY, valToPosX, valToPosY, xOff, yOff, xDim, yDim, moveTo, lineTo, rect, arc) => {
let d = u.data[seriesIdx];
let strokeWidth = 1;
let deg360 = 2 * Math.PI;
/* Alt.: Sizes based on other Data Rows */
// let sizes = disp.size.values(u, seriesIdx, idx0, idx1);
u.ctx.save();
u.ctx.rect(u.bbox.left, u.bbox.top, u.bbox.width, u.bbox.height);
u.ctx.clip();
u.ctx.lineWidth = strokeWidth;
// todo: this depends on direction & orientation
// todo: calc once per redraw, not per path
let filtLft = u.posToVal(-maxSize / 2, scaleX.key);
let filtRgt = u.posToVal(u.bbox.width / pxRatio + maxSize / 2, scaleX.key);
let filtBtm = u.posToVal(u.bbox.height / pxRatio + maxSize / 2, scaleY.key);
let filtTop = u.posToVal(-maxSize / 2, scaleY.key);
for (let i = 0; i < d[0].length; i++) {
let p = new Path2D();
if (useColors) {
u.ctx.strokeStyle = "rgb(0, 0, 0)";
// Jobs: Color based on Duration
if (jobsData) {
//u.ctx.strokeStyle = getRGB(u.data[2][i]);
u.ctx.fillStyle = getRGB(u.data[2][i], transparentFill);
// Nodes: Color based on Idle vs. Allocated
} else if (nodesData) {
// console.log('In Plot Handler NodesData', nodesData)
if (nodesData[i]?.nodeState == "idle") {
//u.ctx.strokeStyle = "rgb(0, 0, 255)";
u.ctx.fillStyle = "rgba(0, 0, 255, 0.5)";
} else if (nodesData[i]?.nodeState == "allocated") {
//u.ctx.strokeStyle = "rgb(0, 255, 0)";
u.ctx.fillStyle = "rgba(0, 255, 0, 0.5)";
} else if (nodesData[i]?.nodeState == "notindb") {
//u.ctx.strokeStyle = "rgb(0, 0, 0)";
u.ctx.fillStyle = "rgba(0, 0, 0, 0.5)";
} else { // Fallback: All other DEFINED states
//u.ctx.strokeStyle = "rgb(255, 0, 0)";
u.ctx.fillStyle = "rgba(255, 0, 0, 0.5)";
}
}
} else {
// No Colors: Use Black
u.ctx.strokeStyle = "rgb(0, 0, 0)";
u.ctx.fillStyle = "rgba(0, 0, 0, 0.5)";
}
// Get Values
let xVal = d[0][i];
let yVal = d[1][i];
u.ctx.strokeStyle = getRGB(u.data[2][i]);
u.ctx.fillStyle = getRGB(u.data[2][i]);
if (
xVal >= scaleX.min &&
xVal <= scaleX.max &&
yVal >= scaleY.min &&
yVal <= scaleY.max
) {
// Calc Size; Alt.: size = sizes[i] * pxRatio
let size = 1;
// Jobs: Size based on Resourcecount
if (jobsData) {
const scaling = jobsData[i].numNodes > 12
? 24 // Capped Dot Size
: jobsData[i].numNodes > 1
? jobsData[i].numNodes * 2 // MultiNode Scaling
: jobsData[i]?.numAcc ? jobsData[i].numAcc : jobsData[i].numNodes * 2 // Single Node or Scale by Accs
size = sizeBase + scaling
// Nodes: Size based on Jobcount
} else if (nodesData) {
size = sizeBase + (nodesData[i]?.numJobs * 1.5) // Max Jobs Scale: 8 * 1.5 = 12
};
if (xVal >= filtLft && xVal <= filtRgt && yVal >= filtBtm && yVal <= filtTop) {
let cx = valToPosX(xVal, scaleX, xDim, xOff);
let cy = valToPosY(yVal, scaleY, yDim, yOff);
p.moveTo(cx + size / 2, cy);
arc(p, cx, cy, size / 2, 0, deg360);
u.ctx.moveTo(cx + size/2, cy);
u.ctx.beginPath();
u.ctx.arc(cx, cy, size/2, 0, deg360);
u.ctx.fill();
u.ctx.stroke();
each(u, seriesIdx, i,
cx - size/2 - strokeWidth/2,
cy - size/2 - strokeWidth/2,
size + strokeWidth,
size + strokeWidth
);
}
u.ctx.fill(p);
}
},
);
return null;
u.ctx.restore();
});
return null;
};
};
const drawPoints = (u, seriesIdx, idx0, idx1) => {
const size = 5 * devicePixelRatio;
uPlot.orient(
u,
seriesIdx,
(
series,
dataX,
dataY,
scaleX,
scaleY,
valToPosX,
valToPosY,
xOff,
yOff,
xDim,
yDim,
moveTo,
lineTo,
rect,
arc,
) => {
let d = u.data[seriesIdx];
u.ctx.strokeStyle = getRGB(0);
u.ctx.fillStyle = getRGB(0);
let deg360 = 2 * Math.PI;
let p = new Path2D();
for (let i = 0; i < d[0].length; i++) {
let xVal = d[0][i];
let yVal = d[1][i];
if (
xVal >= scaleX.min &&
xVal <= scaleX.max &&
yVal >= scaleY.min &&
yVal <= scaleY.max
) {
let cx = valToPosX(xVal, scaleX, xDim, xOff);
let cy = valToPosY(yVal, scaleY, yDim, yOff);
p.moveTo(cx + size / 2, cy);
arc(p, cx, cy, size / 2, 0, deg360);
let drawPoints = makeDrawPoints({
// disp: {
// size: {
// // unit: 3, // raw CSS pixels
// // discr: true,
// values: (u, seriesIdx, idx0, idx1) => {
// /* Func to get sizes from additional subSeries [series][2...x] ([0,1] is [x,y]) */
// // TODO: only run once per setData() call
// let [minValue, maxValue] = getSizeMinMax(u);
// return u.data[seriesIdx][2].map(v => getSize(v, minValue, maxValue));
// },
// },
// },
transparentFill: true,
each: (u, seriesIdx, dataIdx, lft, top, wid, hgt) => {
// we get back raw canvas coords (included axes & padding). translate to the plotting area origin
lft -= u.bbox.left;
top -= u.bbox.top;
qt.add({x: lft, y: top, w: wid, h: hgt, sidx: seriesIdx, didx: dataIdx});
},
});
const legendValues = (u, seriesIdx, dataIdx) => {
// when data null, it's initial schema probe (also u.status == 0)
if (u.data == null || dataIdx == null || hRect == null || hRect.sidx != seriesIdx) {
return {
"Intensity [FLOPS/Byte]": '-',
"":'',
"Performace [GFLOPS]": '-'
};
}
return {
"Intensity [FLOPS/Byte]": roundTwoDigits(u.data[seriesIdx][0][dataIdx]),
"":'',
"Performace [GFLOPS]": roundTwoDigits(u.data[seriesIdx][1][dataIdx]),
};
};
// Tooltip Plugin
function tooltipPlugin({onclick, getLegendData, shiftX = 10, shiftY = 10}) {
let tooltipLeftOffset = 0;
let tooltipTopOffset = 0;
const tooltip = document.createElement("div");
// Build Manual Class By Styles
tooltip.style.fontSize = "10pt";
tooltip.style.position = "absolute";
tooltip.style.background = "#fcfcfc";
tooltip.style.display = "none";
tooltip.style.border = "2px solid black";
tooltip.style.padding = "4px";
tooltip.style.pointerEvents = "none";
tooltip.style.zIndex = "100";
tooltip.style.whiteSpace = "pre";
tooltip.style.fontFamily = "monospace";
const tipSeriesIdx = 1; // Scatter: Series IDX is always 1
let tipDataIdx = null;
// const fmtDate = uPlot.fmtDate("{M}/{D}/{YY} {h}:{mm}:{ss} {AA}");
let over;
let tooltipVisible = false;
function showTooltip() {
if (!tooltipVisible) {
tooltip.style.display = "block";
over.style.cursor = "pointer";
tooltipVisible = true;
}
}
function hideTooltip() {
if (tooltipVisible) {
tooltip.style.display = "none";
over.style.cursor = null;
tooltipVisible = false;
}
}
function setTooltip(u, i) {
showTooltip();
let top = u.valToPos(u.data[tipSeriesIdx][1][i], 'y');
let lft = u.valToPos(u.data[tipSeriesIdx][0][i], 'x');
tooltip.style.top = (tooltipTopOffset + top + shiftX) + "px";
tooltip.style.left = (tooltipLeftOffset + lft + shiftY) + "px";
if (useColors) {
// Jobs: Color based on Duration
if (jobsData) {
tooltip.style.borderColor = getRGB(u.data[2][i]);
// Nodes: Color based on Idle vs. Allocated
} else if (nodesData) {
if (nodesData[i]?.nodeState == "idle") {
tooltip.style.borderColor = "rgb(0, 0, 255)";
} else if (nodesData[i]?.nodeState == "allocated") {
tooltip.style.borderColor = "rgb(0, 255, 0)";
} else if (nodesData[i]?.nodeState == "notindb") { // Missing from DB table
tooltip.style.borderColor = "rgb(0, 0, 0)";
} else { // Fallback: All other DEFINED states
tooltip.style.borderColor = "rgb(255, 0, 0)";
}
}
u.ctx.fill(p);
},
);
return null;
};
} else {
// No Colors: Use Black
tooltip.style.borderColor = "rgb(0, 0, 0)";
}
if (jobsData) {
tooltip.textContent = (
// Tooltip Content as String for Job
`Job ID: ${getLegendData(u, i).jobId}\nRuntime: ${getLegendData(u, i).duration}\nNodes: ${getLegendData(u, i).numNodes}${getLegendData(u, i)?.numAcc?`\nAccelerators: ${getLegendData(u, i).numAcc}`:''}`
);
