Repository navigation
feat(ci): run plugin-sanity daily for the release branch indexes - #5704
Closed
gustavolira wants to merge 3 commits into
Closed
gustavolira wants to merge 3 commits into
gustavolira wants to merge 3 commits into
Conversation
plugin-sanity only ran on schedule on main, against the next catalog index. Release branches could only be checked by a manual workflow_dispatch, which needs repository admin rights, so RC and z-stream indexes were never checked automatically. GitHub only fires `schedule` on the default branch, so a new release-plugin-sanity job on main's schedule dispatches the workflow on the two newest release-X.Y branches that carry the harness and waits for those runs, failing if any fails. Each branch runs its own harness against its own backend and plugin-catalog-index:X.Y tag; mixing main's backend with a release index would test a combination that is never shipped. The dispatch API's return_run_details gives the run id to wait on. Branches older than the harness (1.10 and earlier) stop the search. Also drop four excludes whose reason no longer holds. homepage, both intelligent-assistant packages (RHDHBUGS-3515) and the keycloak auth provider (RHDHBUGS-3678, closed as obsolete) were excluded because their refs could not be pulled anonymously. The next and 2.1 indexes now pin them by digest and `skopeo inspect --no-creds` resolves all four. Fork runs with them enabled passed on main (next index) and on release-2.1 (2.1 index): https://github.com/gustavolira/rhdh/actions/runs/39851389654 https://github.com/gustavolira/rhdh/actions/runs/39851624217 The two orchestrator excludes (RHIDP-15961) stay, now noting where the orchestrator backend is covered instead. The harness doc gains a plugin sanity section. Part of RHIDP-18205. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
gustavolira
added a commit
to gustavolira/rhdh
that referenced
this pull request
Oct 22, 2026
homepage, both intelligent-assistant packages (RHDHBUGS-3515) and the keycloak auth provider (RHDHBUGS-3678) were excluded because their refs could not be pulled anonymously. The 2.1 index now pins them by digest and `skopeo inspect --no-creds` resolves all four. A fork run on this branch with them enabled passed: https://github.com/gustavolira/rhdh/actions/runs/39851624217 The orchestrator excludes (RHIDP-15961) stay, now noting where the orchestrator backend is covered instead. Same change as on main in redhat-developer#5704. Part of RHIDP-18205. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1 of 5 tasks
…ions Review fixes on dispatch-release-plugin-sanity.sh: - has_harness treated every API error as "no harness", so a rate limit or 5xx on the first branch ended the search and the job exited 0 having checked nothing. Only a 404 now means "no harness"; anything else fails. - A failing branch listing inside the mapfile process substitution was swallowed the same way (verified: against a missing repository the old script printed "nothing to dispatch" and exited 0). The listing now runs in a plain assignment, so set -e catches it. - Each run is reported with its real conclusion instead of success/failure, so a run cancelled by a manual dispatch on the same branch (shared concurrency group) shows as cancelled. Anything but success still fails the job. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
|
The container image build workflow finished with status: |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #5704 +/- ##
==========================================
- Coverage 75.70% 73.04% -2.67%
==========================================
Files 65 61 -4
Lines 708 575 -133
Branches 100 83 -17
==========================================
- Hits 536 420 -116
+ Misses 171 154 -17
Partials 1 1
Continue to review full report in Codecov by Harness.
