Hooked job deletion incompatible with Helm --wait and FluxCD by default

[alee792]

The issues described in this comment will manifest automatically when using FluxCD to manage helmreleases. FluxCD waits by default, and the HelmRelease resource needs spec.upgrade.disableWait: true set for the OpenFGA Helm chart to continue functioning. Without it, the job is deleted before the deployment can check the job.

I don't have any suggestions for how to effectively delete/recreate the job, but figured I'd at least create the issue in case others are using FluxCD and are stumped.

[tamalsaha]

The migrate hook never runs because the openfga pods wait for the migrate job to complete when --wait flag is used. Here is a similar issue: fluxcd/flux2#1085 (comment)

I think one way to solve this will be:

• Don't use hook to run the migrate job.
• Increase the timeout and backoffLimit of the migrate job so it waits and retries until the db pods are ready.

I have done this in my fork and it works. Let me know if openfga interested in a pr.

[alee792]

My experience with the team is that they're very open and responsive to contributions! If you put it up for a PR, I'll help promote and review it.