} else if (nodesData && useColors) {
tooltip.textContent = (
// Tooltip Content as String for Node
`Host: ${getLegendData(u, i).nodeName}\nState: ${getLegendData(u, i).nodeState}\nJobs: ${getLegendData(u, i).numJobs}`
);
} else if (nodesData && !useColors) {
tooltip.textContent = (
// Tooltip Content as String for Node
`Host: ${getLegendData(u, i).nodeName}\nJobs: ${getLegendData(u, i).numJobs}`
);
}
}
return {
hooks: {
ready: [
u => {
over = u.over;
tooltipLeftOffset = parseFloat(over.style.left);
tooltipTopOffset = parseFloat(over.style.top);
u.root.querySelector(".u-wrap").appendChild(tooltip);
let clientX;
let clientY;
over.addEventListener("mousedown", e => {
clientX = e.clientX;
clientY = e.clientY;
});
over.addEventListener("mouseup", e => {
// clicked in-place
if (e.clientX == clientX && e.clientY == clientY) {
if (tipDataIdx != null) {
onclick(u, tipDataIdx);
}
}
});
}
],
setCursor: [
u => {
let i = u.legend.idxs[1];
if (i != null) {
tipDataIdx = i;
setTooltip(u, i);
} else {
tipDataIdx = null;
hideTooltip();
}
}
]
}
};
}
// Main Functions
function sizeChanged() {
if (timeoutId != null) clearTimeout(timeoutId);
timeoutId = setTimeout(() => {
timeoutId = null;
if (uplot) uplot.destroy();
render(data);
render(roofData, jobsData, nodesData);
}, 200);
}
function render(plotData) {
if (plotData) {
function render(roofData, jobsData, nodesData) {
let plotTitle = "CPU Roofline Diagram";
if (jobsData) plotTitle = "Job Average Roofline Diagram";
if (nodesData) plotTitle = "Node Average Roofline Diagram";
if (roofData) {
const opts = {
title: "CPU Roofline Diagram",
title: plotTitle,
mode: 2,
width: width,
height: height,
legend: {
show: false,
show: true,
},
cursor: {
dataIdx: (u, seriesIdx) => {
if (seriesIdx == 1) {
hRect = null;
let dist = Infinity;
let area = Infinity;
let cx = u.cursor.left * pxRatio;
let cy = u.cursor.top * pxRatio;
qt.get(cx, cy, 1, 1, o => {
if (pointWithin(cx, cy, o.x, o.y, o.x + o.w, o.y + o.h)) {
let ocx = o.x + o.w / 2;
let ocy = o.y + o.h / 2;
let dx = ocx - cx;
let dy = ocy - cy;
let d = Math.sqrt(dx ** 2 + dy ** 2);
// test against radius for actual hover
if (d <= o.w / 2) {
let a = o.w * o.h;
// prefer smallest
if (a < area) {
area = a;
dist = d;
hRect = o;
}
// only hover bbox with closest distance
else if (a == area && d <= dist) {
dist = d;
hRect = o;
}
}
}
});
}
return hRect && seriesIdx == hRect.sidx ? hRect.didx : null;
},
/* Render "Fill" on Data Point Hover: Works in Example Bubble, does not work here? Guess: Interference with tooltip */
// points: {
// size: (u, seriesIdx) => {
// return hRect && seriesIdx == hRect.sidx ? hRect.w / pxRatio : 0;
// }
// },
/* Make all non-focused series semi-transparent: Useless unless more than one series rendered */
// focus: {
// prox: 1e3,
// alpha: 0.3,
// dist: (u, seriesIdx) => {
// let prox = (hRect?.sidx === seriesIdx ? 0 : Infinity);
// return prox;
// },
// },
drag: { // Activates Zoom: Only one Dimension; YX Breaks Zoom Reset (Reason TBD)
x: true,
y: false
},
},
cursor: { drag: { x: false, y: false } },
axes: [
{
label: "Intensity [FLOPS/Byte]",
@@ -228,7 +616,7 @@
},
y: {
range: [
1.0,
0.01,
subCluster?.flopRateSimd?.value
? nearestThousand(subCluster.flopRateSimd.value)
: 10000,
@@ -237,12 +625,36 @@
log: 10, // log exp
},
},
series: [{}, { paths: renderTime ? drawColorPoints : drawPoints }],
series: [
null,
{
/* Facets: Define Purpose of Sub-Arrays in Series-Array, e.g. x, y, size, label, color, ... */
// facets: [
// {
// scale: 'x',
// auto: true,
// },
// {
// scale: 'y',
// auto: true,
// }
// ],
paths: drawPoints,
values: legendValues
}
],
hooks: {
// setSeries: [ (u, seriesIdx) => console.log('setSeries', seriesIdx) ],
// setLegend: [ u => console.log('setLegend', u.legend.idxs) ],
drawClear: [
(u) => {
qt = qt || new Quadtree(0, 0, u.bbox.width, u.bbox.height);
qt.clear();
// force-clear the path cache to cause drawBars() to rebuild new quadtree
u.series.forEach((s, i) => {
if (i > 0) s._paths = null;
if (i > 0)
s._paths = null;
});
},
],
@@ -334,30 +746,92 @@
// Reset grid lineWidth
u.ctx.lineWidth = 0.15;
}
if (renderTime) {
// The Color Scale For Time Information
const posX = u.valToPos(0.1, "x", true)
const posXLimit = u.valToPos(100, "x", true)
const posY = u.valToPos(14000.0, "y", true)
u.ctx.fillStyle = 'black'
u.ctx.fillText('Start', posX, posY)
const start = posX + 10
for (let x = start; x < posXLimit; x += 10) {
/* Render Scales */
if (useColors) {
// Jobs: The Color Scale For Time Information
if (jobsData) {
const posX = u.valToPos(0.1, "x", true)
const posXLimit = u.valToPos(100, "x", true)
const posY = u.valToPos(17500.0, "y", true)
u.ctx.fillStyle = 'black'
u.ctx.fillText('0 Hours', posX, posY)
const start = posX + 10
for (let x = start; x < posXLimit; x += 10) {
let c = (x - start) / (posXLimit - start)
u.ctx.fillStyle = getRGB(c)
u.ctx.beginPath()
u.ctx.arc(x, posY, 3, 0, Math.PI * 2, false)
u.ctx.fill()
}
u.ctx.fillStyle = 'black'
u.ctx.fillText('24 Hours', posXLimit + 55, posY)
}
// Nodes: The Colors Of NodeStates
if (nodesData) {
const posY = u.valToPos(17500.0, "y", true)
const posAllocDot = u.valToPos(0.03, "x", true)
const posAllocText = posAllocDot + 60
const posIdleDot = u.valToPos(0.3, "x", true)
const posIdleText = posIdleDot + 30
const posOtherDot = u.valToPos(3, "x", true)
const posOtherText = posOtherDot + 40
const posMissingDot = u.valToPos(30, "x", true)
const posMissingText = posMissingDot + 80
u.ctx.fillStyle = "rgb(0, 255, 0)"
u.ctx.beginPath()
u.ctx.arc(posAllocDot, posY, 3, 0, Math.PI * 2, false)
u.ctx.fill()
u.ctx.fillStyle = 'black'
u.ctx.fillText('Allocated', posAllocText, posY)
u.ctx.fillStyle = "rgb(0, 0, 255)"
u.ctx.beginPath()
u.ctx.arc(posIdleDot, posY, 3, 0, Math.PI * 2, false)
u.ctx.fill()
u.ctx.fillStyle = 'black'
u.ctx.fillText('Idle', posIdleText, posY)
u.ctx.fillStyle = "rgb(255, 0, 0)"
u.ctx.beginPath()
u.ctx.arc(posOtherDot, posY, 3, 0, Math.PI * 2, false)
u.ctx.fill()
u.ctx.fillStyle = 'black'
u.ctx.fillText('Other', posOtherText, posY)
u.ctx.fillStyle = 'black'
u.ctx.beginPath()
u.ctx.arc(posMissingDot, posY, 3, 0, Math.PI * 2, false)
u.ctx.fill()
u.ctx.fillText('Missing in DB', posMissingText, posY)
}
u.ctx.fillStyle = 'black'
u.ctx.fillText('End', posXLimit + 23, posY)
}
},
],
},
// cursor: { drag: { x: true, y: true } } // Activate zoom
plugins: [
tooltipPlugin({
onclick(u, dataIdx) {
if (jobsData) {
window.open(`/monitoring/job/${jobsData[dataIdx].id}`)
} else if (nodesData) {
window.open(`/monitoring/node/${cluster}/${nodesData[dataIdx].nodeName}`)
}
},
getLegendData: (u, dataIdx) => {
if (jobsData) {
return jobsData[dataIdx]
} else if (nodesData) {
return nodesData[dataIdx]
}
}
}),
],
};
uplot = new uPlot(opts, plotData, plotWrapper);
uplot = new uPlot(opts, roofData, plotWrapper);
} else {
// console.log("No data for roofline!");
}
@@ -365,7 +839,7 @@
/* On Mount */
onMount(() => {
render(data);
render(roofData, jobsData, nodesData);
});
/* On Destroy */
@@ -375,10 +849,8 @@
});
</script>
{#if data != null}
{#if roofData != null}
<div bind:this={plotWrapper} class="p-2"></div>
{:else}
<Card class="mx-4" body color="warning">Cannot render roofline: No data!</Card
>
<Card class="mx-4" body color="warning">Cannot render roofline: No data!</Card>
{/if}

