From 243ac62e3fea67659b74bf3bf4df520773895da3 Mon Sep 17 00:00:00 2001
From: kathpeony <68385123+kathpeony@users.noreply.github.com>
Date: Thu, 16 Jul 2020 19:13:17 +0200
Subject: [PATCH] Add backupValidationFailureTotal to metrics (#2714)

* Add backValidationFailureTotal to metrics

Signed-off-by: Kathrin Mao <kathrin.mao@sap.com>
---
 changelogs/unreleased/2714-kathpeony |  1 +
 pkg/controller/backup_controller.go  |  2 ++
 pkg/metrics/metrics.go               | 19 +++++++++++++++++++
 3 files changed, 22 insertions(+)
 create mode 100644 changelogs/unreleased/2714-kathpeony

diff --git a/changelogs/unreleased/2714-kathpeony b/changelogs/unreleased/2714-kathpeony
new file mode 100644
index 000000000..89e261d9e
--- /dev/null
+++ b/changelogs/unreleased/2714-kathpeony
@@ -0,0 +1 @@
+Add backupValidationFailureTotal to metrics
diff --git a/pkg/controller/backup_controller.go b/pkg/controller/backup_controller.go
index 366e81951..e958229ec 100644
--- a/pkg/controller/backup_controller.go
+++ b/pkg/controller/backup_controller.go
@@ -287,6 +287,8 @@ func (c *backupController) processBackup(key string) error {
 		c.metrics.RegisterBackupPartialFailure(backupScheduleName)
 	case velerov1api.BackupPhaseFailed:
 		c.metrics.RegisterBackupFailed(backupScheduleName)
+	case velerov1api.BackupPhaseFailedValidation:
+		c.metrics.RegisterBackupValidationFailure(backupScheduleName)
 	}
 
 	log.Debug("Updating backup's final status")
diff --git a/pkg/metrics/metrics.go b/pkg/metrics/metrics.go
index d3ed0cbee..2e18c4345 100644
--- a/pkg/metrics/metrics.go
+++ b/pkg/metrics/metrics.go
@@ -35,6 +35,7 @@ const (
 	backupSuccessTotal            = "backup_success_total"
 	backupPartialFailureTotal     = "backup_partial_failure_total"
 	backupFailureTotal            = "backup_failure_total"
+	backupValidationFailureTotal  = "backup_validation_failure_total"
 	backupDurationSeconds         = "backup_duration_seconds"
 	backupDeletionAttemptTotal    = "backup_deletion_attempt_total"
 	backupDeletionSuccessTotal    = "backup_deletion_success_total"
@@ -115,6 +116,14 @@ func NewServerMetrics() *ServerMetrics {
 				},
 				[]string{scheduleLabel},
 			),
+			backupValidationFailureTotal: prometheus.NewCounterVec(
+				prometheus.CounterOpts{
+					Namespace: metricNamespace,
+					Name:      backupValidationFailureTotal,
+					Help:      "Total number of validation failed backups",
+				},
+				[]string{scheduleLabel},
+			),
 			backupDeletionAttemptTotal: prometheus.NewCounterVec(
 				prometheus.CounterOpts{
 					Namespace: metricNamespace,
@@ -254,6 +263,9 @@ func (m *ServerMetrics) InitSchedule(scheduleName string) {
 	if c, ok := m.metrics[backupFailureTotal].(*prometheus.CounterVec); ok {
 		c.WithLabelValues(scheduleName).Add(0)
 	}
+	if c, ok := m.metrics[backupValidationFailureTotal].(*prometheus.CounterVec); ok {
+		c.WithLabelValues(scheduleName).Add(0)
+	}
 	if c, ok := m.metrics[backupDeletionAttemptTotal].(*prometheus.CounterVec); ok {
 		c.WithLabelValues(scheduleName).Add(0)
 	}
@@ -339,6 +351,13 @@ func (m *ServerMetrics) RegisterBackupFailed(backupSchedule string) {
 	}
 }
 
+// RegisterBackupValidationFailure records a validation failed backup.
+func (m *ServerMetrics) RegisterBackupValidationFailure(backupSchedule string) {
+	if c, ok := m.metrics[backupValidationFailureTotal].(*prometheus.CounterVec); ok {
+		c.WithLabelValues(backupSchedule).Inc()
+	}
+}
+
 // RegisterBackupDuration records the number of seconds a backup took.
 func (m *ServerMetrics) RegisterBackupDuration(backupSchedule string, seconds float64) {
 	if c, ok := m.metrics[backupDurationSeconds].(*prometheus.HistogramVec); ok {