Skip to content

Commit

Permalink
Add Metrics for Running PipelinesRuns at Pipeline and Namespace level
Browse files Browse the repository at this point in the history 
Currently metrics shown for Running Pipeline Count is at cluster level.
There is no way we can get that metric at namespace or pipeline level.
We have added the PipelineRun metric at pipeline and namespace level.
Level of PipelineRun Can be configured via ConfigMap. Default is "" . but it can be set to namespace or Pipelinerun.
  • Loading branch information
@pramodbindal
pramodbindal committed Oct 3, 2024
1 parent 0649270 commit d9c0b48
Show file tree
Hide file tree
Showing 8 changed files with 152 additions and 18 deletions.
1 change: 1 addition & 0 deletions config/config-observability.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -59,3 +59,4 @@ data:
metrics.pipelinerun.level: "pipeline"
metrics.pipelinerun.duration-type: "histogram"
metrics.count.enable-reason: "false"
metrics.running-pipelinerun.level: ""
31 changes: 18 additions & 13 deletions docs/metrics.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -41,26 +41,31 @@ A sample config-map has been provided as [config-observability](./../config/conf
metrics.taskrun.level: "task"
metrics.taskrun.duration-type: "histogram"
metrics.pipelinerun.level: "pipeline"
metrics.running-pipelinerun.level: ""
metrics.pipelinerun.duration-type: "histogram"
metrics.count.enable-reason: "false"
```
Following values are available in the configmap:
| configmap data | value | description |
| -- | ----------- | ----------- |
| metrics.taskrun.level | `taskrun` | Level of metrics is taskrun |
| metrics.taskrun.level | `task` | Level of metrics is task and taskrun label isn't present in the metrics |
| metrics.taskrun.level | `namespace` | Level of metrics is namespace, and task and taskrun label isn't present in the metrics
| metrics.pipelinerun.level | `pipelinerun` | Level of metrics is pipelinerun |
| metrics.pipelinerun.level | `pipeline` | Level of metrics is pipeline and pipelinerun label isn't present in the metrics |
| metrics.pipelinerun.level | `namespace` | Level of metrics is namespace, pipeline and pipelinerun label isn't present in the metrics |
| metrics.taskrun.duration-type | `histogram` | `tekton_pipelines_controller_pipelinerun_taskrun_duration_seconds` and `tekton_pipelines_controller_taskrun_duration_seconds` is of type histogram |
| configmap data | value | description |
| -- | ----------- |--------------------------------------------------------------------------------------------------------------------------------------------------------------|
| metrics.taskrun.level | `taskrun` | Level of metrics is taskrun |
| metrics.taskrun.level | `task` | Level of metrics is task and taskrun label isn't present in the metrics |
| metrics.taskrun.level | `namespace` | Level of metrics is namespace, and task and taskrun label isn't present in the metrics
| metrics.pipelinerun.level | `pipelinerun` | Level of metrics is pipelinerun |
| metrics.pipelinerun.level | `pipeline` | Level of metrics is pipeline and pipelinerun label isn't present in the metrics |
| metrics.pipelinerun.level | `namespace` | Level of metrics is namespace, pipeline and pipelinerun label isn't present in the metrics |
| metrics.running-pipelinerun.level | `pipelinerun` | Level of running-pipelinerun metrics is pipelinerun |
| metrics.running-pipelinerun.level | `pipeline` | Level of running-pipelinerun metrics is pipeline and pipelinerun label isn't present in the metrics |
| metrics.running-pipelinerun.level | `namespace` | Level of running-pipelinerun metrics is namespace, pipeline and pipelinerun label isn't present in the metrics |
| metrics.running-pipelinerun.level | `` | Level of running-pipelinerun metrics is cluster, namespace, pipeline and pipelinerun label isn't present in the metrics. |
| metrics.taskrun.duration-type | `histogram` | `tekton_pipelines_controller_pipelinerun_taskrun_duration_seconds` and `tekton_pipelines_controller_taskrun_duration_seconds` is of type histogram |
| metrics.taskrun.duration-type | `lastvalue` | `tekton_pipelines_controller_pipelinerun_taskrun_duration_seconds` and `tekton_pipelines_controller_taskrun_duration_seconds` is of type gauge or lastvalue |
| metrics.pipelinerun.duration-type | `histogram` | `tekton_pipelines_controller_pipelinerun_duration_seconds` is of type histogram |
| metrics.pipelinerun.duration-type | `lastvalue` | `tekton_pipelines_controller_pipelinerun_duration_seconds` is of type gauge or lastvalue |
| metrics.count.enable-reason | `false` | Sets if the `reason` label should be included on count metrics |
| metrics.taskrun.throttle.enable-namespace | `false` | Sets if the `namespace` label should be included on the `tekton_pipelines_controller_running_taskruns_throttled_by_quota` metric |
| metrics.pipelinerun.duration-type | `histogram` | `tekton_pipelines_controller_pipelinerun_duration_seconds` is of type histogram |
| metrics.pipelinerun.duration-type | `lastvalue` | `tekton_pipelines_controller_pipelinerun_duration_seconds` is of type gauge or lastvalue |
| metrics.count.enable-reason | `false` | Sets if the `reason` label should be included on count metrics |
| metrics.taskrun.throttle.enable-namespace | `false` | Sets if the `namespace` label should be included on the `tekton_pipelines_controller_running_taskruns_throttled_by_quota` metric |

Histogram value isn't available when pipelinerun or taskrun labels are selected. The Lastvalue or Gauge will be provided. Histogram would serve no purpose because it would generate a single bar. TaskRun and PipelineRun level metrics aren't recommended because they lead to an unbounded cardinality which degrades the observability database.

