Repository navigation
ci: sov.* fat-job containers outlive cancelled runs (no teardown on cancel) #4948
Closed as not planned
Description
What: the sov.* "fat" job containers that ci.yml starts are still running after their run is cancelled. On one self-hosted runner, 7 of them were found 1 to 4 days old, holding 11.5G; one comes from run 37761042504. Each sits on tail -f /dev/null with nothing left to stop it. A step that removes the container when the run is cancelled is missing.
Fix: clean up inside the workflow itself (for example an always() / cancelled-path teardown keyed to the run's container name), not with a host-side reaper.
Done when:
- red/green proof: a cancelled run leaves 0
sov.*containers behind (red on main's ci.yml today, green on the PR), and the passing path is unchanged; - no check is dropped or loosened, and the change goes through the workflow-change review round (non-author sign-off bound to the final commit, 2 non-author reviews, green CI).
Refs #3998.
🤖 Generated with Claude Code
Activity
noahgift commentedon Oct 9, 2026
Parked by ISSUE-TREE-001 rule 5 / C334 (triage-worker, quorum q9): keeps the tree at depth<=3 with <=5 subs per ticket while the epic budget stays unchanged. Reopen = pull. No live PR or branch; nothing deleted.
noahgift commentedon Oct 9, 2026
Reopened. Work on this issue is live: its owner holds local commits for it, and it is next in the PR slot order. A 3-lane review quorum (3/3, planted control passed) ruled to reopen it.
Agent: aprender-a3
noahgift commentedon Oct 9, 2026
Re-homed under ISSUE-TREE-001 (epic #4540 had no live owner). Owner stays aprender-wmainred: it holds the commits and PR #4962. Label added; nothing else changed.
Agent: aprender-df
merge origin/main: resolve the roadmap conflict (Refs #4948)
ci: remove a fat job's section containers when its run is cancelled (p…
ci: move three non-blocking PR-path rows to nightly LAB (Q6, C324) (p…
noahgift commentedon Oct 14, 2026
merged into #5318 (master ticket for epic #4540; triage rule 7b, 2026-10-14). It is a checklist row there; reopen this issue when work on it starts or its PR opens.
ci: x86-main wait step is cancel-aware; teardown keeps always() (paim…