Hello,
I know that you've already opened a PR with a fix, but I'm interested in why you need to set MaxReplicaCount: 0.
Is it for pausing the autoscaling? If yes, KEDA already supports a better approach

Hi @JorTurFer,
there is a huge difference between pausing Keda and setting MaxReplicaCount to 0:
pausing leaves all the running pods running;
setting MaxReplicaCount to 0 shuts all the pods down.

In some cases we need to scale the service down completely, currently we have to add pause annotation and run kubectl scale --replicas=0 deployment .... This is inconvenient, requires people to run manual unreviewed commands (there is always chance of a mistake) and leaves records in different places, so it's harder to debug during incidents.

Yeah, I totally understand the requirement of scaling to 0 for some reasons, but that's already supported too xD

If you annotate the ScaledObject with autoscaling.keda.sh/paused: "true", you are right and KEDA will lock current instances as the desired instances, leaving all the pods running but, If you annotate the ScaledObject with autoscaling.keda.sh/paused-replicas: "0", KEDA will kill all the pods scaling to 0 the workload instantly and locking the scaling out.

In fact, you can use autoscaling.keda.sh/paused-replicas: "YOUR_DESIRED_VALUE" to enforce an arbitrary amount of pods meanwhile the ScaledObject is annotated. Then when you remove the annotation, the scaling will start from where it is on the annotation removal

I missed that part somehow, thanks!

We are migrating from https://github.com/jthomperoo/custom-pod-autoscaler and we used to scale down by setting maxReplicas to 0.

I still think that setting MaxReplicaCount to 0 is simpler for the task, no need to create custom annotations (it's easy to make small mistake in them and spend a lot of time trying to understand what's going on) and we won't need to re-educate people in our company. Also my PR would fix very strange error MinReplicaCount=1 must be less than MaxReplicaCount=0 for all people trying to set MaxReplicaCount to 0. I spend hours debugging what's going on.

Yeah, definitivelly, I don't have anything against the PR, I just suggested as an already available option to speed up your processes xD

autoscaling.keda.sh/paused-replicas: "0" solves this, it also removes the HPA. I am not sure we should allow max == 0. Fixint the error would be nice though.

Removing HPA is undesired in our case, if HPA would stay in place it will make sure that the deployment will stay scaled down, even if someone scaled it up by mistake.

BTW, there are more people expecting this behaviour, see #3013 and #3014.

What would be a downside of supporting it?

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.

This issue can be fixed by the #5571, any reason not to merge it?

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.

This issue can be fixed by the #5571, any reason not to merge it?

This issue has been automatically closed due to inactivity.