Skip to content

Auto sync dev to stable (up to 28a60bd7)#119

Merged
github-actions[bot] merged 2 commits intostablefrom
lake-gate-20260320-144054
Mar 20, 2026
Merged

Auto sync dev to stable (up to 28a60bd7)#119
github-actions[bot] merged 2 commits intostablefrom
lake-gate-20260320-144054

Conversation

@github-actions
Copy link
Copy Markdown

Automated Sync from dev to stable

This PR automatically syncs the dev branch to the stable branch by creating a temporary branch directly from dev.

Latest commit: 28a60bd - fix: Add Secret to ByObject label filter for defense-in-depth
Total commits to sync: 2

⚠️ IMPORTANT: How to Merge

🚫 DO NOT use the GitHub merge button!

Once all checks are complete, comment /approve on this PR to automatically fast-forward merge the stable branch to point to the same commit as this temporary branch.

Only use the /approve command - the GitHub merge button will not work correctly for this automated sync process.

Pre-merge Checklist

Before approving this PR, please ensure the following tasks are completed:

  • PR checks: PR checks passed
  • E2E tests: Smoke and sanity tests passed against cluster with training operator image built from this PR
    • Build operator image locally (podman build -f build/images/training-operator/Dockerfile.rhoai .), upload it to quay
    • Adjust params.env file to reference uploaded image
    • Provision OCP cluster with latest ODH nightly
    • Install modified operator using update-training.sh (or manually)
    • Wait until training operator Pod is reprovisioned
    • Run smoke and sanity tests using Jenkins job or manually

Commits to be synced:

  • 22b0f90: fix: Restrict informer cache to operator-managed resources
  • 28a60bd: fix: Add Secret to ByObject label filter for defense-in-depth

This PR was created automatically by the sync workflow.

Add label-selector filtering on the controller-runtime informer cache
so only resources carrying the operator-name label are cached. Without
filtering, the cache loads all Pods, Services, ConfigMaps, and RBAC
objects cluster-wide, causing unbounded memory growth under load.

Changes:
- Add newCacheOptions() with ByObject label selectors using "exists"
  requirement on the operator-name label
- Strip managed fields from cached objects (TransformStripManagedFields)
- Disable cache for ConfigMap and Secret client reads (DisableFor)
- Add operator-name label to MPI controller resources (ConfigMap,
  ServiceAccount, Role, RoleBinding) that were missing it
- Set GOMEMLIMIT=460MiB and add resource limits (512Mi) to deployment
Add Secret to the ByObject cache configuration with the same
operator-name label filter used for other core types. While no
controller currently creates a Secret informer, this prevents
an unfiltered cluster-wide Secret watch if one is added in the future.
@sutaakar
Copy link
Copy Markdown

/approve

@github-actions github-actions Bot merged commit 28a60bd into stable Mar 20, 2026
11 of 13 checks passed
@github-actions
Copy link
Copy Markdown
Author

✅ Approved and merged! The stable branch has been fast-forwarded to point to the same commit as this temporary branch.

@github-actions github-actions Bot deleted the lake-gate-20260320-144054 branch March 20, 2026 14:42
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants