fix: use UpdatedReplicas instead of ReadyReplicas in observeCurrentImage

ReadyReplicas requires pod readiness probes to pass, which depends on
the sidecar reporting ready via /v0/healthz. But the sidecar only
becomes ready after the MarkReady plan step runs, which is sequenced
after AwaitSpecUpdate, which reads currentImage, which requires
observeCurrentImage to succeed. Circular dependency.

UpdatedReplicas counts pods running the desired template regardless of
readiness probe status. This correctly signals "the image change took
effect" without requiring the full sidecar initialization to complete.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: bdchatham

internal/controller/node/controller.go

@@ -197,10 +197,7 @@ func (r *SeiNodeReconciler) observeCurrentImage(ctx context.Context, node *seiv1
 		return err
 	}
 
-	if sts.Status.CurrentRevision == "" || sts.Status.CurrentRevision != sts.Status.UpdateRevision {
-		return nil
-	}
-	if sts.Status.ReadyReplicas < 1 {
+	if sts.Spec.Replicas == nil || sts.Status.UpdatedReplicas < *sts.Spec.Replicas {
 		return nil
 	}
 

internal/controller/node/reconciler_test.go

@@ -249,9 +249,7 @@ func TestObserveCurrentImage_UpdatesWhenConverged(t *testing.T) {
 		WithStatusSubresource(&seiv1alpha1.SeiNode{}, &appsv1.StatefulSet{}).
 		Build()
 
-	sts.Status.CurrentRevision = testRevision
-	sts.Status.UpdateRevision = testRevision
-	sts.Status.ReadyReplicas = 1
+	sts.Status.UpdatedReplicas = 1
 	g.Expect(c.Status().Update(ctx, sts)).To(Succeed())
 
 	r := &SeiNodeReconciler{Client: c, Scheme: s}