fix: ensure GPU is retrieved before starting worker job (#26)

0x5457 · web-flow · commit 6b276e8fd8e2 · 2025-01-02T22:43:41.000+08:00
diff --git a/internal/controller/tensorfusionconnection_controller.go b/internal/controller/tensorfusionconnection_controller.go
@@ -116,7 +116,13 @@ func (r *TensorFusionConnectionReconciler) Reconcile(ctx context.Context, req ct
 		}
 	}
 
-	if connection.Status.Phase != tfv1.TensorFusionConnectionPending {
+	if gpu == nil && connection.Status.GPU != "" {
+		if err := r.Get(ctx, client.ObjectKey{Name: connection.Status.GPU}, gpu); err != nil {
+			log.Error(err, "Failed to get GPU.", "gpu", connection.Status.GPU)
+		}
+	}
+
+	if connection.Status.Phase != tfv1.TensorFusionConnectionPending && gpu != nil {
 		// Start worker job
 		workerPod, err := r.tryStartWorker(ctx, gpu, connection, types.NamespacedName{Name: connection.Name, Namespace: connection.Namespace})
 		if err != nil {

Original file line number	Diff line number	Diff line change
`@@ -116,7 +116,13 @@ func (r *TensorFusionConnectionReconciler) Reconcile(ctx context.Context, req ct`
`116`	`116`	`}`
`117`	`117`	`}`
`118`	`118`
`119`		`- if connection.Status.Phase != tfv1.TensorFusionConnectionPending {`
	`119`	`+ if gpu == nil && connection.Status.GPU != "" {`
	`120`	`+ if err := r.Get(ctx, client.ObjectKey{Name: connection.Status.GPU}, gpu); err != nil {`
	`121`	`+ log.Error(err, "Failed to get GPU.", "gpu", connection.Status.GPU)`
	`122`	`+ }`
	`123`	`+ }`
	`124`	`+`
	`125`	`+ if connection.Status.Phase != tfv1.TensorFusionConnectionPending && gpu != nil {`
`120`	`126`	`// Start worker job`
`121`	`127`	`workerPod, err := r.tryStartWorker(ctx, gpu, connection, types.NamespacedName{Name: connection.Name, Namespace: connection.Namespace})`
`122`	`128`	`if err != nil {`