Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add a branch label to workflow metrics #164

Merged
merged 2 commits into from
Feb 18, 2024
Merged

Add a branch label to workflow metrics #164

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6997424
Add a branch label to workflow metrics
rguliyev Jan 25, 2024
6f218c3
Merge branch 'main' into branch_labels
rguliyev Feb 18, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
40 changes: 20 additions & 20 deletions internal/server/metrics.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,36 +8,36 @@ var (
Help: "Time that a workflow job took to reach a given state.",
Buckets: prometheus.ExponentialBuckets(1, 1.4, 30),
},
[]string{"org", "repo", "state", "runner_group", "workflow_name", "job_name"},
[]string{"org", "repo", "branch", "state", "runner_group", "workflow_name", "job_name"},
)

workflowJobDurationCounter = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "workflow_job_duration_seconds_total",
Help: "The total duration of jobs.",
},
[]string{"org", "repo", "status", "conclusion", "runner_group", "workflow_name", "job_name"},
[]string{"org", "repo", "branch", "status", "conclusion", "runner_group", "workflow_name", "job_name"},
)

workflowJobStatusCounter = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "workflow_job_status_count",
Help: "Count of workflow job events.",
},
[]string{"org", "repo", "status", "conclusion", "runner_group", "workflow_name", "job_name"},
[]string{"org", "repo", "branch", "status", "conclusion", "runner_group", "workflow_name", "job_name"},
)

workflowRunHistogramVec = prometheus.NewHistogramVec(prometheus.HistogramOpts{
Name: "workflow_execution_time_seconds",
Help: "Time that a workflow took to run.",
Buckets: prometheus.ExponentialBuckets(1, 1.4, 30),
},
[]string{"org", "repo", "workflow_name", "conclusion"},
[]string{"org", "repo", "branch", "workflow_name", "conclusion"},
)

workflowRunStatusCounter = prometheus.NewCounterVec(prometheus.CounterOpts{
Name: "workflow_status_count",
Help: "Count of the occurrences of different workflow states.",
},
[]string{"org", "repo", "status", "conclusion", "workflow_name"},
[]string{"org", "repo", "branch", "status", "conclusion", "workflow_name"},
)

totalMinutesUsedActions = prometheus.NewGaugeVec(prometheus.GaugeOpts{
Expand Down Expand Up @@ -83,36 +83,36 @@ func init() {
}

type WorkflowObserver interface {
ObserveWorkflowJobDuration(org, repo, state, runnerGroup, workflowName, jobName string, seconds float64)
CountWorkflowJobStatus(org, repo, status, conclusion, runnerGroup, workflowName, jobName string)
CountWorkflowJobDuration(org, repo, status, conclusion, runnerGroup, workflowName, jobName string, seconds float64)
ObserveWorkflowJobDuration(org, repo, branch, state, runnerGroup, workflowName, jobName string, seconds float64)
CountWorkflowJobStatus(org, repo, branch, status, conclusion, runnerGroup, workflowName, jobName string)
CountWorkflowJobDuration(org, repo, branch, status, conclusion, runnerGroup, workflowName, jobName string, seconds float64)

ObserveWorkflowRunDuration(org, repo, workflow, conclusion string, seconds float64)
CountWorkflowRunStatus(org, repo, status, conclusion, workflow string)
ObserveWorkflowRunDuration(org, repo, branch, workflow, conclusion string, seconds float64)
CountWorkflowRunStatus(org, repo, branch, status, conclusion, workflow string)
}

var _ WorkflowObserver = (*PrometheusObserver)(nil)

type PrometheusObserver struct{}

func (o *PrometheusObserver) ObserveWorkflowJobDuration(org, repo, state, runnerGroup, workflowName, jobName string, seconds float64) {
workflowJobHistogramVec.WithLabelValues(org, repo, state, runnerGroup, workflowName, jobName).
func (o *PrometheusObserver) ObserveWorkflowJobDuration(org, repo, branch, state, runnerGroup, workflowName, jobName string, seconds float64) {
workflowJobHistogramVec.WithLabelValues(org, repo, branch, state, runnerGroup, workflowName, jobName).
Observe(seconds)
}

func (o *PrometheusObserver) CountWorkflowJobStatus(org, repo, status, conclusion, runnerGroup, workflowName, jobName string) {
workflowJobStatusCounter.WithLabelValues(org, repo, status, conclusion, runnerGroup, workflowName, jobName).Inc()
func (o *PrometheusObserver) CountWorkflowJobStatus(org, repo, branch, status, conclusion, runnerGroup, workflowName, jobName string) {
workflowJobStatusCounter.WithLabelValues(org, repo, branch, status, conclusion, runnerGroup, workflowName, jobName).Inc()
}

