Fix not-found handling on delete and fix event msg

GoogleCloudPlatform · Mar 6, 2024 · 4fe6ac9 · 4fe6ac9
1 parent 33b641a
commit 4fe6ac9
Show file tree

Hide file tree

Showing 3 changed files with 5 additions and 2 deletions.
diff --git a/tpu-provisioner/internal/cloud/gke.go b/tpu-provisioner/internal/cloud/gke.go
@@ -105,6 +105,9 @@ func (g *GKE) DeleteNodePoolForNode(node *corev1.Node) error {
 
 	op, err := g.Service.Projects.Locations.Clusters.Delete(g.ClusterContext.NodePoolName(name)).Do()
 	if err != nil {
+		if gerr, ok := err.(*googleapi.Error); ok && gerr.Code == http.StatusNotFound {
+			return nil
+		}
 		return fmt.Errorf("deleting node pool %q: %w", name, err)
 	}
 

diff --git a/tpu-provisioner/internal/controller/deletion_controller.go b/tpu-provisioner/internal/controller/deletion_controller.go
@@ -149,7 +149,7 @@ func (r *DeletionReconciler) Reconcile(ctx context.Context, req ctrl.Request) (c
 			return ctrl.Result{}, nil
 		} else {
 			r.Recorder.Event(&node, corev1.EventTypeWarning, EventFailedDeletingNodePool, "Failed to delete Node Pool: "+err.Error())
-			return ctrl.Result{}, client.IgnoreNotFound(err)
+			return ctrl.Result{}, err
 		}
 	}
 	r.Recorder.Event(&node, corev1.EventTypeNormal, EventNodePoolDeleted, DeletedNodePoolEventMessage)

diff --git a/tpu-provisioner/internal/controller/events.go b/tpu-provisioner/internal/controller/events.go
@@ -9,6 +9,6 @@ const (
 	EventNodePoolEnsured         = "NodePoolEnsured"
 	EventDeletingNodePool        = "DeletingNodePool"
 	EventNodePoolDeleted         = "NodePoolDeleted"
-	DeletingNodePoolEventMessage = "Deleted Node Pool."
+	DeletingNodePoolEventMessage = "Deleting Node Pool."
 	DeletedNodePoolEventMessage  = "Deleted Node Pool."
 )