feat: cumulative workflow job completion counters

docs/partials/metrics.md

@@ -241,6 +241,9 @@ github_runner_repo_busy{owner, id, name, os, status}
 github_runner_repo_online{owner, id, name, os, status}
 : Static metrics of runner is online or not
 
+github_workflow_job_completed_total{owner, repo, workflow_name, name, conclusion}
+: Total number of completed workflow jobs
+
 github_workflow_job_created_timestamp{owner, repo, name, title, branch, sha, identifier, run_id, run_attempt, labels, runner_id, runner_name, runner_group_id, runner_group_name, workflow_name, conclusion}
 : Timestamp when the workflow job have been created
 
@@ -250,6 +253,9 @@ github_workflow_job_duration_ms{owner, repo, name, title, branch, sha, identifie
 github_workflow_job_duration_run_created_minutes{owner, repo, name, title, branch, sha, identifier, run_id, run_attempt, labels, runner_id, runner_name, runner_group_id, runner_group_name, workflow_name, conclusion}
 : Duration since the workflow run creation time in minutes
 
+github_workflow_job_duration_seconds_total{owner, repo, workflow_name, name, conclusion}
+: Total duration of completed workflow jobs in seconds
+
 github_workflow_job_started_timestamp{owner, repo, name, title, branch, sha, identifier, run_id, run_attempt, labels, runner_id, runner_name, runner_group_id, runner_group_name, workflow_name, conclusion}
 : Timestamp when the workflow job have been started
 

go.mod

@@ -15,6 +15,7 @@ require (
 	github.com/lib/pq v1.12.3
 	github.com/oklog/run v1.2.0
 	github.com/prometheus/client_golang v1.23.2
+	github.com/prometheus/client_model v0.6.2
 	github.com/prometheus/exporter-toolkit v0.17.1
 	github.com/ryanuber/go-glob v1.0.0
 	github.com/stretchr/testify v1.11.1
@@ -200,7 +201,6 @@ require (
 	github.com/pelletier/go-toml/v2 v2.3.1 // indirect
 	github.com/pkg/errors v0.9.1 // indirect
 	github.com/pmezard/go-difflib v1.0.0 // indirect
-	github.com/prometheus/client_model v0.6.2 // indirect
 	github.com/prometheus/common v0.69.0 // indirect
 	github.com/prometheus/procfs v0.17.0 // indirect
 	github.com/quasilyte/go-ruleguard v0.4.5 // indirect

pkg/exporter/workflow_job.go

@@ -19,11 +19,13 @@ type WorkflowJobCollector struct {
 	duration *prometheus.HistogramVec
 	config   config.Target
 
-	Status   *prometheus.Desc
-	Duration *prometheus.Desc
-	Creation *prometheus.Desc
-	Created  *prometheus.Desc
-	Started  *prometheus.Desc
+	Status               *prometheus.Desc
+	Duration             *prometheus.Desc
+	Creation             *prometheus.Desc
+	Created              *prometheus.Desc
+	Started              *prometheus.Desc
+	CompletedTotal       *prometheus.Desc
+	DurationSecondsTotal *prometheus.Desc
 }
 
 // NewWorkflowJobCollector returns a new WorkflowCollector.
@@ -33,6 +35,14 @@ func NewWorkflowJobCollector(logger *slog.Logger, client *github.Client, db stor
 	}
 
 	labels := cfg.WorkflowJobs.Labels
+	completionLabels := []string{
+		"owner",
+		"repo",
+		"workflow_name",
+		"name",
+		"conclusion",
+	}
+
 	return &WorkflowJobCollector{
 		client:   client,
 		logger:   logger.With("collector", "workflow_job"),
@@ -71,6 +81,18 @@ func NewWorkflowJobCollector(logger *slog.Logger, client *github.Client, db stor
 			labels,
 			nil,
 		),
+		CompletedTotal: prometheus.NewDesc(
+			"github_workflow_job_completed_total",
+			"Total number of completed workflow jobs",
+			completionLabels,
+			nil,
+		),
+		DurationSecondsTotal: prometheus.NewDesc(
+			"github_workflow_job_duration_seconds_total",
+			"Total duration of completed workflow jobs in seconds",
+			completionLabels,
+			nil,
+		),
 	}
 }
 
