Add backup_active_total metric.

This counts the number of backups currently in progress. Signed-off-by: Matthew Arnold <[email protected]> Better metrics function names. Signed-off-by: Matthew Arnold <[email protected]> Update copyright notice as per contribution guidelines. Signed-off-by: Matthew Arnold <[email protected]> Add changelog file. Signed-off-by: Matthew Arnold <[email protected]> Rework backup_active_total after rebase. Signed-off-by: Matthew Arnold <[email protected]>
vmware-tanzu · Oct 15, 2024 · fd6f375 · fd6f375
1 parent 34d4f18
commit fd6f375
Show file tree

Hide file tree

Showing 4 changed files with 27 additions and 1 deletion.
diff --git a/changelogs/unreleased/5703-mrnold b/changelogs/unreleased/5703-mrnold
@@ -0,0 +1 @@
+Add velero_backup_active_total metric
diff --git a/pkg/controller/backup_controller.go b/pkg/controller/backup_controller.go
@@ -269,10 +269,12 @@ func (b *backupReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctr
 	original = request.Backup.DeepCopy()
 
 	b.backupTracker.Add(request.Namespace, request.Name)
+	b.metrics.IncrementActiveBackupTotal()
 	defer func() {
 		switch request.Status.Phase {
 		case velerov1api.BackupPhaseCompleted, velerov1api.BackupPhasePartiallyFailed, velerov1api.BackupPhaseFailed, velerov1api.BackupPhaseFailedValidation:
 			b.backupTracker.Delete(request.Namespace, request.Name)
+			b.metrics.DecrementActiveBackupTotal()
 		}
 	}()
 

diff --git a/pkg/controller/backup_finalizer_controller.go b/pkg/controller/backup_finalizer_controller.go
@@ -121,6 +121,7 @@ func (r *backupFinalizerReconciler) Reconcile(ctx context.Context, req ctrl.Requ
 			velerov1api.BackupPhaseFailed,
 			velerov1api.BackupPhaseFailedValidation:
 			r.backupTracker.Delete(backup.Namespace, backup.Name)
+			r.metrics.DecrementActiveBackupTotal()
 		}
 		// Always attempt to Patch the backup object and status after each reconciliation.
 		//

diff --git a/pkg/metrics/metrics.go b/pkg/metrics/metrics.go
@@ -1,5 +1,5 @@
 /*
-Copyright 2018 the Velero contributors.
+Copyright the Velero contributors.
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -33,6 +33,7 @@ const (
 	//Velero metrics
 	backupTarballSizeBytesGauge   = "backup_tarball_size_bytes"
 	backupTotal                   = "backup_total"
+	backupActiveTotal             = "backup_active_total"
 	backupAttemptTotal            = "backup_attempt_total"
 	backupSuccessTotal            = "backup_success_total"
 	backupPartialFailureTotal     = "backup_partial_failure_total"
@@ -113,6 +114,13 @@ func NewServerMetrics() *ServerMetrics {
 					Help:      "Current number of existent backups",
 				},
 			),
+			backupActiveTotal: prometheus.NewGauge(
+				prometheus.GaugeOpts{
+					Namespace: metricNamespace,
+					Name:      backupActiveTotal,
+					Help:      "Number of backups currently in progress",
+				},
+			),
 			backupAttemptTotal: prometheus.NewCounterVec(
 				prometheus.CounterOpts{
 					Namespace: metricNamespace,
@@ -706,6 +714,20 @@ func (m *ServerMetrics) SetBackupTotal(numberOfBackups int64) {
 	}
 }
 
+// IncrementActiveBackupTotal increments the number of in-progress backups
+func (m *ServerMetrics) IncrementActiveBackupTotal() {
+	if g, ok := m.metrics[backupActiveTotal].(prometheus.Gauge); ok {
+		g.Inc()
+	}
+}
+
+// DecrementActiveBackupTotal decrements the number of in-progress backups
+func (m *ServerMetrics) DecrementActiveBackupTotal() {
+	if g, ok := m.metrics[backupActiveTotal].(prometheus.Gauge); ok {
+		g.Dec()
+	}
+}
+
 // RegisterBackupAttempt records an backup attempt.
 func (m *ServerMetrics) RegisterBackupAttempt(backupSchedule string) {
 	if c, ok := m.metrics[backupAttemptTotal].(*prometheus.CounterVec); ok {