View File

@@ -0,0 +1,384 @@
<!--
@component Roofline Model Plot based on uPlot
Properties:
- `data [null, [], []]`: Roofline Data Structure, see below for details [Default: null]
- `renderTime Bool?`: If time information should be rendered as colored dots [Default: false]
- `allowSizeChange Bool?`: If dimensions of rendered plot can change [Default: false]
- `subCluster GraphQL.SubCluster?`: SubCluster Object; contains required topology information [Default: null]
- `width Number?`: Plot width (reactively adaptive) [Default: 600]
- `height Number?`: Plot height (reactively adaptive) [Default: 380]
Data Format:
- `data = [null, [], []]`
- Index 0: null-axis required for scatter
- Index 1: Array of XY-Arrays for Scatter
- Index 2: Optional Time Info
- `data[1][0] = [100, 200, 500, ...]`
- X Axis: Intensity (Vals up to clusters' flopRateScalar value)
- `data[1][1] = [1000, 2000, 1500, ...]`
- Y Axis: Performance (Vals up to clusters' flopRateSimd value)
- `data[2] = [0.1, 0.15, 0.2, ...]`
- Color Code: Time Information (Floats from 0 to 1) (Optional)
-->
<script>
import uPlot from "uplot";
import { formatNumber } from "../units.js";
import { onMount, onDestroy } from "svelte";
import { Card } from "@sveltestrap/sveltestrap";
/* Svelte 5 Props */
let {
data = null,
renderTime = false,
allowSizeChange = false,
subCluster = null,
width = 600,
height = 380,
} = $props();
/* Const Init */
const lineWidth = clusterCockpitConfig?.plot_general_lineWidth || 2;
const cbmode = clusterCockpitConfig?.plot_general_colorblindMode || false;
/* Var Init */
let timeoutId = null;
/* State Init */
let plotWrapper = $state(null);
let uplot = $state(null);
/* Effect */
$effect(() => {
if (allowSizeChange) sizeChanged(width, height);
});
/* Functions */
// Helper
function getGradientR(x) {
if (x < 0.5) return 0;
if (x > 0.75) return 255;
x = (x - 0.5) * 4.0;
return Math.floor(x * 255.0);
}
function getGradientG(x) {
if (x > 0.25 && x < 0.75) return 255;
if (x < 0.25) x = x * 4.0;
else x = 1.0 - (x - 0.75) * 4.0;
return Math.floor(x * 255.0);
}
function getGradientB(x) {
if (x < 0.25) return 255;
if (x > 0.5) return 0;
x = 1.0 - (x - 0.25) * 4.0;
return Math.floor(x * 255.0);
}
function getRGB(c) {
return `rgb(${cbmode ? '0' : getGradientR(c)}, ${getGradientG(c)}, ${getGradientB(c)})`;
}
function nearestThousand(num) {
return Math.ceil(num / 1000) * 1000;
}
function lineIntersect(x1, y1, x2, y2, x3, y3, x4, y4) {
let l = (y4 - y3) * (x2 - x1) - (x4 - x3) * (y2 - y1);
let a = ((x4 - x3) * (y1 - y3) - (y4 - y3) * (x1 - x3)) / l;
return {
x: x1 + a * (x2 - x1),
y: y1 + a * (y2 - y1),
};
}
// Dot Renderers
const drawColorPoints = (u, seriesIdx, idx0, idx1) => {
const size = 5 * devicePixelRatio;
uPlot.orient(
u,
seriesIdx,
(
series,
dataX,
dataY,
scaleX,
scaleY,
valToPosX,
valToPosY,
xOff,
yOff,
xDim,
yDim,
moveTo,
lineTo,
rect,
arc,
) => {
let d = u.data[seriesIdx];
let deg360 = 2 * Math.PI;
for (let i = 0; i < d[0].length; i++) {
let p = new Path2D();
let xVal = d[0][i];
let yVal = d[1][i];
u.ctx.strokeStyle = getRGB(u.data[2][i]);
u.ctx.fillStyle = getRGB(u.data[2][i]);
if (
xVal >= scaleX.min &&
xVal <= scaleX.max &&
yVal >= scaleY.min &&
yVal <= scaleY.max
) {
let cx = valToPosX(xVal, scaleX, xDim, xOff);
let cy = valToPosY(yVal, scaleY, yDim, yOff);
p.moveTo(cx + size / 2, cy);
arc(p, cx, cy, size / 2, 0, deg360);
}
u.ctx.fill(p);
}
},
);
return null;
};
const drawPoints = (u, seriesIdx, idx0, idx1) => {
const size = 5 * devicePixelRatio;
uPlot.orient(
u,
seriesIdx,
(
series,
dataX,
dataY,
scaleX,
scaleY,
valToPosX,
valToPosY,
xOff,
yOff,
xDim,
yDim,
moveTo,
lineTo,
rect,
arc,
) => {
let d = u.data[seriesIdx];
u.ctx.strokeStyle = getRGB(0);
u.ctx.fillStyle = getRGB(0);
let deg360 = 2 * Math.PI;
let p = new Path2D();
for (let i = 0; i < d[0].length; i++) {
let xVal = d[0][i];
let yVal = d[1][i];
if (
xVal >= scaleX.min &&
xVal <= scaleX.max &&
yVal >= scaleY.min &&
yVal <= scaleY.max
) {
let cx = valToPosX(xVal, scaleX, xDim, xOff);
let cy = valToPosY(yVal, scaleY, yDim, yOff);
p.moveTo(cx + size / 2, cy);
arc(p, cx, cy, size / 2, 0, deg360);
}
}
u.ctx.fill(p);
},
);
return null;
};
// Main Functions
function sizeChanged() {
if (timeoutId != null) clearTimeout(timeoutId);
timeoutId = setTimeout(() => {
timeoutId = null;
if (uplot) uplot.destroy();
render(data);
}, 200);
}
function render(plotData) {
if (plotData) {
const opts = {
title: "CPU Roofline Diagram",
mode: 2,
width: width,
height: height,
legend: {
show: false,
},
cursor: { drag: { x: false, y: false } },
axes: [
{
label: "Intensity [FLOPS/Byte]",
values: (u, vals) => vals.map((v) => formatNumber(v)),
},
{
label: "Performace [GFLOPS]",
values: (u, vals) => vals.map((v) => formatNumber(v)),
},
],
scales: {
x: {
time: false,
range: [0.01, 1000],
distr: 3, // Render as log
log: 10, // log exp
},
y: {
range: [
1.0,
subCluster?.flopRateSimd?.value
? nearestThousand(subCluster.flopRateSimd.value)
: 10000,
],
distr: 3, // Render as log
log: 10, // log exp
},
},
series: [{}, { paths: renderTime ? drawColorPoints : drawPoints }],
hooks: {
drawClear: [
(u) => {
u.series.forEach((s, i) => {
if (i > 0) s._paths = null;
});
},
],
draw: [
(u) => {
// draw roofs when subCluster set
if (subCluster != null) {
const padding = u._padding; // [top, right, bottom, left]
u.ctx.strokeStyle = "black";
u.ctx.lineWidth = lineWidth;
u.ctx.beginPath();
const ycut = 0.01 * subCluster.memoryBandwidth.value;
const scalarKnee =
(subCluster.flopRateScalar.value - ycut) /
subCluster.memoryBandwidth.value;
const simdKnee =
(subCluster.flopRateSimd.value - ycut) /
subCluster.memoryBandwidth.value;
const scalarKneeX = u.valToPos(scalarKnee, "x", true), // Value, axis, toCanvasPixels
simdKneeX = u.valToPos(simdKnee, "x", true),
flopRateScalarY = u.valToPos(
subCluster.flopRateScalar.value,
"y",
true,
),
flopRateSimdY = u.valToPos(
subCluster.flopRateSimd.value,
"y",
true,
);
if (
scalarKneeX <
width * window.devicePixelRatio -
padding[1] * window.devicePixelRatio
) {
// Lower horizontal roofline
u.ctx.moveTo(scalarKneeX, flopRateScalarY);
u.ctx.lineTo(
width * window.devicePixelRatio -
padding[1] * window.devicePixelRatio,
flopRateScalarY,
);
}
if (
simdKneeX <
width * window.devicePixelRatio -
padding[1] * window.devicePixelRatio
) {
// Top horitontal roofline
u.ctx.moveTo(simdKneeX, flopRateSimdY);
u.ctx.lineTo(
width * window.devicePixelRatio -
padding[1] * window.devicePixelRatio,
flopRateSimdY,
);
}
let x1 = u.valToPos(0.01, "x", true),
y1 = u.valToPos(ycut, "y", true);
let x2 = u.valToPos(simdKnee, "x", true),
y2 = flopRateSimdY;
let xAxisIntersect = lineIntersect(
x1,
y1,
x2,
y2,
u.valToPos(0.01, "x", true),
u.valToPos(1.0, "y", true), // X-Axis Start Coords
u.valToPos(1000, "x", true),
u.valToPos(1.0, "y", true), // X-Axis End Coords
);
if (xAxisIntersect.x > x1) {
x1 = xAxisIntersect.x;
y1 = xAxisIntersect.y;
}
// Diagonal
u.ctx.moveTo(x1, y1);
u.ctx.lineTo(x2, y2);
u.ctx.stroke();
// Reset grid lineWidth
u.ctx.lineWidth = 0.15;
}
if (renderTime) {
// The Color Scale For Time Information
const posX = u.valToPos(0.1, "x", true)
const posXLimit = u.valToPos(100, "x", true)
const posY = u.valToPos(14000.0, "y", true)
u.ctx.fillStyle = 'black'
u.ctx.fillText('Start', posX, posY)
const start = posX + 10
for (let x = start; x < posXLimit; x += 10) {
let c = (x - start) / (posXLimit - start)
u.ctx.fillStyle = getRGB(c)
u.ctx.beginPath()
u.ctx.arc(x, posY, 3, 0, Math.PI * 2, false)
u.ctx.fill()
}
u.ctx.fillStyle = 'black'
u.ctx.fillText('End', posXLimit + 23, posY)
}
},
],
},
// cursor: { drag: { x: true, y: true } } // Activate zoom
};
uplot = new uPlot(opts, plotData, plotWrapper);
} else {
// console.log("No data for roofline!");
}
}
/* On Mount */
onMount(() => {
render(data);
});
/* On Destroy */
onDestroy(() => {
if (uplot) uplot.destroy();
if (timeoutId != null) clearTimeout(timeoutId);
});
</script>
{#if data != null}
<div bind:this={plotWrapper} class="p-2"></div>
{:else}
<Card class="mx-4" body color="warning">Cannot render roofline: No data!</Card
>
{/if}

