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.
We have added the PipelineRun metric at pipeline and namespace level.

Fix Lint Error

Fix Lint Error
  • Loading branch information
pramodbindal committed Oct 10, 2024
1 parent 0649270 commit d1757a6
Show file tree
Hide file tree
Showing 8 changed files with 200 additions and 18 deletions.
1 change: 1 addition & 0 deletions config/config-observability.yaml
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
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
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
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
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"
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"
63 changes: 58 additions & 5 deletions pkg/pipelinerunmetrics/metrics.go
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"
runningPRLevelCluster = ""
)

var (
pipelinerunTag = tag.MustNewKey("pipelinerun")
pipelineTag = tag.MustNewKey("pipeline")
Expand Down Expand Up @@ -134,6 +141,7 @@ func NewRecorder(ctx context.Context) (*Recorder, error) {
}

cfg := config.FromContextOrDefaults(ctx)
r.cfg = cfg.Metrics
errRegistering = viewRegister(cfg.Metrics)
if errRegistering != nil {
r.initialized = false
Expand All @@ -149,7 +157,6 @@ 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 +171,18 @@ 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 +232,7 @@ 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 +346,7 @@ 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 +399,10 @@ 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 +415,37 @@ 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 runningPRLevelCluster:
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 +456,13 @@ func (r *Recorder) RunningPipelineRuns(lister listers.PipelineRunLister) error {
prsWaitResolvingPipelineRef++
}
}
} else {
// In case there are no running PipelineRuns for the pipelineRunKey, set the metric value to 0 to ensure
// the metric is set for the key.
if _, exists := countMap[pipelineRunKey]; !exists {
countMap[pipelineRunKey] = 0
metrics.Record(ctx_, runningPRs.M(0))
}
}
}

Expand All @@ -421,7 +475,6 @@ 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
Loading

0 comments on commit d1757a6

Please sign in to comment.