ws-manager: remove annotation gitpod/failedBeforeStopping

Use annotation `workspaceExplicitFailAnnotation` instead.

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

Author: jenting

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

@@ -50,11 +50,6 @@ const (
 	//   https://kubernetes.io/docs/concepts/overview/working-with-objects/annotations/#attaching-metadata-to-objects
 	workspaceInitializerAnnotation = "gitpod/contentInitializer"
 
-	// workspaceFailedBeforeStoppingAnnotation marks a workspace as failed even before we tried
-	// to stop it. We do not extract the failure state from this annotation, but just stabilize
-	// the state computation.
-	workspaceFailedBeforeStoppingAnnotation = "gitpod/failedBeforeStopping"
-
 	// customTimeoutAnnotation configures the activity timeout of a workspace, i.e. the timeout a user experiences when not using an otherwise active workspace for some time.
 	// This is handy if you want to prevent a workspace from timing out during lunch break.
 	customTimeoutAnnotation = "gitpod/customTimeout"
@@ -113,13 +108,6 @@ func (m *Manager) markWorkspace(ctx context.Context, workspaceID string, annotat
 
 		for _, a := range annotations {
 			a.Apply(pod.Annotations)
-
-			// Optimization: if we're failing the workspace explicitly, we might as well add the workspaceFailedBeforeStoppingAnnotation
-			// as well. If we didin't do this here, the monitor would do that for us down the road, but this way need one fewer modification
-			// of the pod.
-			if a.Name == workspaceExplicitFailAnnotation {
-				pod.Annotations[workspaceFailedBeforeStoppingAnnotation] = "true"
-			}
 		}
 
 		return m.Clientset.Update(ctx, pod)

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

@@ -277,22 +277,22 @@ func actOnPodEvent(ctx context.Context, m actingManager, manager *Manager, statu
 		// the workspace has failed to run/start - shut it down
 		// we should mark the workspace as failedBeforeStopping - this way the failure status will persist
 		// while we stop the workspace
-		_, hasFailureAnnotation := pod.Annotations[workspaceFailedBeforeStoppingAnnotation]
+		_, hasFailureAnnotation := pod.Annotations[workspaceExplicitFailAnnotation]
 		if status.Conditions.Failed != "" && !hasFailureAnnotation {
 			span.LogKV("event", "failed and no failed before stopping annotation")
 			// If this marking operation failes that's ok - we'll still continue to shut down the workspace.
 			// The failure message won't persist while stopping the workspace though.
-			err := m.markWorkspace(ctx, workspaceID, addMark(workspaceFailedBeforeStoppingAnnotation, util.BooleanTrueString))
+			err := m.markWorkspace(ctx, workspaceID, addMark(workspaceExplicitFailAnnotation, status.Conditions.Failed))
 			if err != nil {
-				log.WithError(err).Debug("cannot mark workspace as workspaceFailedBeforeStoppingAnnotation")
+				log.WithError(err).Debug("cannot mark workspace as workspaceExplicitFailAnnotation")
 			}
 		}
 
 		// At the moment we call stopWorkspace on the same workspace at least twice:
 		// First when the workspace originally failed, and
-		// second when adding the workspaceFailedBeforeStoppingAnnotation which in turn triggers a new pod event.
+		// second when adding the workspaceExplicitFailAnnotation which in turn triggers a new pod event.
 		//
-		// The alternative is to stop the pod only when the workspaceFailedBeforeStoppingAnnotation is present.
+		// The alternative is to stop the pod only when the workspaceExplicitFailAnnotation is present.
 		// However, that's much more brittle than stopping the workspace twice (something that Kubernetes can handle).
 		// It is important that we do not fail here if the pod is already gone, i.e. when we lost the race.
 		err = m.stopWorkspace(ctx, workspaceID, stopWorkspaceNormallyGracePeriod)
@@ -441,7 +441,7 @@ func actOnPodEvent(ctx context.Context, m actingManager, manager *Manager, statu
 
 		span.LogKV("disposalStatusAnnotation", ds)
 		if terminated && !ds.Status.IsDisposed() {
-			if wso.Pod.Annotations[workspaceFailedBeforeStoppingAnnotation] == util.BooleanTrueString && wso.Pod.Annotations[workspaceNeverReadyAnnotation] == util.BooleanTrueString {
+			if wso.Pod.Annotations[workspaceNeverReadyAnnotation] == util.BooleanTrueString {
 				span.LogKV("event", "failed before stopping and never ready")
 				// The workspace is never ready, so there is no need for a finalizer.
 				if _, ok := pod.Annotations[workspaceExplicitFailAnnotation]; !ok {

components/ws-manager/pkg/manager/testdata/actOnPodEvent_brokenScheduler_UNKNOWN00.golden

