Repository navigation
Conversation
The org's GitHub Actions billing block stops GitHub-hosted jobs from starting, so on ubuntu-latest these checks never ran at all -- every PR on this repo has sat with checks that never leave the queue. gha-ci-runners (the ARC fleet on the OVH cluster) is where the org's actually-running CI already lives. gitleaks drops its sudo install (the runner container has no sudo) and runs the extracted binary from the workspace by path instead; nothing else changes shape.
The self-hosted runner container has no sudo, so this step never ran successfully there. pywr-moea's and pywr-borg's own CI already confirm the Nexsys-Analytics/ci-runners image prebakes libglpk-dev among its other native build dependencies, so the install step is unnecessary rather than needing a no-sudo replacement.
Moves each job's step logic out of ci.yml/gitleaks.yml into
.github/actions/{typecheck,test,test-minimal-install,gitleaks},
parameterised by a working-directory input defaulting to the repo
root. This lets mono's synced-in copy of this repo call the identical
steps locally against its subtree-prefixed path, matching the pattern
already shipped for infrastructure -- no behaviour change to this
repo's own CI, which still resolves working-directory to ".".
Also applies the org's CI hardening convention established on
cluster-api-v2 #91: every job gets timeout-minutes (15 for the
typecheck-shaped job, 30 for test, 10 for the minimal-install and
gitleaks jobs, since self-hosted runner pods have no GitHub-imposed
job limit), and the gitleaks download is checksum-verified and
time-bounded (--max-time 60, sha256sum -c against the pinned v8.21.2
release digest) rather than piped straight from curl into tar.
|
This branch's checks have queued and then auto-cancelled at the 24h GitHub Actions timeout twice now (2026-08-21 and 2026-08-27) rather than failing -- nothing ever picked the jobs up. Can't push a fix myself either way: my access on this repo is currently pull-only (no push, no admin), unlike every other repo in this set. Flagging both -- the runner-group scope and the access level -- rather than guessing at either. |
|
Filed as Nexsys-Analytics/ci-runners#1, assigned to Gio. |
The org's GitHub Actions billing block stops GitHub-hosted jobs from starting, so on
ubuntu-latestthis repo's checks never ran at all. This moves both workflows ontogha-ci-runners, the ARC fleet on the OVH cluster, with the same adaptations as the sibling repos' runner PRs (no sudo on the runner container: gitleaks runs from the workspace by path, shellcheck becomes a static workspace binary).Blocked, not ready: this repo's two runs have sat queued while an idle, listening fleet runner held the matching
gha-ci-runnerslabel and every sibling repo's runs were served -- the signature of thenexsys-ci-runnersrunner group's repository-access list not including pywr-utils. That list is an org-admin setting (Organization Settings -> Actions -> Runner groups -> nexsys-ci-runners -> Repository access), so this PR goes ready once pywr-utils is added and these checks run.