View File

@@ -3,8 +3,9 @@
Properties:
- `cluster String`: Currently selected cluster
- `selectedHistograms [String]`: The currently selected metrics to display as histogram
- `ìsOpen Bool`: Is selection opened [Bindable]
- `configName String`: The config id string to be updated in database on selection change
- `presetSelectedHistograms [String]`: The currently selected metrics to display as histogram
- `applyChange Func`: The callback function to apply current selection
-->
@@ -25,6 +26,7 @@
let {
cluster,
isOpen = $bindable(),
configName,
presetSelectedHistograms,
applyChange
} = $props();
@@ -67,8 +69,8 @@
applyChange(selectedHistograms)
updateConfiguration({
name: cluster
? `user_view_histogramMetrics:${cluster}`
: "user_view_histogramMetrics",
? `${configName}:${cluster}`
: configName,
value: selectedHistograms,
});
}

View File

@@ -96,9 +96,9 @@
function printAvailability(metric, cluster) {
const avail = globalMetrics.find((gm) => gm.name === metric)?.availability
if (!cluster) {
return avail.map((av) => av.cluster).join(',')
return avail.map((av) => av.cluster).join(', ')
} else {
return avail.find((av) => av.cluster === cluster).subClusters.join(',')
return avail.find((av) => av.cluster === cluster).subClusters.join(', ')
}
}
@@ -208,7 +208,7 @@
/>
{/if}
{metric}
<span style="float: right;">
<span style="float: right; text-align: justify;">
{printAvailability(metric, cluster)}
</span>
</li>

View File

@@ -19,7 +19,7 @@
import {
transformDataForRoofline,
} from "../generic/utils.js";
import Roofline from "../generic/plots/Roofline.svelte";
import Roofline from "../generic/plots/RooflineLegacy.svelte";
/* Svelte 5 Props */
let {

View File

@@ -5,7 +5,7 @@ import Status from './Status.root.svelte'
mount(Status, {
target: document.getElementById('svelte-app'),
props: {
cluster: infos.cluster,
presetCluster: infos.cluster,
},
context: new Map([
['cc-config', clusterCockpitConfig]

