Add counter metrics to record error count

Add two metrics to record the `mount device failed` and `cannt mount volume`
error. So, we could know the frequency the error happened.

Signed-off-by: JenTing Hsiao <[email protected]>

Author: jenting

components/ws-manager/pkg/manager/metrics.go

@@ -52,6 +52,8 @@ type metrics struct {
 	totalRestoreCounterVec                    *prometheus.CounterVec
 	totalRestoreFailureCounterVec             *prometheus.CounterVec
 	totalUnintentionalWorkspaceStopCounterVec *prometheus.CounterVec
+	totalMountDeviceFailedVec                 *prometheus.CounterVec
+	totalCannotMountVolumeVec                 *prometheus.CounterVec
 
 	// Gauge
 	totalOpenPortGauge prometheus.GaugeFunc
@@ -142,6 +144,18 @@ func newMetrics(m *Manager) *metrics {
 			Name:      "workspace_unintentional_stop_total",
 			Help:      "total number of workspaces when container stopped without being deleted prior",
 		}, []string{"type", "class"}),
+		totalMountDeviceFailedVec: prometheus.NewCounterVec(prometheus.CounterOpts{
+			Namespace: metricsNamespace,
+			Subsystem: metricsWorkspaceSubsystem,
+			Name:      "workspace_mount_device_failed",
+			Help:      "total number of workspace mount device failed",
+		}, []string{"type", "class"}),
+		totalCannotMountVolumeVec: prometheus.NewCounterVec(prometheus.CounterOpts{
+			Namespace: metricsNamespace,
+			Subsystem: metricsWorkspaceSubsystem,
+			Name:      "workspace_cannot_mount_volume",
+			Help:      "total number of workspace cannot mount volume",
+		}, []string{"type", "class"}),
 		totalOpenPortGauge: prometheus.NewGaugeFunc(prometheus.GaugeOpts{
 			Namespace: metricsNamespace,
 			Subsystem: metricsWorkspaceSubsystem,
@@ -205,6 +219,8 @@ func (m *metrics) Register(reg prometheus.Registerer) error {
 		m.totalRestoreCounterVec,
 		m.totalRestoreFailureCounterVec,
 		m.totalUnintentionalWorkspaceStopCounterVec,
+		m.totalMountDeviceFailedVec,
+		m.totalCannotMountVolumeVec,
 		m.totalOpenPortGauge,
 	}
 	for _, c := range collectors {

components/ws-manager/pkg/manager/status.go

@@ -528,6 +528,8 @@ func (m *Manager) extractStatusFromPod(result *api.WorkspaceStatus, wso workspac
 // one should extract the phase themselves. If the pod has not failed, this function returns "", nil.
 func extractFailure(wso workspaceObjects, metrics *metrics) (string, *api.WorkspacePhase) {
 	pod := wso.Pod
+	wsType := strings.ToUpper(pod.Labels[wsk8s.TypeLabel])
+	wsClass := pod.Labels[workspaceClassLabel]
 
 	// if the workspace was explicitely marked as failed that also constitutes a failure reason
 	reason, explicitFailure := pod.Annotations[workspaceExplicitFailAnnotation]
@@ -590,8 +592,6 @@ func extractFailure(wso workspaceObjects, metrics *metrics) (string, *api.Worksp
 			} else if terminationState.Reason == "Completed" {
 				// container terminated successfully - this is not a failure
 				if !isPodBeingDeleted(pod) {
-					wsType := strings.ToUpper(pod.Labels[wsk8s.TypeLabel])
-					wsClass := pod.Labels[workspaceClassLabel]
 					if metrics != nil && !wso.IsWorkspaceHeadless() {
 						metrics.totalUnintentionalWorkspaceStopCounterVec.WithLabelValues(wsType, wsClass).Inc()
 					}
@@ -618,9 +618,17 @@ func extractFailure(wso workspaceObjects, metrics *metrics) (string, *api.Worksp
 		if strings.Contains(evt.Message, "MountVolume.MountDevice failed for volume") {
 			// ref: https://github.com/gitpod-io/gitpod/issues/13353
 			// ref: https://github.com/kubernetes-sigs/gcp-compute-persistent-disk-csi-driver/issues/608
+			log.WithField("pod", pod.Name).Warnf("%s", evt.Message)
+			if metrics != nil {
+				metrics.totalMountDeviceFailedVec.WithLabelValues(wsType, wsClass).Inc()
+			}
 			return "", nil
 		} else if strings.Contains(evt.Message, workspaceVolumeName) {
 			// ref: https://github.com/gitpod-io/gitpod/issues/14032
+			log.WithField("pod", pod.Name).Warnf("%s", evt.Message)
+			if metrics != nil {
+				metrics.totalCannotMountVolumeVec.WithLabelValues(wsType, wsClass).Inc()
+			}
 			return "", nil
 		} else {
 			// if this happens we did not do a good job because that means we've introduced another volume to the pod

components/ws-manager/pkg/manager/testdata/actOnPodEvent_failedWorkspaceMount_PENDING00.golden

@@ -1,24 +1,3 @@
 {
-    "actions": [
-        {
-            "Func": "markWorkspace",
-            "Params": {
-                "annotations": [
-                    {
-                        "Name": "gitpod/failedBeforeStopping",
-                        "Value": "true",
-                        "Delete": false
-                    }
-                ],
-                "workspaceID": "b3242d9b-6920-41b5-8e72-c3d5637ca148"
-            }
-        },
-        {
-            "Func": "stopWorkspace",
-            "Params": {
-                "gracePeriod": 30000000000,
-                "workspaceID": "b3242d9b-6920-41b5-8e72-c3d5637ca148"
-            }
-        }
-    ]
-}
+    "actions": null
+}

components/ws-manager/pkg/manager/testdata/status_failedWorkspaceMount_PENDING00.golden

@@ -16,7 +16,6 @@
         },
         "phase": 1,
         "conditions": {
-            "failed": "cannot mount workspace",
             "volume_snapshot": {}
         },
         "message": "pod is pending",