@@ -82,6 +104,8 @@ func (c *WorkflowJobCollector) Metrics() []*prometheus.Desc {
 		c.Creation,
 		c.Created,
 		c.Started,
+		c.CompletedTotal,
+		c.DurationSecondsTotal,
 	}
 }
 
@@ -92,6 +116,8 @@ func (c *WorkflowJobCollector) Describe(ch chan<- *prometheus.Desc) {
 	ch <- c.Creation
 	ch <- c.Created
 	ch <- c.Started
+	ch <- c.CompletedTotal
+	ch <- c.DurationSecondsTotal
 }
 
 // Collect is called by the Prometheus registry when collecting metrics.
@@ -174,6 +200,45 @@ func (c *WorkflowJobCollector) Collect(ch chan<- prometheus.Metric) {
 			labels...,
 		)
 	}
+
+	completions, err := c.db.GetWorkflowJobCompletions()
+
+	if err != nil {
+		c.logger.Error("Failed to fetch workflow job completions",
+			"err", err,
+		)
+
+		c.failures.WithLabelValues("workflow_job").Inc()
+		return
+	}
+
+	c.logger.Debug("Fetched workflow job completions",
+		"count", len(completions),
+	)
+
+	for _, completion := range completions {
+		labels := []string{
+			completion.Owner,
+			completion.Repo,
+			completion.WorkflowName,
+			completion.Name,
+			completion.Conclusion,
+		}
+
+		ch <- prometheus.MustNewConstMetric(
+			c.CompletedTotal,
+			prometheus.CounterValue,
+			float64(completion.Count),
+			labels...,
+		)
+
+		ch <- prometheus.MustNewConstMetric(
+			c.DurationSecondsTotal,
+			prometheus.CounterValue,
+			completion.DurationSecondsTotal,
+			labels...,
+		)
+	}
 }
 
 func jobStatusToGauge(conclusion string) float64 {

pkg/exporter/workflow_job_test.go

@@ -1,7 +1,6 @@
 package exporter
 
 import (
-	"fmt"
 	"log/slog"
 	"os"
 	"reflect"
@@ -10,24 +9,13 @@ import (
 
 	"github.com/google/go-github/v89/github"
 	"github.com/prometheus/client_golang/prometheus"
+	dto "github.com/prometheus/client_model/go"
 	"github.com/promhippie/github_exporter/pkg/config"
 	"github.com/promhippie/github_exporter/pkg/store"
 )
 
 type StaticStore struct{}
 
-func (s StaticStore) GetWorkflowJobRuns(owner, repo, workflow string) ([]*store.WorkflowRun, error) {
-	_, _ = fmt.Fprintf(
-		os.Stdout,
-		"GetWorkflowJobRuns for %s/%s %s \n",
-		owner,
-		repo,
-		workflow,
-	)
-
-	return nil, nil
-}
-
 func (s StaticStore) StoreWorkflowRunEvent(*github.WorkflowRunEvent) error {
 	return nil
 }
@@ -52,6 +40,10 @@ func (s StaticStore) PruneWorkflowJobs(time.Duration) error {
 	return nil
 }
 
+func (s StaticStore) GetWorkflowJobCompletions() ([]*store.WorkflowJobCompletionAggregate, error) {
+	return nil, nil
+}
+
 func (s StaticStore) Open() (bool, error) {
 	return true, nil
 }
@@ -118,6 +110,21 @@ func TestWorkflowJobCollector(t *testing.T) {
 			"Created time of the workflow job",
 			nil, nil,
 		),
+		Started: prometheus.NewDesc(
+			"github_workflow_job_started_timestamp",
+			"Timestamp when the workflow job have been started",
+			nil, nil,
+		),
+		CompletedTotal: prometheus.NewDesc(
+			"github_workflow_job_completed_total",
+			"Total number of completed workflow jobs",
+			nil, nil,
+		),
+		DurationSecondsTotal: prometheus.NewDesc(
+			"github_workflow_job_duration_seconds_total",
+			"Total duration of completed workflow jobs in seconds",
+			nil, nil,
+		),
 	}
 