View File

@@ -0,0 +1,159 @@
<!--
@component Main cluster status view component; renders current system-usage information
Properties:
- `presetCluster String`: The cluster to show status information for
-->
<script>
import { getContext } from "svelte";
import {
Row,
Col,
Spinner,
Card,
Icon,
Button,
} from "@sveltestrap/sveltestrap";
import {
queryStore,
gql,
getContextClient,
} from "@urql/svelte";
import {
init,
convert2uplot,
} from "../generic/utils.js";
import PlotGrid from "../generic/PlotGrid.svelte";
import Histogram from "../generic/plots/Histogram.svelte";
import HistogramSelection from "../generic/select/HistogramSelection.svelte";
import Refresher from "../generic/helper/Refresher.svelte";
/* Svelte 5 Props */
let {
presetCluster
} = $props();
/* Const Init */
const { query: initq } = init();
const ccconfig = getContext("cc-config");
const client = getContextClient();
/* State Init */
let cluster = $state(presetCluster);
// Histogram
let isHistogramSelectionOpen = $state(false);
let from = $state(new Date(Date.now() - (30 * 24 * 60 * 60 * 1000))); // Simple way to retrigger GQL: Jobs Started last Month
let to = $state(new Date(Date.now()));
/* Derived */
let selectedHistograms = $derived(cluster
? ccconfig[`status_view_selectedHistograms:${cluster}`] || ( ccconfig['status_view_selectedHistograms'] || [] )
: ccconfig['status_view_selectedHistograms'] || []);
// Note: nodeMetrics are requested on configured $timestep resolution
const metricStatusQuery = $derived(queryStore({
client: client,
query: gql`
query (
$filter: [JobFilter!]!
$selectedHistograms: [String!]
) {
jobsStatistics(filter: $filter, metrics: $selectedHistograms) {
histMetrics {
metric
unit
data {
min
max
count
bin
}
}
}
}
`,
variables: {
filter: [{ state: ["running"] }, { cluster: { eq: cluster}}, {startTime: { from, to }}],
selectedHistograms: selectedHistograms,
},
}));
</script>
<!-- Loading indicators & Metric Sleect -->
<Row class="justify-content-between">
<Col class="mb-2 mb-md-0" xs="12" md="5" lg="4" xl="3">
<Button
outline
color="secondary"
onclick={() => (isHistogramSelectionOpen = true)}
>
<Icon name="bar-chart-line" /> Select Histograms
</Button>
</Col>
<Col xs="12" md="5" lg="4" xl="3">
<Refresher
initially={120}
onRefresh={() => {
from = new Date(Date.now() - (30 * 24 * 60 * 60 * 1000)); // Triggers GQL
to = new Date(Date.now());
}}
/>
</Col>
</Row>
<Row cols={1} class="text-center mt-3">
<Col>
{#if $initq.fetching || $metricStatusQuery.fetching}
<Spinner />
{:else if $initq.error}
<Card body color="danger">{$initq.error.message}</Card>
{:else}
<!-- ... -->
{/if}
</Col>
</Row>
{#if $metricStatusQuery.error}
<Row cols={1}>
<Col>
<Card body color="danger">{$metricStatusQuery.error.message}</Card>
</Col>
</Row>
{/if}
{#if $initq.data && $metricStatusQuery.data}
<!-- Selectable Stats as Histograms : Average Values of Running Jobs -->
{#if selectedHistograms}
<!-- Note: Ignore '#snippet' Error in IDE -->
{#snippet gridContent(item)}
<Histogram
data={convert2uplot(item.data)}
title="Distribution of '{item.metric}' averages"
xlabel={`${item.metric} bin maximum ${item?.unit ? `[${item.unit}]` : ``}`}
xunit={item.unit}
ylabel="Number of Jobs"
yunit="Jobs"
usesBins
/>
{/snippet}
{#key $metricStatusQuery.data.jobsStatistics[0].histMetrics}
<PlotGrid
items={$metricStatusQuery.data.jobsStatistics[0].histMetrics}
itemsPerRow={2}
{gridContent}
/>
{/key}
{/if}
{/if}
<HistogramSelection
{cluster}
bind:isOpen={isHistogramSelectionOpen}
presetSelectedHistograms={selectedHistograms}
configName="status_view_selectedHistograms"
applyChange={(newSelection) => {
selectedHistograms = [...newSelection];
}}
/>

