Merge pull request #246 from whywaita/feat/metrisc-exponential-backoff

Add metrics about exponential backoff

Author: whywaita

pkg/runner/metrics.go

@@ -0,0 +1,25 @@
+package runner
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
+)
+
+var (
+	// DeleteRunnerBackoffDuration is histogram of exponential backoff duration for deleting runner
+	DeleteRunnerBackoffDuration = promauto.NewHistogramVec(prometheus.HistogramOpts{
+		Namespace: "myshoes",
+		Subsystem: "runner",
+		Name:      "delete_runner_backoff_duration_seconds",
+		Help:      "Histogram of exponential backoff duration in seconds for deleting runner",
+		Buckets:   prometheus.ExponentialBuckets(1, 2, 10), // 1s, 2s, 4s, 8s, 16s, 32s, 64s, 128s, 256s, 512s
+	}, []string{"runner_uuid"})
+
+	// DeleteRunnerRetryTotal is counter of total retries for deleting runner
+	DeleteRunnerRetryTotal = promauto.NewCounterVec(prometheus.CounterOpts{
+		Namespace: "myshoes",
+		Subsystem: "runner",
+		Name:      "delete_runner_retry_total",
+		Help:      "Total number of retries for deleting runner",
+	}, []string{"runner_uuid"})
+)

pkg/runner/runner_delete.go

@@ -118,7 +118,12 @@ func (m *Manager) removeRunners(ctx context.Context, t datastore.Target) error {
 				sem.Release(1)
 				ConcurrencyDeleting.Add(-1)
 			}()
-			time.Sleep(util.CalcRetryTime(count))
+			sleep := util.CalcRetryTime(count)
+			if count > 0 {
+				DeleteRunnerRetryTotal.WithLabelValues(runner.UUID.String()).Inc()
+				DeleteRunnerBackoffDuration.WithLabelValues(runner.UUID.String()).Observe(sleep.Seconds())
+			}
+			time.Sleep(sleep)
 
 			if err := m.removeRunner(cctx, t, runner, ghRunners); err != nil {
 				DeleteRetryCount.Store(runner.UUID, count+1)

pkg/starter/metrics.go

@@ -0,0 +1,25 @@
+package starter
+
+import (
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promauto"
+)
+
+var (
+	// AddInstanceBackoffDuration is histogram of exponential backoff duration for adding instance
+	AddInstanceBackoffDuration = promauto.NewHistogramVec(prometheus.HistogramOpts{
+		Namespace: "myshoes",
+		Subsystem: "starter",
+		Name:      "add_instance_backoff_duration_seconds",
+		Help:      "Histogram of exponential backoff duration in seconds for adding instance",
+		Buckets:   prometheus.ExponentialBuckets(1, 2, 10), // 1s, 2s, 4s, 8s, 16s, 32s, 64s, 128s, 256s, 512s
+	}, []string{"job_uuid"})
+
+	// AddInstanceRetryTotal is counter of total retries for adding instance
+	AddInstanceRetryTotal = promauto.NewCounterVec(prometheus.CounterOpts{
+		Namespace: "myshoes",
+		Subsystem: "starter",
+		Name:      "add_instance_retry_total",
+		Help:      "Total number of retries for adding instance",
+	}, []string{"job_uuid"})
+)

pkg/starter/starter.go

@@ -163,6 +163,10 @@ func (s *Starter) run(ctx context.Context, ch chan datastore.Job) error {
 			inProgress.Store(job.UUID, struct{}{})
 
 			sleep := util.CalcRetryTime(count)
+			if count > 0 {
+				AddInstanceRetryTotal.WithLabelValues(job.UUID.String()).Inc()
+				AddInstanceBackoffDuration.WithLabelValues(job.UUID.String()).Observe(sleep.Seconds())
+			}
 			go func(job datastore.Job, sleep time.Duration) {
 				defer func() {
 					sem.Release(1)