pachyderm / openshift-operator Goto Github PK

In some user environments, custom resources that represent deployments may be maintained as files in github or other git repositories. The Pachyderm CR includes credentials for S3 buckets, which would need to either be stored separately in a kubernetes Secret, or be stored on the Pachyderm CR in an encrypted form, so that the s3 credentials are not exposed on git repositories.

It should be possible to recover from a bad state without needing to delete the CRD instance and start over.

TODO: Add Repro steps

If I try and deploy the sample in the current repo, I get an index out of bounds exception. This can be traced to the new pachyderm version code, we should present a better error message here if the pach version is not specified in the given CRD instance. We should also add Pachyderm version to config/samples/aiml_v1beta1_pachyderm.yaml

Values in the values.yaml should only be passed through if they are specified in the given CRD. That way they can rely on chart defaults. As an example, the chart fails to apply by default as reverse gets set to empty string, which is not a valid boolean value

Error message:

values don't meet the specifications of the schema(s) in the following chart(s):\npachyderm:\n- pachd.storage.amazon.reverse: Invalid type. Expected: boolean, given: string\n

This is because the value is rendered out as

reverse: <nil>

https://github.com/pachyderm/openshift-operator/blob/main/hack/charts/2.0.3/values.yaml#L297

We have inputs in the CRD for image repository / tag, but we don't use those values if they are specified in the CRD (instead using the values in images.json) I believe the ordering should be

1. Use image and tag values from the crd if specified
2. Fall back on the images.json values