🚀 New features to boost your workflow:
|
Script (dispatch-release-plugin-sanity.sh): - A failed dispatch no longer aborts the script under set -e and leaves the runs already dispatched unwatched. It is recorded as "dispatch failed", the other runs are still waited on, and the job fails. - The run id is checked before watching, and a failing `gh run view` reports "unknown" instead of ending the job with half a summary. - The run URL comes from the dispatch response's html_url instead of being rebuilt. - RELEASE_COUNT is an env var with its default in one place, validated as a positive integer; the workflow no longer repeats the 2. - Bash 3 compatible (no mapfile or associative arrays), so the unit test runs with /bin/bash on macOS too; the branch listing asks for 100 per page. Tests: unit/dispatch-release-plugin-sanity.test.ts drives the script with a stub gh on PATH: default count, stop at the first branch without the harness, nothing to dispatch, API error vs 404, failed conclusion, failed dispatch, and RELEASE_COUNT validation. Each case was seen failing under a mutation of the behaviour it covers. Docs: the X.Y index tag follows the newest build of the stream (plugin- catalog-index:2.1 currently resolves to the 2.1-111 digest), so RC builds are covered by the schedule; a manual dispatch is for one specific build. The doc now says the job fails unless all runs succeed, "up to two" branches, and that manual and scheduled dispatches on a branch cancel each other. The loki exclude notes its coverage. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
gustavolira
added a commit
to gustavolira/rhdh
that referenced
this pull request
Oct 22, 2026
Same wording as on main in redhat-developer#5704. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Contributor
|
The container image build workflow finished with status: |
Contributor
|
@gustavolira the openshift-ci bot did its random assigning of reviewers and selected me, but I have no prior context for ci in this repo - you may want to tag someone from git history for the changed files |
Member
Author
|
Closing: the release plugin sanity check is moving to rhdh-plugin-export-overlays, where one workflow will check every catalog index and community plugin in the real RHDH image of a release (including RCs). The rhdh plugin sanity checks will be removed once that workflow is in place, so this change is no longer needed. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
plugin-sanity(cluster-free,e2e-cluster-free.yaml) only ran on main's schedule, against thenextcatalog index. Release branches could only be checked with a manualworkflow_dispatch, which needs admin rights on the repository, so RC and z-stream indexes were never checked automatically. For 2.1.0 it had to be run from a fork (RHIDP-13884).Release branches on a schedule. GitHub only fires
scheduleon the default branch. A newrelease-plugin-sanityjob on main's schedule runsdispatch-release-plugin-sanity.sh, which:release-X.Ybranches that carry the harness (it stops at the first branch without it, so 1.10 and earlier are skipped);plugin-catalog-index:X.Y;return_run_detailsreturns the run id) and fails unless all succeed, with each run's conclusion and link in the job summary. Without this, a dispatched run that fails notifies nobody. A failed dispatch or API error fails the job rather than reading as "nothing to check".plugin-catalog-index:X.Yfollows the newest index build of the stream (:2.1currently resolves to the2.1-111digest), so RC and z-stream builds are checked as they land. A manual dispatch withcatalog_index_imageremains for checking one specific build.The job needs
actions: writefor the dispatch, scoped to that job.I chose not to use a matrix on main that checks out the release refs: main's workflow and backend would boot a release index, a combination that is never shipped.
Four stale excludes removed. homepage, both intelligent-assistant packages (RHDHBUGS-3515) and the keycloak auth provider (RHDHBUGS-3678, closed as obsolete) were excluded because their refs could not be pulled anonymously. The
nextand2.1indexes now pin them by digest, andskopeo inspect --no-credsresolves all four. The two orchestrator excludes (RHIDP-15961) stay. Their notes say what covers them today: the orchestrator E2E in rhdh-plugin-export-overlays, which loads the overlay build, not the index artifact. No rhdh job loads them, and no spec exercises the Loki module.docs/e2e-tests/local-e2e-harness.mdgains a plugin-sanity section.unit/dispatch-release-plugin-sanity.test.tscovers the script with a stubghonPATH. Each case was seen failing under a mutation of the behaviour it covers.Which issue(s) does this PR fix
PR acceptance criteria
How to test changes / Special notes to the reviewer
None of this runs on
pull_request. Verified from the fork:plugin-sanitypassed on main againstnext(run; its first attempt hit a transient quay.io blob EOF) and on release-2.1 against2.1with the [release-2.1] fix(e2e): unblock plugin-sanity and drop stale excludes #5703 stub (run).dispatch-release-plugin-sanity.shrun locally againstgustavolira/rhdh: it picked onlyrelease-2.1, dispatched it (run), waited, and exited 1 with the summary line. That run failed on the/jira/apiproxy target, as expected on arelease-2.1without [release-2.1] fix(e2e): unblock plugin-sanity and drop stale excludes #5703. This also shows the failure path.Merge #5703 first, otherwise the first scheduled release run fails on that proxy.
🤖 Generated with Claude Code