View File

@@ -0,0 +1,580 @@
<!--
@component Main cluster status view component; renders current system-usage information
Properties:
- `presetCluster String`: The cluster to show status information for
-->
<script>
import {
Row,
Col,
Card,
CardHeader,
CardTitle,
CardBody,
Table,
Progress,
Icon,
} from "@sveltestrap/sveltestrap";
import {
queryStore,
gql,
getContextClient,
} from "@urql/svelte";
import {
init,
} from "../generic/utils.js";
import { scaleNumbers, formatTime } from "../generic/units.js";
import Refresher from "../generic/helper/Refresher.svelte";
import Roofline from "../generic/plots/Roofline.svelte";
import Pie, { colors } from "../generic/plots/Pie.svelte";
/* Svelte 5 Props */
let {
presetCluster,
useCbColors = false,
useAltColors = false,
} = $props();
/* Const Init */
const { query: initq } = init();
const client = getContextClient();
/* State Init */
let cluster = $state(presetCluster);
let pieWidth = $state(0);
let plotWidths = $state([]);
let from = $state(new Date(Date.now() - 5 * 60 * 1000));
let to = $state(new Date(Date.now()));
// Bar Gauges
let allocatedNodes = $state({});
let allocatedAccs = $state({});
let flopRate = $state({});
let flopRateUnitPrefix = $state({});
let flopRateUnitBase = $state({});
let memBwRate = $state({});
let memBwRateUnitPrefix = $state({});
let memBwRateUnitBase = $state({});
// Plain Infos
let runningJobs = $state({});
let activeUsers = $state({});
let totalAccs = $state({});
/* Derived */
// Accumulated NodeStates for Piecharts
const nodesStateCounts = $derived(queryStore({
client: client,
query: gql`
query ($filter: [NodeFilter!]) {
nodeStates(filter: $filter) {
state
count
}
}
`,
variables: {
filter: { cluster: { eq: cluster }}
},
}));
const refinedStateData = $derived.by(() => {
return $nodesStateCounts?.data?.nodeStates.filter((e) => ['allocated', 'reserved', 'idle', 'mixed','down', 'unknown'].includes(e.state))
});
const refinedHealthData = $derived.by(() => {
return $nodesStateCounts?.data?.nodeStates.filter((e) => ['full', 'partial', 'failed'].includes(e.state))
});
// Note: nodeMetrics are requested on configured $timestep resolution
// Result: The latest 5 minutes (datapoints) for each node independent of job
const statusQuery = $derived(queryStore({
client: client,
query: gql`
query (
$cluster: String!
$metrics: [String!]
$from: Time!
$to: Time!
$jobFilter: [JobFilter!]!
$nodeFilter: [NodeFilter!]!
$paging: PageRequest!
$sorting: OrderByInput!
) {
# Node 5 Minute Averages for Roofline
nodeMetrics(
cluster: $cluster
metrics: $metrics
from: $from
to: $to
) {
host
subCluster
metrics {
name
metric {
series {
statistics {
avg
}
}
}
}
}
# Running Job Metric Average for Rooflines
jobsMetricStats(filter: $jobFilter, metrics: $metrics) {
id
jobId
duration
numNodes
numAccelerators
subCluster
stats {
name
data {
avg
}
}
}
# Get Jobs for Per-Node Counts
jobs(filter: $jobFilter, order: $sorting, page: $paging) {
items {
jobId
resources {
hostname
}
}
count
}
# Only counts shared nodes once
allocatedNodes(cluster: $cluster) {
name
count
}
# Get States for Node Roofline; $sorting unused in backend: Use placeholder
nodes(filter: $nodeFilter, order: $sorting) {
count
items {
hostname
cluster
subCluster
nodeState
}
}
# totalNodes includes multiples if shared jobs
jobsStatistics(
filter: $jobFilter
page: $paging
sortBy: TOTALJOBS
groupBy: SUBCLUSTER
) {
id
totalJobs
totalUsers
totalAccs
}
}
`,
variables: {
cluster: cluster,
metrics: ["flops_any", "mem_bw"], // Fixed names for roofline and status bars
from: from.toISOString(),
to: to.toISOString(),
jobFilter: [{ state: ["running"] }, { cluster: { eq: cluster } }],
nodeFilter: { cluster: { eq: cluster }},
paging: { itemsPerPage: -1, page: 1 }, // Get all: -1
sorting: { field: "startTime", type: "col", order: "DESC" }
},
}));
/* Effects */
$effect(() => {
if ($initq.data && $statusQuery.data) {
let subClusters = $initq.data.clusters.find(
(c) => c.name == cluster,
).subClusters;
for (let subCluster of subClusters) {
// Allocations
allocatedNodes[subCluster.name] =
$statusQuery.data.allocatedNodes.find(
({ name }) => name == subCluster.name,
)?.count || 0;
allocatedAccs[subCluster.name] =
$statusQuery.data.jobsStatistics.find(
({ id }) => id == subCluster.name,
)?.totalAccs || 0;
// Infos
activeUsers[subCluster.name] =
$statusQuery.data.jobsStatistics.find(
({ id }) => id == subCluster.name,
)?.totalUsers || 0;
runningJobs[subCluster.name] =
$statusQuery.data.jobsStatistics.find(
({ id }) => id == subCluster.name,
)?.totalJobs || 0;
totalAccs[subCluster.name] =
(subCluster?.numberOfNodes * subCluster?.topology?.accelerators?.length) || null;
// Keymetrics
flopRate[subCluster.name] =
Math.floor(
sumUp($statusQuery.data.nodeMetrics, subCluster.name, "flops_any") *
100,
) / 100;
flopRateUnitPrefix[subCluster.name] = subCluster.flopRateSimd.unit.prefix;
flopRateUnitBase[subCluster.name] = subCluster.flopRateSimd.unit.base;
memBwRate[subCluster.name] =
Math.floor(
sumUp($statusQuery.data.nodeMetrics, subCluster.name, "mem_bw") * 100,
) / 100;
memBwRateUnitPrefix[subCluster.name] =
subCluster.memoryBandwidth.unit.prefix;
memBwRateUnitBase[subCluster.name] = subCluster.memoryBandwidth.unit.base;
}
}
});
/* Const Functions */
const sumUp = (data, subcluster, metric) =>
data.reduce(
(sum, node) =>
node.subCluster == subcluster
? sum +
(node.metrics
.find((m) => m.name == metric)
?.metric?.series[0]?.statistics?.avg || 0
)
: sum,
0,
);
/* Functions */
function transformJobsStatsToData(subclusterData) {
/* c will contain values from 0 to 1 representing the duration */
let data = null
const x = [], y = [], c = [], day = 86400.0
if (subclusterData) {
for (let i = 0; i < subclusterData.length; i++) {
const flopsData = subclusterData[i].stats.find((s) => s.name == "flops_any")
const memBwData = subclusterData[i].stats.find((s) => s.name == "mem_bw")
const f = flopsData.data.avg
const m = memBwData.data.avg
const d = subclusterData[i].duration / day
const intensity = f / m
if (Number.isNaN(intensity) || !Number.isFinite(intensity))
continue
x.push(intensity)
y.push(f)
// Long Jobs > 1 Day: Use max Color
if (d > 1.0) c.push(1.0)
else c.push(d)
}
} else {
console.warn("transformJobsStatsToData: metrics for 'mem_bw' and/or 'flops_any' missing!")
}
if (x.length > 0 && y.length > 0 && c.length > 0) {
data = [null, [x, y], c] // for dataformat see roofline.svelte
}
return data
}
function transformNodesStatsToData(subclusterData) {
let data = null
const x = [], y = []
if (subclusterData) {
for (let i = 0; i < subclusterData.length; i++) {
const flopsData = subclusterData[i].metrics.find((s) => s.name == "flops_any")
const memBwData = subclusterData[i].metrics.find((s) => s.name == "mem_bw")
const f = flopsData.metric.series[0].statistics.avg
const m = memBwData.metric.series[0].statistics.avg