func (o *PrometheusObserver) CountWorkflowJobDuration(org, repo, status, conclusion, runnerGroup, workflowName, jobName string, seconds float64) {
workflowJobDurationCounter.WithLabelValues(org, repo, status, conclusion, runnerGroup, workflowName, jobName).Add(seconds)
func (o *PrometheusObserver) CountWorkflowJobDuration(org, repo, branch, status, conclusion, runnerGroup, workflowName, jobName string, seconds float64) {
workflowJobDurationCounter.WithLabelValues(org, repo, branch, status, conclusion, runnerGroup, workflowName, jobName).Add(seconds)
}

func (o *PrometheusObserver) ObserveWorkflowRunDuration(org, repo, workflowName, conclusion string, seconds float64) {
workflowRunHistogramVec.WithLabelValues(org, repo, workflowName, conclusion).
func (o *PrometheusObserver) ObserveWorkflowRunDuration(org, repo, branch, workflowName, conclusion string, seconds float64) {
workflowRunHistogramVec.WithLabelValues(org, repo, branch, workflowName, conclusion).
Observe(seconds)
}

func (o *PrometheusObserver) CountWorkflowRunStatus(org, repo, status, conclusion, workflowName string) {
workflowRunStatusCounter.WithLabelValues(org, repo, status, conclusion, workflowName).Inc()
func (o *PrometheusObserver) CountWorkflowRunStatus(org, repo, branch, status, conclusion, workflowName string) {
workflowRunStatusCounter.WithLabelValues(org, repo, branch, status, conclusion, workflowName).Inc()
}
6 changes: 4 additions & 2 deletions internal/server/server_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -87,6 +87,7 @@ func Test_Server_MetricsRouteAfterWorkflowJob(t *testing.T) {
}()

repo := "some-repo"
branch := "some-branch"
org := "someone"
expectedDuration := 10.0
jobStartedAt := time.Unix(1650308740, 0)
Expand All @@ -104,6 +105,7 @@ func Test_Server_MetricsRouteAfterWorkflowJob(t *testing.T) {
},
},
WorkflowJob: &github.WorkflowJob{
HeadBranch: &branch,
Status: github.String("completed"),
Conclusion: github.String("success"),
StartedAt: &github.Timestamp{Time: jobStartedAt},
Expand All @@ -129,6 +131,6 @@ func Test_Server_MetricsRouteAfterWorkflowJob(t *testing.T) {

payload, err := io.ReadAll(metricsRes.Body)
require.NoError(t, err)
assert.Contains(t, string(payload), `workflow_job_duration_seconds_bucket{job_name="Test",org="someone",repo="some-repo",runner_group="runner-group",state="in_progress",workflow_name="Build and test",le="10.541350399999995"} 1`)
assert.Contains(t, string(payload), `workflow_job_duration_seconds_total{conclusion="success",job_name="Test",org="someone",repo="some-repo",runner_group="runner-group",status="completed",workflow_name="Build and test"} 10`)
assert.Contains(t, string(payload), `workflow_job_duration_seconds_bucket{branch="some-branch",job_name="Test",org="someone",repo="some-repo",runner_group="runner-group",state="in_progress",workflow_name="Build and test",le="10.541350399999995"} 1`)
assert.Contains(t, string(payload), `workflow_job_duration_seconds_total{branch="some-branch",conclusion="success",job_name="Test",org="someone",repo="some-repo",runner_group="runner-group",status="completed",workflow_name="Build and test"} 10`)
}
17 changes: 10 additions & 7 deletions internal/server/workflow_metrics_exporter.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,14 +77,15 @@ func (c *WorkflowMetricsExporter) HandleGHWebHook(w http.ResponseWriter, r *http
_ = level.Info(c.Logger).Log("msg", "got workflow_job event",
"org", event.GetRepo().GetOwner().GetLogin(),
"repo", event.GetRepo().GetName(),
"branch", event.GetWorkflowJob().GetHeadBranch(),
"runId", event.GetWorkflowJob().GetRunID(),
"action", event.GetAction(),
"workflow_name", event.GetWorkflowJob().GetWorkflowName(),
"job_name", event.GetWorkflowJob().GetName())
go c.CollectWorkflowJobEvent(event)
case "workflow_run":
event := model.WorkflowRunEventFromJSON(io.NopCloser(bytes.NewBuffer(buf)))
_ = level.Info(c.Logger).Log("msg", "got workflow_run event", "org", event.GetRepo().GetOwner().GetLogin(), "repo", event.GetRepo().GetName(), "workflow_name", event.GetWorkflow().GetName(), "runNumber", event.GetWorkflowRun().GetRunNumber(), "action", event.GetAction())
_ = level.Info(c.Logger).Log("msg", "got workflow_run event", "org", event.GetRepo().GetOwner().GetLogin(), "repo", event.GetRepo().GetName(), "branch", event.GetWorkflowRun().GetHeadBranch(), "workflow_name", event.GetWorkflow().GetName(), "runNumber", event.GetWorkflowRun().GetRunNumber(), "action", event.GetAction())
go c.CollectWorkflowRunEvent(event)
default:
_ = level.Info(c.Logger).Log("msg", "not implemented", "eventType", eventType)
Expand All @@ -99,6 +100,7 @@ func (c *WorkflowMetricsExporter) HandleGHWebHook(w http.ResponseWriter, r *http
func (c *WorkflowMetricsExporter) CollectWorkflowJobEvent(event *github.WorkflowJobEvent) {
repo := event.GetRepo().GetName()
org := event.GetRepo().GetOwner().GetLogin()
branch := event.WorkflowJob.GetHeadBranch()
action := event.GetAction()

