Support csv-format for WATCH_NAMESPACE env var

[jkremser]

the current implementation gives a false impression that multiple namespaces are supported (method called GetWatchNamespaces + plular used here). However, it's not the case. No matter what's in the WATCH_NAMESPACE env var, it just passes its content to the cache options when calling NewManager in operator's main.go.

Let's change that :)

This couple of lines actually make it work so that user can specify WATCH_NAMESPACE=ns1,default,foo or just a singe ns and keda will watch for events in those whiteallowlisted namespaces. If left empty, it still watches for all.

I tested this locally by:

make run ARGS="--zap-log-level=debug --cert-dir=/Users/jkremser/workspace/keda" WATCH_NAMESPACE="ns1,ns2,default" KEDA_CLUSTER_OBJECT_NAMESPACE="keda"

# then creating ScaledObjects in those ^ namespaces worked, while those in other namespaces were ignored

Fix #5670

[zroubalik]

Thanks, we should probably apply this also to webhook: https://github.com/kedacore/keda/blob/main/cmd/webhooks/main.go

[zroubalik]

Thanks, we should probably apply this also to webhook: https://github.com/kedacore/keda/blob/main/cmd/webhooks/main.go

Not a good idea actually. Let's not do that at the moment.

[jkremser]

👍 for the record:

having WATCH_NAMESPACE implemented also for the validating admission controller would lead to a situation where in certain namespace the malformed, say, ScaledObject could be created and in other it would be blocked. This feels too magic and isn't very standard.

If we changed the logic of the webhook is a way that it would be even blocking creation of those CRs in namespaces that are not watched (w/ some sensible message) that would complicate the possible scenario of having multiple KEDAs (with various configuration) installed in the same k8s cluster watching different namespaces.

Thanks, @zroubalik

[zroubalik]

@jkremser please update changelog

[jkremser]

@zroubalik should be there now

[zroubalik]

Could you please open a doc PR to document this setting? Probably a new section on this page:
https://keda.sh/docs/2.14/operate/cluster/

[jkremser]

docs: kedacore/keda-docs#1357

[zroubalik]

/run-e2e internal
Update: You can check the progress here