Checklist:

• I've included steps to reproduce the bug.
• I've included the version of argo rollouts.

Describe the bug

Rollout controller in a infinity loop trying to reconcile a still referenced replicaset.

Deploy 1 -> Deploy2-> delay on service switch (pods start time) -> scale event

All reconciles in this rollout execute the isEscaleEvent, and will never make the service switch.

To Reproduce

I made this e2e test, need to use a E2E_POD_DELAY=3 for the service switch delay

func (s *CanarySuite) TestRolloutScalingWhenChanged() {
	s.Given().
		RolloutObjects(`@functional/alb-canary-rollout.yaml`).
		When().
		ApplyManifests().
		WaitForRolloutStatus("Healthy").
		UpdateSpec().
		WaitForRolloutStatus("Paused").
		UpdateSpec().
                Sleep(2*time.Second). // Should be less than E2E_POD_DELAY
		ScaleRollout(3).
                Sleep(30*time.Second). // Just to stop the execution before delete the rollout
		Then().
		ExpectRevisionPodCount("1", 3).
		ExpectRevisionPodCount("2", 0).
		ExpectRevisionPodCount("3", 1)
}

Expected behavior

The controller should be able to reconcile the rollout/services.

Version

1.6.6

Message from the maintainers:

Impacted by this bug? Give it a 👍. We prioritize the issues with the most 👍.