[ws-manager] Remove finalizer once WS is stopped

Author: csweichel

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

@@ -372,11 +372,6 @@ func (m *Monitor) actOnPodEvent(ctx context.Context, status *api.WorkspaceStatus
 	}
 
 	if status.Phase == api.WorkspacePhase_STOPPING {
-		if status.Conditions.FinalBackupComplete == api.WorkspaceConditionBool_TRUE {
-			// we've disposed already - try to remove the finalizer and call it a day
-			return m.manager.modifyFinalizer(ctx, workspaceID, gitpodFinalizerName, false)
-		}
-
 		var terminated bool
 		for _, c := range wso.Pod.Status.ContainerStatuses {
 			if c.Name == "workspace" {
@@ -396,6 +391,11 @@ func (m *Monitor) actOnPodEvent(ctx context.Context, status *api.WorkspaceStatus
 		}
 	}
 
+	if status.Phase == api.WorkspacePhase_STOPPED {
+		// we've disposed already - try to remove the finalizer and call it a day
+		return m.manager.modifyFinalizer(ctx, workspaceID, gitpodFinalizerName, false)
+	}
+
 	return nil
 }
 

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

@@ -394,6 +394,14 @@ func (m *Manager) extractStatusFromPod(result *api.WorkspaceStatus, wso workspac
 
 			if ds.BackupComplete {
 				result.Conditions.FinalBackupComplete = api.WorkspaceConditionBool_TRUE
+
+				// Finalizer or not - once the backup is complete we consider the pod stopped.
+				// Once the finalizer is removed, there's no guarantee we see the pod again, because it might be
+				// deleted too quickly for us to handle. Hence, we consider the workspace stoppped once the backup
+				// is done, even though the finalizer might still be present.
+				//
+				// This runs the risk that a pod might still be present, but is considered stopped.
+				result.Phase = api.WorkspacePhase_STOPPED
 			}
 			result.Repo = ds.GitStatus
 

components/ws-manager/pkg/manager/testdata/status_containerd4214_STOPPING00.golden

@@ -0,0 +1,35 @@
+{
+    "status": {
+        "id": "cc1a3979-4e0e-42cc-bba5-a8d66485bdee",
+        "metadata": {
+            "owner": "26488517-271b-44f5-8739-96a6d9f6759a",
+            "meta_id": "amber-swallow-sbhsd9w0",
+            "started_at": {
+                "seconds": 1617004536
+            }
+        },
+        "spec": {
+            "workspace_image": "eu.gcr.io/gitpod-core-dev/registry/workspace-images:4821db66dd61e6944f8940c4328386612ccdb60d542d65f0456d4fd38f51f36a",
+            "ide_image": "eu.gcr.io/gitpod-core-dev/build/ide/code:cw-wsdaemon-patch.0",
+            "url": "https://amber-swallow-sbhsd9w0.ws-dev.cw-wsdaemon-patch.staging.gitpod-dev.com",
+            "timeout": "30m"
+        },
+        "phase": 6,
+        "conditions": {
+            "final_backup_complete": 1,
+            "deployed": 1
+        },
+        "repo": {
+            "branch": "master",
+            "latest_commit": "82457ccee52cfec99c3d491063f9fadfdd78607d"
+        },
+        "runtime": {
+            "node_name": "gke-dev-workload-1-49d27f81-hhx0",
+            "pod_name": "ws-cc1a3979-4e0e-42cc-bba5-a8d66485bdee",
+            "node_ip": "10.132.0.14"
+        },
+        "auth": {
+            "owner_token": "{pKaZ75.$$hIiW2z2!-h#HcmldG#U?Dl"
+        }
+    }
+}