Fix #191 - Changing deployment strategy to recreate and setting replicas to max 1

[ricardozanini]

Fixes #191

@LCaparelli do you mind taking a look? Maybe improving the doc a little bit? I'm lacking creativity...

@bszeti I'll push an image with this fix soon, do you mind taking testing in your local environment? As soon as we review and test this PR, I'll release a new version with this fix for you.

Thanks!

[ricardozanini]

@LCaparelli @bszeti the image is available at: quay.io/m88i/nexus-operator:0.5.0-cr1 can be verified with the [nexus-operator.yaml file](https://github.com/m88i/nexus-operator/blob/main/nexus-operator.yaml). Just change the image tag there.

[ricardozanini]

OLM test may fail since they changed some bits. @Kaitou786 since you've changed that on Kogito, mind replicating the fix here as well? 🤭

[LCaparelli]

Thanks a lot for this @ricardozanini! I didn't find the doc section lacking, it's straightforward and clear, nothing to add on my behalf but linking the HA page from Sonatype.

I do think the logic that controls spec.replica should be placed at a default setting method, not a validation one.

[LCaparelli]

@ricardozanini just tested the release candidate and things are working as expected. Setting spec.replicas to 3 changes it back to 1, it gets logged fine and any changes which cause a new ReplicaSet to be created now destroy the existing pods from the previous RS first. :-D

[ricardozanini]

@LCaparelli can you take a final look? I've moved to defaults instead. Hopefully, with admission controllers, we wouldn't need that, since the CR won't be created.

[LCaparelli]

@LCaparelli can you take a final look? I've moved to defaults instead.

This looks great man, thanks!

Hopefully, with admission controllers, we wouldn't need that, since the CR won't be created.

Well, we will, but it will be part of the mutating webhook, not part of the reconciliation logic.