let intensity = f / m
if (Number.isNaN(intensity) || !Number.isFinite(intensity)) {
intensity = 0.0 // Set to Float Zero: Will not show in Log-Plot (Always below render limit)
}
x.push(intensity)
y.push(f)
}
} else {
// console.warn("transformNodesStatsToData: metrics for 'mem_bw' and/or 'flops_any' missing!")
}
if (x.length > 0 && y.length > 0) {
data = [null, [x, y]] // for dataformat see roofline.svelte
}
return data
}
function transformJobsStatsToInfo(subclusterData) {
if (subclusterData) {
return subclusterData.map((sc) => { return {id: sc.id, jobId: sc.jobId, numNodes: sc.numNodes, numAcc: sc?.numAccelerators? sc.numAccelerators : 0, duration: formatTime(sc.duration)} })
} else {
console.warn("transformJobsStatsToInfo: jobInfo missing!")
return []
}
}
function transformNodesStatsToInfo(subClusterData) {
let result = [];
if (subClusterData) { // && $nodesState?.data) {
// Use Nodes as Returned from CCMS, *NOT* as saved in DB via SlurmState-API!
for (let j = 0; j < subClusterData.length; j++) {
const nodeName = subClusterData[j]?.host ? subClusterData[j].host : "unknown"
const nodeMatch = $statusQuery?.data?.nodes?.items?.find((n) => n.hostname == nodeName && n.subCluster == subClusterData[j].subCluster);
const nodeState = nodeMatch?.nodeState ? nodeMatch.nodeState : "notindb"
let numJobs = 0
if ($statusQuery?.data) {
const nodeJobs = $statusQuery?.data?.jobs?.items?.filter((job) => job.resources.find((res) => res.hostname == nodeName))
numJobs = nodeJobs?.length ? nodeJobs.length : 0
}
result.push({nodeName: nodeName, nodeState: nodeState, numJobs: numJobs})
};
};
return result
}
function legendColors(targetIdx) {
// Reuses first color if targetIdx overflows
let c;
if (useCbColors) {
c = [...colors['colorblind']];
} else if (useAltColors) {
c = [...colors['alternative']];
} else {
c = [...colors['default']];
}
return c[(c.length + targetIdx) % c.length];
}
</script>
<!-- Refresher and space for other options -->
<Row class="justify-content-end">
<Col xs="12" md="5" lg="4" xl="3">
<Refresher
initially={120}
onRefresh={() => {
from = new Date(Date.now() - 5 * 60 * 1000);
to = new Date(Date.now());
}}
/>
</Col>
</Row>
<hr/>
<!-- Node Health Pis, later Charts -->
{#if $initq.data && $nodesStateCounts.data}
<Row cols={{ lg: 4, md: 2 , sm: 1}} class="mb-3 justify-content-center">
<Col class="px-3 mt-2 mt-lg-0">
<div bind:clientWidth={pieWidth}>
{#key refinedStateData}
<h4 class="text-center">
{cluster.charAt(0).toUpperCase() + cluster.slice(1)} Node States
</h4>
<Pie
{useAltColors}
canvasId="hpcpie-slurm"
size={pieWidth * 0.55}
sliceLabel="Nodes"
quantities={refinedStateData.map(
(sd) => sd.count,
)}
entities={refinedStateData.map(
(sd) => sd.state,
)}
/>
{/key}
</div>
</Col>
<Col class="px-4 py-2">
{#key refinedStateData}
<Table>
<tr class="mb-2">
<th></th>
<th>Current State</th>
<th>Nodes</th>
</tr>
{#each refinedStateData as sd, i}
<tr>
<td><Icon name="circle-fill" style="color: {legendColors(i)};"/></td>
<td>{sd.state}</td>
<td>{sd.count}</td>
</tr>
{/each}
</Table>
{/key}
</Col>
<Col class="px-3 mt-2 mt-lg-0">
<div bind:clientWidth={pieWidth}>
{#key refinedHealthData}
<h4 class="text-center">
{cluster.charAt(0).toUpperCase() + cluster.slice(1)} Node Health
</h4>
<Pie
{useAltColors}
canvasId="hpcpie-health"
size={pieWidth * 0.55}
sliceLabel="Nodes"
quantities={refinedHealthData.map(
(sd) => sd.count,
)}
entities={refinedHealthData.map(
(sd) => sd.state,
)}
/>
{/key}
</div>
</Col>
<Col class="px-4 py-2">
{#key refinedHealthData}
<Table>
<tr class="mb-2">
<th></th>
<th>Current Health</th>
<th>Nodes</th>
</tr>
{#each refinedHealthData as hd, i}
<tr>
<td><Icon name="circle-fill" style="color: {legendColors(i)};" /></td>
<td>{hd.state}</td>
<td>{hd.count}</td>
</tr>
{/each}
</Table>
{/key}
</Col>
</Row>
{/if}
<hr/>
<!-- Gauges & Roofline per Subcluster-->
{#if $initq.data && $statusQuery.data}
{#each $initq.data.clusters.find((c) => c.name == cluster).subClusters as subCluster, i}
<Row cols={{ lg: 3, md: 1 , sm: 1}} class="mb-3 justify-content-center">
<Col class="px-3">
<Card class="h-auto mt-1">
<CardHeader>
<CardTitle class="mb-0">SubCluster "{subCluster.name}"</CardTitle>
<span>{subCluster.processorType}</span>
</CardHeader>
<CardBody>
<Table borderless>
<tr class="py-2">
<td style="font-size:x-large;">{runningJobs[subCluster.name]} Running Jobs</td>
<td colspan="2" style="font-size:x-large;">{activeUsers[subCluster.name]} Active Users</td>
</tr>
<hr class="my-1"/>
<tr class="pt-2">
<td style="font-size: large;">
Flop Rate (<span style="cursor: help;" title="Flops[Any] = (Flops[Double] x 2) + Flops[Single]">Any</span>)
</td>
<td colspan="2" style="font-size: large;">
Memory BW Rate
</td>
</tr>
<tr class="pb-2">
<td style="font-size:x-large;">
{flopRate[subCluster.name]}
{flopRateUnitPrefix[subCluster.name]}{flopRateUnitBase[subCluster.name]}
</td>
<td colspan="2" style="font-size:x-large;">
{memBwRate[subCluster.name]}
{memBwRateUnitPrefix[subCluster.name]}{memBwRateUnitBase[subCluster.name]}
</td>
</tr>
<hr class="my-1"/>
<tr class="py-2">
<th scope="col">Allocated Nodes</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={allocatedNodes[subCluster.name]}
max={subCluster.numberOfNodes}
/>
</div></td
>
<td
>{allocatedNodes[subCluster.name]} / {subCluster.numberOfNodes}
Nodes</td
>
</tr>
{#if totalAccs[subCluster.name] !== null}
<tr class="py-2">
<th scope="col">Allocated Accelerators</th>
<td style="min-width: 100px;"
><div class="col">
<Progress
value={allocatedAccs[subCluster.name]}
max={totalAccs[subCluster.name]}
/>
</div></td
>
<td
>{allocatedAccs[subCluster.name]} / {totalAccs[subCluster.name]}
Accelerators</td
>
</tr>
{/if}
</Table>
</CardBody>
</Card>
</Col>
<Col class="px-3 mt-2 mt-lg-0">
<div bind:clientWidth={plotWidths[i]}>
{#key $statusQuery?.data?.nodeMetrics}
<Roofline
useColors={true}
allowSizeChange
width={plotWidths[i] - 10}
height={300}
cluster={cluster}
subCluster={subCluster}
roofData={transformNodesStatsToData($statusQuery?.data?.nodeMetrics.filter(
(data) => data.subCluster == subCluster.name,
)
)}
nodesData={transformNodesStatsToInfo($statusQuery?.data?.nodeMetrics.filter(
(data) => data.subCluster == subCluster.name,
)
)}
/>
{/key}
</div>
</Col>
<Col class="px-3 mt-2 mt-lg-0">
<div bind:clientWidth={plotWidths[i]}>
{#key $statusQuery?.data?.jobsMetricStats}
<Roofline
useColors={true}
allowSizeChange
width={plotWidths[i] - 10}
height={300}
subCluster={subCluster}
roofData={transformJobsStatsToData($statusQuery?.data?.jobsMetricStats.filter(
(data) => data.subCluster == subCluster.name,
)
)}
jobsData={transformJobsStatsToInfo($statusQuery?.data?.jobsMetricStats.filter(
(data) => data.subCluster == subCluster.name,
)
)}
/>
{/key}
</div>
</Col>
</Row>
{/each}
{:else}
<Card class="mx-4" body color="warning">Cannot render status rooflines: No data!</Card>
{/if}

