[ws-manager-mk2] Remember existing workspaces after a restart (#17066)

Author: WVerlaek

components/ws-daemon/pkg/controller/workspace_controller.go

@@ -302,7 +302,9 @@ func (wsc *WorkspaceController) handleWorkspaceStop(ctx context.Context, ws *wor
 		wsc.metrics.recordFinalizeTime(time.Since(disposeStart).Seconds(), ws)
 	}
 
-	wsc.emitEvent(ws, "Backup", fmt.Errorf("failed to backup workspace: %w", disposeErr))
+	if disposeErr != nil {
+		wsc.emitEvent(ws, "Backup", fmt.Errorf("failed to backup workspace: %w", disposeErr))
+	}
 
 	err = wsc.operations.DeleteWorkspace(ctx, ws.Name)
 	if err != nil {

components/ws-manager-mk2/controllers/metrics.go

@@ -186,6 +186,10 @@ func (m *controllerMetrics) countTotalRestoreFailures(log *logr.Logger, ws *work
 	counter.Inc()
 }
 
+func (m *controllerMetrics) containsWorkspace(ws *workspacev1.Workspace) bool {
+	return m.cache.Contains(ws.Name)
+}
+
 func (m *controllerMetrics) rememberWorkspace(ws *workspacev1.Workspace, state *metricState) {
 	var s metricState
 	if state != nil {

components/ws-manager-mk2/controllers/workspace_controller.go

@@ -148,6 +148,11 @@ func (r *WorkspaceReconciler) Reconcile(ctx context.Context, req ctrl.Request) (
 func (r *WorkspaceReconciler) actOnStatus(ctx context.Context, workspace *workspacev1.Workspace, workspacePods corev1.PodList) (ctrl.Result, error) {
 	log := log.FromContext(ctx)
 
+	if workspace.Status.Phase != workspacev1.WorkspacePhaseStopped && !r.metrics.containsWorkspace(workspace) {
+		// If the workspace hasn't stopped yet, and we don't know about this workspace yet, remember it.
+		r.metrics.rememberWorkspace(workspace, nil)
+	}
+
 	if len(workspacePods.Items) == 0 {
 		// if there isn't a workspace pod and we're not currently deleting this workspace,// create one.
 		switch {
@@ -193,7 +198,6 @@ func (r *WorkspaceReconciler) actOnStatus(ctx context.Context, workspace *worksp
 
 				r.Recorder.Event(workspace, corev1.EventTypeNormal, "Creating", "")
 			}
-			r.metrics.rememberWorkspace(workspace, nil)
 
 		case workspace.Status.Phase == workspacev1.WorkspacePhaseStopped:
 			if err := r.deleteWorkspaceSecrets(ctx, workspace); err != nil {