#187 Check Capabilities to render PodDisruptionBudget manifest

[vovtz]

Fixes #187

[joejulian]

The PDB is critical to prevent outages, split brain, and data loss. I think we should just prevent the chart from being used on k8s < 1.21 (which was merged before I saw this issue/PR). What do you think?

[vovtz]

For our current use case there is no HA requirement, we plan to use Redpanda only to create ephemeral one-node ‘clusters’ that are used in component testing. In future we might decide to use Redpanda in production too, but for now we use Kafka.

(BTW, we obviously shouldn’t run K8s versions that are EOL anyway - we urgently requested to upgrade.)

[joejulian]

For our current use case there is no HA requirement, we plan to use Redpanda only to create ephemeral one-node ‘clusters’ that are used in component testing.

In that case, we should probably add a value under "below here be dragons" with a big warning about not using in production, then use that new value in an else, here.

something like:

yesyesyesno:
  pod_disruption_budget:
    enabled: true

Suggested change

	{{- end }}
	{{- else }}
	{{- if .Values.yesyesyesno.pod_disruption_budget.enabled }}
	fail "You must use Kubernetes 1.21+ for a production cluster. If you're not using this in production and you know why doing this is a bad idea, you can set yesyesyesno.pod_disruption_budget.enabled=false to continue."
	{{- end }
	{{- end }}

(The parent key is not what I suggest, it's just a humorous placeholder)

[vovtz]

I committed largely what you propose above, including an attempt for a more serious key name 😄

[joejulian]

Thank you. This seems like a good compromise.

[joejulian]

rebased

[joejulian]

Will rebase and merge after #202