 	if collector.client != mockClient {
@@ -139,3 +146,104 @@ func TestWorkflowJobCollector(t *testing.T) {
 		t.Errorf("Expected config to be %v, got %v", mockConfig, collector.config)
 	}
 }
+
+type completionStore struct {
+	StaticStore
+	completions []*store.WorkflowJobCompletionAggregate
+}
+
+func (s completionStore) GetWorkflowJobCompletions() ([]*store.WorkflowJobCompletionAggregate, error) {
+	return s.completions, nil
+}
+
+func TestWorkflowJobCollectorCounters(t *testing.T) {
+	mockLogger := slog.New(
+		slog.NewTextHandler(os.Stdout, &slog.HandlerOptions{
+			Level: slog.LevelDebug,
+		}),
+	)
+
+	mockFailures := prometheus.NewCounterVec(prometheus.CounterOpts{
+		Name: "test_failures_total",
+		Help: "Total number of test failures",
+	}, []string{"type"})
+
+	mockDuration := prometheus.NewHistogramVec(prometheus.HistogramOpts{
+		Name: "test_duration_seconds",
+		Help: "Duration of test",
+	}, []string{"type"})
+
+	completions := []*store.WorkflowJobCompletionAggregate{
+		{
+			Owner:                "promhippie",
+			Repo:                 "github_exporter",
+			WorkflowName:         "CI",
+			Name:                 "test",
+			Conclusion:           "success",
+			Count:                2,
+			DurationSecondsTotal: 42.5,
+		},
+		{
+			Owner:                "promhippie",
+			Repo:                 "github_exporter",
+			WorkflowName:         "CI",
+			Name:                 "test",
+			Conclusion:           "failure",
+			Count:                1,
+			DurationSecondsTotal: 10.0,
+		},
+	}
+
+	store := completionStore{completions: completions}
+	collector := NewWorkflowJobCollector(
+		mockLogger,
+		nil,
+		store,
+		mockFailures,
+		mockDuration,
+		config.Target{},
+	)
+
+	registry := prometheus.NewRegistry()
+	registry.MustRegister(collector)
+
+	metrics, err := registry.Gather()
+	if err != nil {
+		t.Fatalf("failed to gather metrics: %v", err)
+	}
+
+	expected := map[string]float64{
+		"github_workflow_job_completed_total":        3,
+		"github_workflow_job_duration_seconds_total": 52.5,
+	}
+
+	for name, expectedValue := range expected {
+		value := metricFamilyValue(t, metrics, name)
+		if value != expectedValue {
+			t.Errorf("expected %s to be %v, got %v", name, expectedValue, value)
+		}
+	}
+}
+
+func metricFamilyValue(t *testing.T, metrics []*dto.MetricFamily, name string) float64 {
+	t.Helper()
+
+	for _, mf := range metrics {
+		if mf.GetName() != name {
+			continue
+		}
+
+		var total float64
+
+		for _, m := range mf.GetMetric() {
+			if m.Counter != nil {
+				total += m.Counter.GetValue()
+			}
+		}
+
+		return total
+	}
+
+	t.Errorf("metric family %s not found", name)
+	return 0
+}

pkg/store/chai.go

@@ -73,6 +73,24 @@ var (
 				PRIMARY KEY(owner, repo, identifier)
 			);`,
 		},
+		{
+			Version:     4,
+			Description: "Creating table workflow_job_completions",
+			Script: `CREATE TABLE workflow_job_completions (
+				owner TEXT NOT NULL,
+				repo TEXT NOT NULL,
+				identifier BIGINT NOT NULL,
+				run_attempt INTEGER NOT NULL,
+				workflow_name TEXT,
+				name TEXT,
+				conclusion TEXT,
+				duration_seconds DOUBLE PRECISION,
+				recorded_at INTEGER,
+				PRIMARY KEY(owner, repo, identifier, run_attempt)
+			);
+			CREATE INDEX idx_workflow_job_completions_aggregate
+				ON workflow_job_completions(owner, repo, workflow_name, name, conclusion);`,
+		},
 	}
 )
 
@@ -166,6 +184,11 @@ func (s *chaiStore) PruneWorkflowJobs(timeframe time.Duration) error {
 	return pruneWorkflowJobs(s.handle, timeframe)
 }
 
+// GetWorkflowJobCompletions implements the Store interface.
+func (s *chaiStore) GetWorkflowJobCompletions() ([]*WorkflowJobCompletionAggregate, error) {
+	return getWorkflowJobCompletions(s.handle)
+}
+
 func (s *chaiStore) dsn() string {
 	if len(s.meta) > 0 {
 		return fmt.Sprintf(