Expand Down
11 changes: 11 additions & 0 deletions pkg/apis/config/metrics.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -29,6 +29,9 @@ const (
// metricsPipelinerunLevel determines to what level to aggregate metrics
// for pipelinerun
metricsPipelinerunLevelKey = "metrics.pipelinerun.level"
// metricsRunningPipelinerunLevelKey determines to what level to aggregate metrics
// for running pipelineruns
metricsRunningPipelinerunLevelKey = "metrics.running-pipelinerun.level"
// metricsDurationTaskrunType determines what type of
// metrics to use for aggregating duration for taskrun
metricsDurationTaskrunType = "metrics.taskrun.duration-type"
Expand All @@ -55,6 +58,9 @@ const (
// DefaultPipelinerunLevel determines to what level to aggregate metrics
// when it isn't specified in configmap
DefaultPipelinerunLevel = PipelinerunLevelAtPipeline
// DefaultRunningPipelinerunLevel determines to what level to aggregate metrics
// when it isn't specified in configmap
DefaultRunningPipelinerunLevel = ""
// PipelinerunLevelAtPipelinerun specify that aggregation will be done at
// pipelinerun level
PipelinerunLevelAtPipelinerun = "pipelinerun"
Expand Down Expand Up @@ -96,6 +102,7 @@ var DefaultMetrics, _ = newMetricsFromMap(map[string]string{})
type Metrics struct {
TaskrunLevel string
PipelinerunLevel string
RunningPipelinerunLevel string
DurationTaskrunType string
DurationPipelinerunType string
CountWithReason bool
Expand Down Expand Up @@ -130,6 +137,7 @@ func newMetricsFromMap(cfgMap map[string]string) (*Metrics, error) {
tc := Metrics{
TaskrunLevel: DefaultTaskrunLevel,
PipelinerunLevel: DefaultPipelinerunLevel,
RunningPipelinerunLevel: DefaultRunningPipelinerunLevel,
DurationTaskrunType: DefaultDurationTaskrunType,
DurationPipelinerunType: DefaultDurationPipelinerunType,
CountWithReason: false,
Expand All @@ -143,6 +151,9 @@ func newMetricsFromMap(cfgMap map[string]string) (*Metrics, error) {
if pipelinerunLevel, ok := cfgMap[metricsPipelinerunLevelKey]; ok {
tc.PipelinerunLevel = pipelinerunLevel
}
if runningPipelinerunLevel, ok := cfgMap[metricsRunningPipelinerunLevelKey]; ok {
tc.RunningPipelinerunLevel = runningPipelinerunLevel
}
if durationTaskrun, ok := cfgMap[metricsDurationTaskrunType]; ok {
tc.DurationTaskrunType = durationTaskrun
}
Expand Down
5 changes: 5 additions & 0 deletions pkg/apis/config/metrics_test.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -36,6 +36,7 @@ func TestNewMetricsFromConfigMap(t *testing.T) {
expectedConfig: &config.Metrics{
TaskrunLevel: config.TaskrunLevelAtTaskrun,
PipelinerunLevel: config.PipelinerunLevelAtPipelinerun,
RunningPipelinerunLevel: config.DefaultRunningPipelinerunLevel,
DurationTaskrunType: config.DurationPipelinerunTypeHistogram,
DurationPipelinerunType: config.DurationPipelinerunTypeHistogram,
CountWithReason: false,
Expand All @@ -47,6 +48,7 @@ func TestNewMetricsFromConfigMap(t *testing.T) {
expectedConfig: &config.Metrics{
TaskrunLevel: config.TaskrunLevelAtNS,
PipelinerunLevel: config.PipelinerunLevelAtNS,
