feat: File persistence definition and implementation

[dmartinol]

What this PR does / why we need it:

Add ephemeral persistence configuration to the Feast services deployed with the operator.
A sample manifest is also committed and reported here for your convenience:

apiVersion: feast.dev/v1alpha1
kind: FeatureStore
metadata:
  name: sample-ephemeral-persistence
spec:
  feastProject: my_project
  services:
    onlineStore:
      persistence:
        file:
          path: /data/online_store.db
    offlineStore:
      persistence:
        file:
          type: dask
    registry:
      local:
        persistence:
          file:
            path: /data/registry.db

Next PR will add persistent storage by using mounted volumed (Kubernetes PVC)

Which issue(s) this PR fixes:

Relates to #4561

[dmartinol]

@tchughesiv @tmihalac
Note: validations will be added to the CRD types as the other options will be integrated.

[tchughesiv]

same

[tchughesiv]

would be good to have unit tests ensuring that proper defaults are being set in status when varying pointers around service persistence are set in the spec.

[dmartinol]

would be good to have unit tests ensuring that proper defaults are being set in status when varying pointers around service persistence are set in the spec.

This was the idea of repo_config_test.go, which verifies the repo config created after applying the defaults.

[tchughesiv]

that's a great addition, and necessary, but its different from ensuring the status defaults are properly being reconciled in the CR.

[dmartinol]

I see what you mean, I'll add it in the next PR for PVC so there will be more cases to cover

[tchughesiv]

lgtm

[tchughesiv]

requested a couple changes. also, i'm noticing we're missing new FeatureStore CR checks in featurestore_controller_test.go for these new status fields and defaults after reconcile

[dmartinol]

requested a couple changes. also, i'm noticing we're missing new FeatureStore CR checks in featurestore_controller_test.go for these new status fields and defaults after reconcile

one more commit to add the missing checks

[tchughesiv]

i've left a couple nits, but otherwise lgtm

[dmartinol]

requested a couple changes. also, i'm noticing we're missing new FeatureStore CR checks in featurestore_controller_test.go for these new status fields and defaults after reconcile

one more commit to add the missing checks

Latest commit includes new controller tests.
Please note that I had to change the resource name because some resources (svc?) are probably not deleted by previous tests, so it would fail due to an attempt to change ownership:

Object default/feast-services-offline is already owned by another Service controller feast-services-offline

I did not investigate further. I think we can handle this issue in a separate PR.

  2024-11-07T17:30:30+01:00     INFO    Successfully reconciled {"Service": "feast-services-offline", "operation": "updated"}
  2024-11-07T17:30:30+01:00     ERROR   Error deploying the FeatureStore client service {"error": "Object default/feast-services-offline is already owned by another Service controller feast-services-offline"}
  github.com/feast-dev/feast/infra/feast-operator/internal/controller/services.(*FeastServices).setFeastServiceCondition
        /Users/dmartino/projects/AI/feast/feast/infra/feast-operator/internal/controller/services/services.go:290
  github.com/feast-dev/feast/infra/feast-operator/internal/controller/services.(*FeastServices).deployFeastServiceByType
        /Users/dmartino/projects/AI/feast/feast/infra/feast-operator/internal/controller/services/services.go:95
  github.com/feast-dev/feast/infra/feast-operator/internal/controller/services.(*FeastServices).Deploy

[tchughesiv]

I did not investigate further. I think we can handle this issue in a separate PR.

it's not an "issue" per say ... it's got to do w/ another unit test from a prior run where i've changed object ownership so we can test that ownership conflicts/failures are working properly. that said, we prob should manually cleanup that object after the fact.

so, yea ... i guess it is an issue :)