workflowJob := event.GetWorkflowJob()
Expand All @@ -120,34 +122,35 @@ func (c *WorkflowMetricsExporter) CollectWorkflowJobEvent(event *github.Workflow

firstStep := workflowJob.Steps[0]
queuedSeconds := firstStep.StartedAt.Time.Sub(workflowJob.GetStartedAt().Time).Seconds()
c.PrometheusObserver.ObserveWorkflowJobDuration(org, repo, "queued", runnerGroup, workflowName, jobName, math.Max(0, queuedSeconds))
c.PrometheusObserver.ObserveWorkflowJobDuration(org, repo, branch, "queued", runnerGroup, workflowName, jobName, math.Max(0, queuedSeconds))
case "completed":
if workflowJob.StartedAt == nil || workflowJob.CompletedAt == nil {
_ = level.Debug(c.Logger).Log("msg", "unable to calculate job duration of completed event steps are missing timestamps")
break
}

jobSeconds := math.Max(0, workflowJob.GetCompletedAt().Time.Sub(workflowJob.GetStartedAt().Time).Seconds())
c.PrometheusObserver.ObserveWorkflowJobDuration(org, repo, "in_progress", runnerGroup, workflowName, jobName, jobSeconds)
c.PrometheusObserver.CountWorkflowJobDuration(org, repo, status, conclusion, runnerGroup, workflowName, jobName, jobSeconds)
c.PrometheusObserver.ObserveWorkflowJobDuration(org, repo, branch, "in_progress", runnerGroup, workflowName, jobName, jobSeconds)
c.PrometheusObserver.CountWorkflowJobDuration(org, repo, branch, status, conclusion, runnerGroup, workflowName, jobName, jobSeconds)
}

c.PrometheusObserver.CountWorkflowJobStatus(org, repo, status, conclusion, runnerGroup, workflowName, jobName)
c.PrometheusObserver.CountWorkflowJobStatus(org, repo, branch, status, conclusion, runnerGroup, workflowName, jobName)
}

func (c *WorkflowMetricsExporter) CollectWorkflowRunEvent(event *github.WorkflowRunEvent) {
repo := event.GetRepo().GetName()
org := event.GetRepo().GetOwner().GetLogin()
branch := event.GetWorkflowRun().GetHeadBranch()
workflowName := event.GetWorkflow().GetName()
conclusion := event.GetWorkflowRun().GetConclusion()

if event.GetAction() == "completed" {
seconds := event.GetWorkflowRun().UpdatedAt.Time.Sub(event.GetWorkflowRun().RunStartedAt.Time).Seconds()
c.PrometheusObserver.ObserveWorkflowRunDuration(org, repo, workflowName, conclusion, seconds)
c.PrometheusObserver.ObserveWorkflowRunDuration(org, repo, branch, workflowName, conclusion, seconds)
}

status := event.GetWorkflowRun().GetStatus()
c.PrometheusObserver.CountWorkflowRunStatus(org, repo, status, conclusion, workflowName)
c.PrometheusObserver.CountWorkflowRunStatus(org, repo, branch, status, conclusion, workflowName)
}

// validateSignature validate the incoming github event.
Expand Down
Loading
Loading