RunningPipelinerunLevel: config.PipelinerunLevelAtNS,
DurationTaskrunType: config.DurationTaskrunTypeHistogram,
DurationPipelinerunType: config.DurationPipelinerunTypeLastValue,
CountWithReason: false,
Expand All @@ -58,6 +60,7 @@ func TestNewMetricsFromConfigMap(t *testing.T) {
expectedConfig: &config.Metrics{
TaskrunLevel: config.TaskrunLevelAtNS,
PipelinerunLevel: config.PipelinerunLevelAtNS,
RunningPipelinerunLevel: config.DefaultRunningPipelinerunLevel,
DurationTaskrunType: config.DurationTaskrunTypeHistogram,
DurationPipelinerunType: config.DurationPipelinerunTypeLastValue,
CountWithReason: true,
Expand All @@ -69,6 +72,7 @@ func TestNewMetricsFromConfigMap(t *testing.T) {
expectedConfig: &config.Metrics{
TaskrunLevel: config.TaskrunLevelAtNS,
PipelinerunLevel: config.PipelinerunLevelAtNS,
RunningPipelinerunLevel: config.PipelinerunLevelAtPipeline,
DurationTaskrunType: config.DurationTaskrunTypeHistogram,
DurationPipelinerunType: config.DurationPipelinerunTypeLastValue,
CountWithReason: true,
Expand All @@ -88,6 +92,7 @@ func TestNewMetricsFromEmptyConfigMap(t *testing.T) {
expectedConfig := &config.Metrics{
TaskrunLevel: config.TaskrunLevelAtTask,
PipelinerunLevel: config.PipelinerunLevelAtPipeline,
RunningPipelinerunLevel: config.DefaultRunningPipelinerunLevel,
DurationTaskrunType: config.DurationPipelinerunTypeHistogram,
DurationPipelinerunType: config.DurationPipelinerunTypeHistogram,
CountWithReason: false,
Expand Down
1 change: 1 addition & 0 deletions pkg/apis/config/testdata/config-observability-namespacelevel.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -27,4 +27,5 @@ data:
metrics.taskrun.level: "namespace"
metrics.taskrun.duration-type: "histogram"
metrics.pipelinerun.level: "namespace"
metrics.running-pipelinerun.level: "namespace"
metrics.pipelinerun.duration-type: "lastvalue"
1 change: 1 addition & 0 deletions pkg/apis/config/testdata/config-observability-throttle.yaml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,7 @@ data:
metrics.taskrun.level: "namespace"
metrics.taskrun.duration-type: "histogram"
metrics.pipelinerun.level: "namespace"
metrics.running-pipelinerun.level: "pipeline"
metrics.pipelinerun.duration-type: "lastvalue"
metrics.count.enable-reason: "true"
metrics.taskrun.throttle.enable-namespace: "true"
62 changes: 57 additions & 5 deletions pkg/pipelinerunmetrics/metrics.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -39,6 +39,13 @@ import (
"knative.dev/pkg/metrics"
)

const (
runningPRLevelPipelinerun = "pipelinerun"
runningPRLevelPipeline = "pipeline"
runningPRLevelNamespace = "namespace"
runningPRLevelBlank = ""
)

var (
pipelinerunTag = tag.MustNewKey("pipelinerun")
pipelineTag = tag.MustNewKey("pipeline")
Expand Down Expand Up @@ -149,7 +156,6 @@ func viewRegister(cfg *config.Metrics) error {
defer r.mutex.Unlock()

var prunTag []tag.Key

switch cfg.PipelinerunLevel {
case config.PipelinerunLevelAtPipelinerun:
prunTag = []tag.Key{pipelinerunTag, pipelineTag}
