Related? #22512

/cc @janetkuo

@aronchick is this a dupe of #25238?

This is working as intended. Scaling is orthogonal to deployment.

What if the user set up HPA? Should HPA's changes be undone?

Scale is set/changed at least 3 different ways above (create, scale, apply). In that scenario, there's no way we could infer user intent regarding whether they expected scale to be undone or not. Note that the Deployment API doesn't really know the difference between apply and other commands, and apply doesn't itself cause Deployment versioning to happen. It's changes in the pod template spec that do, and scaling doesn't change the pod template spec.

@janetkuo is this documented in undo?

@janetkuo is this documented in undo?

Yes, it is in http://kubernetes.io/docs/user-guide/deployments/#rolling-back-a-deployment:

You should note that only the update to the Deployment’s “pod template” is seen as a different revision of rollout.

I'll make it more obvious, considering most people would skim the doc.

Updated docs kubernetes/website#484

Closing since doc is updated

I didn't intend to change both # of replicas & image; i overlooked the fact that the # of replicas was different when I rolled out the new image.

I have a new template that has a number of replicas in it. When I applied that new template, it used that new number in it.

It still feels, for better or worse, that ANY changes I made should be "undone" when I roll back given what I'm doing. Documentation will not address this - years of hitting cmd-Z on my laptop tells me that whatever I just did is now rolled back, that's what this felt like.

Alternative may be if you're doing a roll out that changes multiple things, maybe we highlight that?

$ kubectl apply -f new_ds.yml
This change will apply the following changes:
replicas: 5 -> 7
image: foo -> bar
y/n?

Doesn't solve my problem entirely - since when I undo/rollback, I still would have expected 7 to go back to 5 and bar to go back to foo.

We could make it configurable at the very least. Something like --change-scaling-settings (we already do this in OpenShift with oc rollback. ). @pwittrock this is where we need to know the last size for a replica set and https://github.com/kubernetes/kubernetes/pull/28162/files#diff-9035ca88b01f4a0aab7c52b8bf8a753fR79 will be useful

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

Prevent issues from auto-closing with an /lifecycle frozen comment.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or @fejta.
/lifecycle stale

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or @fejta.
/lifecycle rotten
/remove-lifecycle stale

Working as intended