View File

@@ -0,0 +1,547 @@
<!--
@component Main cluster status view component; renders current system-usage information
Properties:
- `presetCluster String`: The cluster to show status information for
-->
<script>
import {
Row,
Col,
Spinner,
Card,
Table,
Icon,
Tooltip,
Input,
InputGroup,
InputGroupText
} from "@sveltestrap/sveltestrap";
import {
queryStore,
gql,
getContextClient,
} from "@urql/svelte";
import {
init,
scramble,
scrambleNames,
convert2uplot,
} from "../generic/utils.js";
import Pie, { colors } from "../generic/plots/Pie.svelte";
import Histogram from "../generic/plots/Histogram.svelte";
import Refresher from "../generic/helper/Refresher.svelte";
/* Svelte 5 Props */
let {
presetCluster,
useCbColors = false,
useAltColors = false
} = $props();
/* Const Init */
const { query: initq } = init();
const client = getContextClient();
const durationBinOptions = ["1m","10m","1h","6h","12h"];
/* State Init */
let cluster = $state(presetCluster)
let from = $state(new Date(Date.now() - (30 * 24 * 60 * 60 * 1000))); // Simple way to retrigger GQL: Jobs Started last Month
let to = $state(new Date(Date.now()));
let colWidthJobs = $state(0);
let colWidthNodes = $state(0);
let colWidthAccs = $state(0);
let numDurationBins = $state("1h");
/* Derived */
const topJobsQuery = $derived(queryStore({
client: client,
query: gql`
query (
$filter: [JobFilter!]!
$paging: PageRequest!
) {
topUser: jobsStatistics(
filter: $filter
page: $paging
sortBy: TOTALJOBS
groupBy: USER
) {
id
name
totalJobs
}
topProjects: jobsStatistics(
filter: $filter
page: $paging
sortBy: TOTALJOBS
groupBy: PROJECT
) {
id
totalJobs
}
}
`,
variables: {
filter: [{ state: ["running"] }, { cluster: { eq: cluster}}, {startTime: { from, to }}],
paging: { itemsPerPage: 10, page: 1 } // Top 10
},
}));
const topNodesQuery = $derived(queryStore({
client: client,
query: gql`
query (
$filter: [JobFilter!]!
$paging: PageRequest!
) {
topUser: jobsStatistics(
filter: $filter
page: $paging
sortBy: TOTALNODES
groupBy: USER
) {
id
name
totalNodes
}
topProjects: jobsStatistics(
filter: $filter
page: $paging
sortBy: TOTALNODES
groupBy: PROJECT
) {
id
totalNodes
}
}
`,
variables: {
filter: [{ state: ["running"] }, { cluster: { eq: cluster }}, {startTime: { from, to }}],
paging: { itemsPerPage: 10, page: 1 } // Top 10
},
}));
const topAccsQuery = $derived(queryStore({
client: client,
query: gql`
query (
$filter: [JobFilter!]!
$paging: PageRequest!
) {
topUser: jobsStatistics(
filter: $filter
page: $paging
sortBy: TOTALACCS
groupBy: USER
) {
id
name
totalAccs
}
topProjects: jobsStatistics(
filter: $filter
page: $paging
sortBy: TOTALACCS
groupBy: PROJECT
) {
id
totalAccs
}
}
`,
variables: {
filter: [{ state: ["running"] }, { cluster: { eq: cluster }}, {startTime: { from, to }}],
paging: { itemsPerPage: 10, page: 1 } // Top 10
},
}));
// Note: nodeMetrics are requested on configured $timestep resolution
const nodeStatusQuery = $derived(queryStore({
client: client,
query: gql`
query (
$filter: [JobFilter!]!
$selectedHistograms: [String!]
$numDurationBins: String
) {
jobsStatistics(filter: $filter, metrics: $selectedHistograms, numDurationBins: $numDurationBins) {
histDuration {
count
value
}
histNumNodes {
count
value
}
histNumAccs {
count
value
}
}
}
`,
variables: {
filter: [{ state: ["running"] }, { cluster: { eq: cluster }}, {startTime: { from, to }}],
selectedHistograms: [], // No Metrics requested for node hardware stats
numDurationBins: numDurationBins,
},
}));
/* Functions */
function legendColors(targetIdx) {
// Reuses first color if targetIdx overflows
let c;
if (useCbColors) {
c = [...colors['colorblind']];
} else if (useAltColors) {
c = [...colors['alternative']];
} else {
c = [...colors['default']];
}
return c[(c.length + targetIdx) % c.length];
}
</script>
<!-- Refresher and space for other options -->
<Row class="justify-content-between">
<Col class="mb-2 mb-md-0" xs="12" md="5" lg="4" xl="3">
<InputGroup>
<InputGroupText>
<Icon name="bar-chart-line-fill" />
</InputGroupText>
<InputGroupText>
Duration Bin Size
</InputGroupText>
<Input type="select" bind:value={numDurationBins}>
{#each durationBinOptions as dbin}
<option value={dbin}>{dbin}</option>
{/each}
</Input>
</InputGroup>
</Col>
<Col xs="12" md="5" lg="4" xl="3">
<Refresher
initially={120}
onRefresh={() => {
from = new Date(Date.now() - (30 * 24 * 60 * 60 * 1000)); // Triggers GQL
to = new Date(Date.now());
}}
/>
</Col>
</Row>
<hr/>
<!-- Job Duration, Top Users and Projects-->
{#if $topJobsQuery.fetching || $nodeStatusQuery.fetching}
<Spinner />
{:else if $topJobsQuery.data && $nodeStatusQuery.data}
<Row>
<Col xs="12" lg="4" class="p-2">
{#key $nodeStatusQuery.data.jobsStatistics[0].histDuration}
<Histogram
data={convert2uplot($nodeStatusQuery.data.jobsStatistics[0].histDuration)}
title="Duration Distribution"
xlabel="Current Job Runtimes"
xunit="Runtime"
ylabel="Number of Jobs"
yunit="Jobs"
height="275"
usesBins
xtime
/>
{/key}
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<div bind:clientWidth={colWidthJobs}>
<h4 class="text-center">
Top Users: Jobs
</h4>
<Pie
{useAltColors}
canvasId="hpcpie-jobs-users"
size={colWidthJobs * 0.75}
sliceLabel="Jobs"
quantities={$topJobsQuery.data.topUser.map(
(tu) => tu['totalJobs'],
)}
entities={$topJobsQuery.data.topUser.map((tu) => scrambleNames ? scramble(tu.id) : tu.id)}
/>
</div>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<Table>
<tr class="mb-2">
<th></th>
<th style="padding-left: 0.5rem;">User</th>
<th>Jobs</th>
</tr>
{#each $topJobsQuery.data.topUser as tu, i}
<tr>
<td><Icon name="circle-fill" style="color: {legendColors(i)};" /></td>
<td id="topName-jobs-{tu.id}">
<a target="_blank" href="/monitoring/user/{tu.id}?cluster={cluster}&state=running"
>{scrambleNames ? scramble(tu.id) : tu.id}
</a>
</td>
{#if tu?.name}
<Tooltip
target={`topName-jobs-${tu.id}`}
placement="left"
>{scrambleNames ? scramble(tu.name) : tu.name}</Tooltip
>
{/if}
<td>{tu['totalJobs']}</td>
</tr>
{/each}
</Table>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<h4 class="text-center">
Top Projects: Jobs
</h4>
<Pie
{useAltColors}
canvasId="hpcpie-jobs-projects"
size={colWidthJobs * 0.75}
sliceLabel={'Jobs'}
quantities={$topJobsQuery.data.topProjects.map(
(tp) => tp['totalJobs'],
)}
entities={$topJobsQuery.data.topProjects.map((tp) => scrambleNames ? scramble(tp.id) : tp.id)}
/>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<Table>
<tr class="mb-2">
<th></th>
<th style="padding-left: 0.5rem;">Project</th>
<th>Jobs</th>
</tr>
{#each $topJobsQuery.data.topProjects as tp, i}
<tr>
<td><Icon name="circle-fill" style="color: {legendColors(i)};" /></td>
<td>
<a target="_blank" href="/monitoring/jobs/?cluster={cluster}&state=running&project={tp.id}&projectMatch=eq"
>{scrambleNames ? scramble(tp.id) : tp.id}
</a>
</td>
<td>{tp['totalJobs']}</td>
</tr>
{/each}
</Table>
</Col>
</Row>
{:else}
<Card class="mx-4" body color="warning">Cannot render job status charts: No data!</Card>
{/if}
<hr/>
<!-- Node Distribution, Top Users and Projects-->
{#if $topNodesQuery.fetching || $nodeStatusQuery.fetching}
<Spinner />
{:else if $topNodesQuery.data && $nodeStatusQuery.data}
<Row>
<Col xs="12" lg="4" class="p-2">
<Histogram
data={convert2uplot($nodeStatusQuery.data.jobsStatistics[0].histNumNodes)}
title="Number of Nodes Distribution"
xlabel="Allocated Nodes"
xunit="Nodes"
ylabel="Number of Jobs"
yunit="Jobs"
height="275"
/>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<div bind:clientWidth={colWidthNodes}>
<h4 class="text-center">
Top Users: Nodes
</h4>
<Pie
{useAltColors}
canvasId="hpcpie-nodes-users"
size={colWidthNodes * 0.75}
sliceLabel="Nodes"
quantities={$topNodesQuery.data.topUser.map(
(tu) => tu['totalNodes'],
)}
entities={$topNodesQuery.data.topUser.map((tu) => scrambleNames ? scramble(tu.id) : tu.id)}
/>
</div>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<Table>
<tr class="mb-2">
<th></th>
<th style="padding-left: 0.5rem;">User</th>
<th>Nodes</th>
</tr>
{#each $topNodesQuery.data.topUser as tu, i}
<tr>
<td><Icon name="circle-fill" style="color: {legendColors(i)};" /></td>
<td id="topName-nodes-{tu.id}">
<a target="_blank" href="/monitoring/user/{tu.id}?cluster={cluster}&state=running"
>{scrambleNames ? scramble(tu.id) : tu.id}
</a>
</td>
{#if tu?.name}
<Tooltip
target={`topName-nodes-${tu.id}`}
placement="left"
>{scrambleNames ? scramble(tu.name) : tu.name}</Tooltip
>
{/if}
<td>{tu['totalNodes']}</td>
</tr>
{/each}
</Table>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<h4 class="text-center">
Top Projects: Nodes
</h4>
<Pie
{useAltColors}
canvasId="hpcpie-nodes-projects"
size={colWidthNodes * 0.75}
sliceLabel={'Nodes'}
quantities={$topNodesQuery.data.topProjects.map(
(tp) => tp['totalNodes'],
)}
entities={$topNodesQuery.data.topProjects.map((tp) => scrambleNames ? scramble(tp.id) : tp.id)}
/>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<Table>
<tr class="mb-2">
<th></th>
<th style="padding-left: 0.5rem;">Project</th>
<th>Nodes</th>
</tr>
{#each $topNodesQuery.data.topProjects as tp, i}
<tr>
<td><Icon name="circle-fill" style="color: {legendColors(i)};" /></td>
<td>
<a target="_blank" href="/monitoring/jobs/?cluster={cluster}&state=running&project={tp.id}&projectMatch=eq"
>{scrambleNames ? scramble(tp.id) : tp.id}
</a>
</td>
<td>{tp['totalNodes']}</td>
</tr>
{/each}
</Table>
</Col>
</Row>
{:else}
<Card class="mx-4" body color="warning">Cannot render node status charts: No data!</Card>
{/if}
<hr/>
<!-- Acc Distribution, Top Users and Projects-->
{#if $topAccsQuery.fetching || $nodeStatusQuery.fetching}
<Spinner />
{:else if $topAccsQuery.data && $nodeStatusQuery.data}
<Row>
<Col xs="12" lg="4" class="p-2">
<Histogram
data={convert2uplot($nodeStatusQuery.data.jobsStatistics[0].histNumAccs)}
title="Number of Accelerators Distribution"
xlabel="Allocated Accs"
xunit="Accs"
ylabel="Number of Jobs"
yunit="Jobs"
height="275"
/>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<div bind:clientWidth={colWidthAccs}>
<h4 class="text-center">
Top Users: GPUs
</h4>
<Pie
{useAltColors}
canvasId="hpcpie-accs-users"
size={colWidthAccs * 0.75}
sliceLabel="GPUs"
quantities={$topAccsQuery.data.topUser.map(
(tu) => tu['totalAccs'],
)}
entities={$topAccsQuery.data.topUser.map((tu) => scrambleNames ? scramble(tu.id) : tu.id)}
/>
</div>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<Table>
<tr class="mb-2">
<th></th>
<th style="padding-left: 0.5rem;">User</th>
<th>GPUs</th>
</tr>
{#each $topAccsQuery.data.topUser as tu, i}
<tr>
<td><Icon name="circle-fill" style="color: {legendColors(i)};" /></td>
<td id="topName-accs-{tu.id}">
<a target="_blank" href="/monitoring/user/{tu.id}?cluster={cluster}&state=running"
>{scrambleNames ? scramble(tu.id) : tu.id}
</a>
</td>
{#if tu?.name}
<Tooltip
target={`topName-accs-${tu.id}`}
placement="left"
>{scrambleNames ? scramble(tu.name) : tu.name}</Tooltip
>
{/if}
<td>{tu['totalAccs']}</td>
</tr>
{/each}
</Table>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<h4 class="text-center">
Top Projects: GPUs
</h4>
<Pie
{useAltColors}
canvasId="hpcpie-accs-projects"
size={colWidthAccs * 0.75}
sliceLabel={'GPUs'}
quantities={$topAccsQuery.data.topProjects.map(
(tp) => tp['totalAccs'],
)}
entities={$topAccsQuery.data.topProjects.map((tp) => scrambleNames ? scramble(tp.id) : tp.id)}
/>
</Col>
<Col xs="6" md="3" lg="2" class="p-2">
<Table>
<tr class="mb-2">
<th></th>
<th style="padding-left: 0.5rem;">Project</th>
<th>GPUs</th>
</tr>
{#each $topAccsQuery.data.topProjects as tp, i}
<tr>
<td><Icon name="circle-fill" style="color: {legendColors(i)};" /></td>
<td>
<a target="_blank" href="/monitoring/jobs/?cluster={cluster}&state=running&project={tp.id}&projectMatch=eq"
>{scrambleNames ? scramble(tp.id) : tp.id}
</a>
</td>
<td>{tp['totalAccs']}</td>
</tr>
{/each}
</Table>
</Col>
</Row>
{:else}
<Card class="mx-4" body color="warning">Cannot render accelerator status charts: No data!</Card>
{/if}