Expand All @@ -164,6 +170,18 @@ func viewRegister(cfg *config.Metrics) error {
return errors.New("invalid config for PipelinerunLevel: " + cfg.PipelinerunLevel)
}

var runningPRTag []tag.Key
switch cfg.RunningPipelinerunLevel {
case config.PipelinerunLevelAtPipelinerun:
runningPRTag = []tag.Key{pipelinerunTag, pipelineTag, namespaceTag}
case config.PipelinerunLevelAtPipeline:
runningPRTag = []tag.Key{pipelineTag, namespaceTag}
case config.PipelinerunLevelAtNS:
runningPRTag = []tag.Key{namespaceTag}
default:
runningPRTag = []tag.Key{}
}

distribution := view.Distribution(10, 30, 60, 300, 900, 1800, 3600, 5400, 10800, 21600, 43200, 86400)

if cfg.PipelinerunLevel == config.PipelinerunLevelAtPipelinerun {
Expand Down Expand Up @@ -213,6 +231,7 @@ func viewRegister(cfg *config.Metrics) error {
Description: runningPRs.Description(),
Measure: runningPRs,
Aggregation: view.LastValue(),
TagKeys: runningPRTag,
}

runningPRsWaitingOnPipelineResolutionCountView = &view.View{
Expand Down Expand Up @@ -326,7 +345,7 @@ func (r *Recorder) updateConfig(cfg *config.Metrics) {

// DurationAndCount logs the duration of PipelineRun execution and
// count for number of PipelineRuns succeed or failed
// returns an error if its failed to log the metrics
// returns an error if it fails to log the metrics
func (r *Recorder) DurationAndCount(pr *v1.PipelineRun, beforeCondition *apis.Condition) error {
if !r.initialized {
return fmt.Errorf("ignoring the metrics recording for %s , failed to initialize the metrics recorder", pr.Name)
Expand Down Expand Up @@ -379,11 +398,10 @@ func (r *Recorder) DurationAndCount(pr *v1.PipelineRun, beforeCondition *apis.Co
}

// RunningPipelineRuns logs the number of PipelineRuns running right now
// returns an error if its failed to log the metrics
// returns an error if it fails to log the metrics
func (r *Recorder) RunningPipelineRuns(lister listers.PipelineRunLister) error {
r.mutex.Lock()
defer r.mutex.Unlock()

if !r.initialized {
return errors.New("ignoring the metrics recording, failed to initialize the metrics recorder")
}
Expand All @@ -396,9 +414,37 @@ func (r *Recorder) RunningPipelineRuns(lister listers.PipelineRunLister) error {
var runningPipelineRuns int
var trsWaitResolvingTaskRef int
var prsWaitResolvingPipelineRef int
countMap := map[string]int{}

for _, pr := range prs {
pipelineName := getPipelineTagName(pr)
pipelineRunKey := ""
mutators := []tag.Mutator{
tag.Insert(namespaceTag, pr.Namespace),
tag.Insert(pipelineTag, pipelineName),
}
if r.cfg != nil {
switch r.cfg.RunningPipelinerunLevel {
case runningPRLevelPipelinerun:
pipelineRunKey = pipelineRunKey + "#" + pr.Name
fallthrough
case runningPRLevelPipeline:
pipelineRunKey = pipelineRunKey + "#" + pipelineName
fallthrough
case runningPRLevelNamespace:
pipelineRunKey = pipelineRunKey + "#" + pr.Namespace
case runningPRLevelBlank:
default:
return fmt.Errorf("RunningPipelineRunLevel value \"%s\" is not valid ", r.cfg.RunningPipelinerunLevel)
}
}
ctx_, err_ := tag.New(context.Background(), mutators...)
