Deploying multiple operator instances fails

[abrodersen]

This appears to be caused by #832.

Deploying the first release of the operator using chart version 1.4.0 works. However, when deploying a second instance, helm returns this error:

Error: UPGRADE FAILED: kind ClusterRole with the name "postgres-pod" already exists in the cluster and wasn't defined in the previous release. Before upgrading, please either delete the resource from the cluster or remove it from the chart

I tried setting rbac.create to false but this causes other necessary RBAC resources to not get created. The chart also does not offer any way to customize the name of the postgres-pod ClusterRole. If this customization were available, multiple instances could be deployed without issue.

[FxKu]

The name of this cluster role should indeed be configurable. Actually the pod_service_account_name field is meant for that. Would it work for you, to change the values.yaml for each operator you are deploying? Or do you want a generated name from helm?

[abrodersen]

@FxKu thank you! It looks like your pull request will resolve the issue.

[FxKu]

@abrodersen we merged it. Thanks for raising this. By default the additional cluster role is still named postgres-pod. But you can remove the name, and it will be generated by helm. With this creating multiple clusters should work.

If you want the two clusters to not interfere with each other either configure the watched_namespace option or enable setting a controller ID. However this is a very new feature and not available in the 1.4.0 release.