@@ -5,8 +5,8 @@
             "Params": {
                 "annotations": [
                     {
-                        "Name": "gitpod/failedBeforeStopping",
-                        "Value": "true",
+                        "Name": "gitpod/explicitFail",
+                        "Value": "MatchNodeSelector: Pod Predicate MatchNodeSelector failed",
                         "Delete": false
                     }
                 ],

components/ws-manager/pkg/manager/testdata/actOnPodEvent_cannotPull_004_CREATING00.golden

@@ -5,8 +5,8 @@
             "Params": {
                 "annotations": [
                     {
-                        "Name": "gitpod/failedBeforeStopping",
-                        "Value": "true",
+                        "Name": "gitpod/explicitFail",
+                        "Value": "cannot pull image: rpc error: code = FailedPrecondition desc = failed to pull and unpack image \"reg.gitpod.io:31001/remote/5031df46-db5e-43ae-91bd-1448305c001d:latest\": failed commit on ref \"layer-sha256:6633ce2524dfae110cac2159a4f8490d198612d12abc1420486c52fbcf30b8b1\": \"layer-sha256:6633ce2524dfae110cac2159a4f8490d198612d12abc1420486c52fbcf30b8b1\" failed size validation: 33554502 != 64598931: failed precondition",
                         "Delete": false
                     }
                 ],

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

@@ -5,8 +5,8 @@
             "Params": {
                 "annotations": [
                     {
-                        "Name": "gitpod/failedBeforeStopping",
-                        "Value": "true",
+                        "Name": "gitpod/explicitFail",
+                        "Value": "cannot pull image: rpc error: code = Unknown desc = Error response from daemon: Get https://registry.staging-cw-minio-core.svc.cluster.local/v1/_ping: dial tcp: lookup registry.staging-cw-minio-core.svc.cluster.local: no such host",
                         "Delete": false
                     }
                 ],

components/ws-manager/pkg/manager/testdata/actOnPodEvent_failedLogs_RUNNING00.golden

@@ -5,8 +5,8 @@
             "Params": {
                 "annotations": [
                     {
-                        "Name": "gitpod/failedBeforeStopping",
-                        "Value": "true",
+                        "Name": "gitpod/explicitFail",
+                        "Value": "IDE failed to start: fork/exec /theia/node_modules/@gitpod/gitpod-ide/startup.sh: no such file or directory",
                         "Delete": false
                     }
                 ],

components/ws-manager/pkg/manager/testdata/actOnPodEvent_failedPending_evicted_UNKNOWN01.golden

@@ -1,5 +1,18 @@
 {
     "actions": [
+        {
+            "Func": "markWorkspace",
+            "Params": {
+                "annotations": [
+                    {
+                        "Name": "gitpod/explicitFail",
+                        "Value": "Evicted: Pod The node was low on resource: [DiskPressure]. ",
+                        "Delete": false
+                    }
+                ],
+                "workspaceID": "cecf2a75-7225-4056-a125-fa3144b9c012"
+            }
+        },
         {
             "Func": "stopWorkspace",
             "Params": {

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

@@ -5,8 +5,8 @@
             "Params": {
                 "annotations": [
                     {
-                        "Name": "gitpod/failedBeforeStopping",
-                        "Value": "true",
+                        "Name": "gitpod/explicitFail",
+                        "Value": "cannot mount workspace",
                         "Delete": false
                     }
                 ],

components/ws-manager/pkg/manager/testdata/actOnPodEvent_ideFailedToStart_005_RUNNING00.golden

@@ -5,8 +5,8 @@
             "Params": {
                 "annotations": [
                     {
-                        "Name": "gitpod/failedBeforeStopping",
-                        "Value": "true",
+                        "Name": "gitpod/explicitFail",
+                        "Value": "IDE failed to start: exit status 127",
                         "Delete": false
                     }
                 ],

components/ws-manager/pkg/manager/testdata/actOnPodEvent_stuckInStopping_RUNNING00.golden

@@ -1,5 +1,18 @@
 {
     "actions": [
+        {
+            "Func": "markWorkspace",
+            "Params": {
+                "annotations": [
+                    {
+                        "Name": "gitpod/explicitFail",
+                        "Value": "container workspace ran with an error: exit code 1",
+                        "Delete": false
+                    }
+                ],
+                "workspaceID": "48f13cc4-6311-48c2-950a-4b7ddedc3037"
+            }
+        },
         {
             "Func": "stopWorkspace",
             "Params": {