if err_ != nil {
return err
}
if !pr.IsDone() {
countMap[pipelineRunKey]++
metrics.Record(ctx_, runningPRs.M(float64(countMap[pipelineRunKey])))
runningPipelineRuns++
succeedCondition := pr.Status.GetCondition(apis.ConditionSucceeded)
if succeedCondition != nil && succeedCondition.Status == corev1.ConditionUnknown {
Expand All @@ -409,6 +455,13 @@ func (r *Recorder) RunningPipelineRuns(lister listers.PipelineRunLister) error {
prsWaitResolvingPipelineRef++
}
}
} else {
// If all PipelineRuns for a Pipeline are completed then record it with 0
// set the key to 0 so that we do not record further completed PipelineRuns for same Pipeline
if _, exists := countMap[pipelineRunKey]; !exists {
countMap[pipelineRunKey] = 0
metrics.Record(ctx_, runningPRs.M(0))
}
}
}

Expand All @@ -421,7 +474,6 @@ func (r *Recorder) RunningPipelineRuns(lister listers.PipelineRunLister) error {
metrics.Record(ctx, runningPRsWaitingOnTaskResolutionCount.M(float64(trsWaitResolvingTaskRef)))
metrics.Record(ctx, runningPRsWaitingOnTaskResolution.M(float64(trsWaitResolvingTaskRef)))
metrics.Record(ctx, runningPRsCount.M(float64(runningPipelineRuns)))
metrics.Record(ctx, runningPRs.M(float64(runningPipelineRuns)))

return nil
}
Expand Down
58 changes: 58 additions & 0 deletions pkg/pipelinerunmetrics/metrics_test.go
View file
Original file line number Diff line number Diff line change
Expand Up @@ -50,6 +50,7 @@ func getConfigContext(countWithReason bool) context.Context {
Metrics: &config.Metrics{
TaskrunLevel: config.TaskrunLevelAtTaskrun,
PipelinerunLevel: config.PipelinerunLevelAtPipelinerun,
RunningPipelinerunLevel: config.DefaultRunningPipelinerunLevel,
DurationTaskrunType: config.DefaultDurationTaskrunType,
DurationPipelinerunType: config.DefaultDurationPipelinerunType,
CountWithReason: countWithReason,
Expand All @@ -58,6 +59,21 @@ func getConfigContext(countWithReason bool) context.Context {
return config.ToContext(ctx, cfg)
}

func getConfigContextRunningPRLevel(runningPipelinerunLevel string) context.Context {
ctx := context.Background()
cfg := &config.Config{
Metrics: &config.Metrics{
TaskrunLevel: config.TaskrunLevelAtTaskrun,
PipelinerunLevel: config.PipelinerunLevelAtPipelinerun,
DurationTaskrunType: config.DefaultDurationTaskrunType,
DurationPipelinerunType: config.DefaultDurationPipelinerunType,
CountWithReason: false,
RunningPipelinerunLevel: runningPipelinerunLevel,
},
}
return config.ToContext(ctx, cfg)
}

func TestUninitializedMetrics(t *testing.T) {
metrics := Recorder{}

Expand Down Expand Up @@ -504,6 +520,48 @@ func TestRecordRunningPipelineRunsCount(t *testing.T) {
metricstest.CheckLastValueData(t, "running_pipelineruns", map[string]string{}, 1)
}

func TestRecordRunningPipelineRunsCountAtPipelineLevel(t *testing.T) {
unregisterMetrics()

newPipelineRun := func(status corev1.ConditionStatus, namespace string) *v1.PipelineRun {
return &v1.PipelineRun{
ObjectMeta: metav1.ObjectMeta{Name: names.SimpleNameGenerator.RestrictLengthWithRandomSuffix("pipelinerun-"), Namespace: namespace},
Status: v1.PipelineRunStatus{
Status: duckv1.Status{
Conditions: duckv1.Conditions{{
Type: apis.ConditionSucceeded,
Status: status,
}},
},
},
}
}

ctx, _ := ttesting.SetupFakeContext(t)
informer := fakepipelineruninformer.Get(ctx)
// Add N randomly-named PipelineRuns with differently-succeeded statuses.
for _, pipelineRun := range []*v1.PipelineRun{
newPipelineRun(corev1.ConditionTrue, "testns"),
newPipelineRun(corev1.ConditionUnknown, "testns"),
newPipelineRun(corev1.ConditionFalse, "testns"),
} {
if err := informer.Informer().GetIndexer().Add(pipelineRun); err != nil {
t.Fatalf("Adding TaskRun to informer: %v", err)
}
}

ctx = getConfigContextRunningPRLevel("pipeline")
recorder, err := NewRecorder(ctx)
if err != nil {
t.Fatalf("NewRecorder: %v", err)
}

if err := recorder.RunningPipelineRuns(informer.Lister()); err != nil {
t.Errorf("RunningPipelineRuns: %v", err)
}
metricstest.CheckLastValueData(t, "running_pipelineruns", map[string]string{"namespace": "testns", "pipeline": "anonymous"}, 1)
}

func TestRecordRunningPipelineRunsResolutionWaitCounts(t *testing.T) {
multiplier := 3
for _, tc := range []struct {
Expand Down

0 comments on commit d9c0b48

Please sign in to comment.