Insure that any resource that is prunable can be pruned

[dturn]

Resources with deploy_method replace* don't go through an apply cycle. Since pruning happens in the apply cycle, they are never pruned. Ensure that everything that is prunable can actually be pruned.

Allow PDBs to be pruned too
#279

[dturn]

It looks like the apply issue has been resolved: kubernetes/kubernetes#53379

Think we should support pruning on replaceable resources?

[KnVerey]

If I'm remembering correctly, the :replace method was add because of that bug. I'd say if the fix was cherry-picked to all currently supported versions (looks like it was), we should go back to applying all CRs. At one point we were discussing relying on the deploy clearing the non-subresource status, but since we decided against that, I don't think we have a good reason to be replacing anymore.

Looks like we still need replace_force though, as the PDB issue is still open. I'd consider the fact that those aren't pruned to be a bug, but not relevant to your current project.

[dturn]

@Shopify/cloudx 👀 please

[KnVerey]

• Statefulservice still has replace in it
• Please make sure to 🎩 the branch before merging by locally deploying a test app with multiple Shopify CRs to a cluster that actually supports them. Make sure to try a deploy that includes a change to the CR's spec (i.e. do a deploy that would have actually triggered the bug we were working around with :replace). There is pretty much zero test coverage for those files because our controllers are not deployed in CI (yet another reason to get rid of them in favour of the dynamic feature!).

[dturn]

Statefulservice still has replace in it

This was intentional. I was unsure why SS are replace since doesn't seem like they would be part of the CRD merge bug.

[dturn]

🎩 with cloud-portal in the staging cluster by increasing & decreasing redis size. Let me know if you want to other resources.

[KnVerey]

This was intentional. I was unsure why SS are replace since doesn't seem like they would be part of the CRD merge bug.

Are you thinking StatefulSet rather than StatefulService, perhaps? StatefulSet are a core resource and are already applied; StatefulService are CRs and were replaced for the same reason as the others. Confusing naming for sure.

with cloud-portal in the staging cluster by increasing & decreasing redis size. Let me know if you want to other resources

I'd suggest testing a change with a more complicated patch just in case. For example. a change inside an array or a hash (I'm actually curious which strategy it uses by default... for core resources it is field-specific). ES probably has the most elaborate schema, and cumulus-cat has one so could be a good target for testing.

[dturn]

Are you thinking StatefulSet rather than StatefulService, perhaps? StatefulSet are a core resource and are already applied; StatefulService are CRs and were replaced for the same reason as the others. Confusing naming for sure.

Yep I definitely got StatefulService confused with StatefulSet

I'll 🎩 something more complicated after fixing up StatefulSet

[dturn]

🎩 a more complicated change went smoothly.

[dturn]

Ready for final 👀

[KnVerey]

🚢