Skip to content
Permalink

Comparing changes

Choose two branches to see what’s changed or to start a new pull request. If you need to, you can also or learn more about diff comparisons.

Open a pull request

Create a new pull request by comparing changes across two branches. If you need to, you can also . Learn more about diff comparisons here.
base repository: feast-dev/feast
Failed to load repositories. Confirm that selected base ref is valid, then try again.
Loading
base: master@{1day}
Choose a base ref
...
head repository: feast-dev/feast
Failed to load repositories. Confirm that selected head ref is valid, then try again.
Loading
compare: master
Choose a head ref
  • 2 commits
  • 6 files changed
  • 3 contributors

Commits on Aug 19, 2026

  1. docs: Add end-to-end registry deletion lifecycle example and unit test ( 

    #5360) (#6504)

* docs: add end-to-end registry deletion lifecycle snippet (#5360)

The existing registry deletion docs cover the CLI and individual Python
SDK delete methods, but lack a single copy-pasteable example showing the
full create -> verify -> delete -> confirm flow.

- Add an 'End-to-end example' snippet to registry.md that lists a feature
  view, deletes it with delete_feature_view(), and lists again to confirm.
- Add a unit test for FeatureView deletion via apply(objects_to_delete=...,
  partial=False) to guard the programmatic deletion path.

Signed-off-by: Linda Oraegbunam <obielinda@gmail.com>

* test: cover delete_feature_view, the API the docs example uses

Addresses review feedback on #6504: the new docs snippet demonstrates
store.delete_feature_view(name), but the only test added went through
apply(objects_to_delete=..., partial=False), so the documented API was
still untested.

- Add test_delete_feature_view, mirroring the registry.md snippet
  step for step: list, delete by name, list again, then assert
  get_feature_view raises FeatureViewNotFoundException.
- Add test_delete_feature_view_raises_when_missing, covering the
  FeatureViewNotFoundException that delete_feature_view's own docstring
  promises for an unregistered name.
- Keep the apply(objects_to_delete=...) test for the `feast apply` path
  documented in the hint block, and note in its docstring that it is
  deliberately distinct from delete_feature_view.
- Lift the shared source frame and registration into two helpers so the
  two lifecycle tests do not duplicate ~30 lines of setup.

Signed-off-by: Linda Oraegbunam <obielinda@gmail.com>

---------

Signed-off-by: Linda Oraegbunam <obielinda@gmail.com>
    @obielin
    obielin authored Aug 19, 2026
    Configuration menu
    Copy the full SHA
    aade119 View commit details
    Browse the repository at this point in the history

  2. ci: Add bundle-sync verification to operator PR workflow (#6751) 

    * ci: Add bundle-sync verification to operator PR workflow

Signed-off-by: ntkathole <nikhilkathole2683@gmail.com>

* fix: Regenerate operator bundle to sync RBAC permissions

Run `make bundle` to pick up refined clusterrole/clusterrolebinding
RBAC rules and drop stale subjectaccessreviews permission.

Signed-off-by: ntkathole <nikhilkathole2683@gmail.com>

* fix: Retry on AlreadyExists for cluster-scoped RBAC resources

retry.RetryOnConflict only handles Conflict (resource version mismatch).
When concurrent reconcile loops both GET a NotFound resource and race
to Create it, the loser gets AlreadyExists which was not retried.

Switch to retry.OnError with a predicate covering both IsConflict and
IsAlreadyExists so the retry re-GETs the now-existing resource and
proceeds with an update.

Signed-off-by: ntkathole <nikhilkathole2683@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

* fix: Regenerate operator bundle to sync RBAC and CRD changes

Run make bundle to pick up the clusterrole get/list RBAC verbs and
updated CRD field descriptions.

Signed-off-by: ntkathole <nikhilkathole2683@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

* fix: Handle AlreadyExists and Conflict as non-errors for cluster RBAC

The controller-runtime cached client can return stale NotFound for
cluster-scoped resources when the informer cache has not yet synced.
This causes CreateOrUpdate to attempt a Create that fails with
AlreadyExists. Retrying does not help because the cache remains
stale during the short retry window.

Treat AlreadyExists and Conflict as non-errors since the resource
exists in the desired state. The next reconcile cycle will update
its contents once the cache has synced. This follows the standard
Kubernetes operator eventual-consistency pattern.

Signed-off-by: ntkathole <nikhilkathole2683@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

---------

Signed-off-by: ntkathole <nikhilkathole2683@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
    @ntkathole @cursoragent
    ntkathole and cursoragent authored Aug 19, 2026
    Configuration menu
    Copy the full SHA
    e79bd33 View commit details
    Browse the repository at this point in the history
Loading