Skip to content

ci: Keep package artifacts for a week instead of a day - #2502

Merged
cgwalters merged 1 commit into
bootc-dev:mainfrom
cgwalters-forge:bot/ci-artifact-retention
Sep 25, 2026
Merged

cgwalters merged 1 commit into
bootc-dev:mainfrom
cgwalters-forge:bot/ci-artifact-retention

Conversation

@cgwalters-bot

Copy link
Copy Markdown
Contributor

The integration test jobs download the packages-<os> RPMs built earlier in
the same workflow run, and those artifacts are uploaded with
retention-days: 1. So re-running a flaked test job on a PR that has sat for
more than a day fails with "Artifact not found for name: packages-", and
the only way out is re-running the whole workflow, package builds included.
This is the "Package expiration" item in the flakes tracker.

This raises the retention to 7 days. The only consumers are the
download-artifact steps in the same run, so nothing depends on the short
expiry. At about 8MB per artifact and recent run volume that's roughly
3.3GB at steady state, which is free for public repositories and small next
to the tmt logs kept for the default 90 days.

Testing: CI-only change, so not exercised beyond actionlint, whose findings
on ci.yml are the same as on main. The real test is re-running a test job
on a PR older than a day.

CI note (2026-09-24): the test-integration (fedora-44, composefs, xfs, grub, bls, unsealed) failure is unrelated to this change: plan-44-shadow-fixup failed with "expected exactly one testbootcgroup in /etc/group" because finalization failed with Failed to open /boot: Host is down (os error 112) after the /boot automount expired, the known flake fixed by cgwalters-forge#6. Failed jobs will be rerun once the run completes.

Related: #2177

The Signed-off-by: Colin Walters <walters@verbum.org> on these commits was added on cgwalters's approval of the review draft: cgwalters-forge#3 (review)

Generated-by: https://github.com/cgwalters/#llms

The integration test jobs consume the RPMs built by the package job
in the same workflow run. With a 1 day retention, re-running a flaked
test job on a PR that has been sitting for more than a day fails with
"Artifact not found for name: packages-<os>", and the only way out is
re-running the whole workflow, including rebuilding the packages.

The tradeoff is storage: each packages-<os> artifact is about 8MB, and
at recent volume (~10-25 runs a day, 5-6 OSes) a week of them is a few
GB at steady state rather than a few hundred MB. That's still small
next to the tmt-log artifacts each run uploads with the default
90 day retention (tens of MB each, dozens per run), and artifact storage
isn't billed for public repositories anyway. We could go longer if
a week turns out not to be enough.

Related: bootc-dev#2177
Generated-by: AI
Signed-off-by: Colin Walters <walters@verbum.org>
@cgwalters
cgwalters enabled auto-merge (rebase) September 25, 2026 19:11
@cgwalters
cgwalters merged commit 8f348ae into bootc-dev:main Sep 25, 2026
54 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants