Add retrying OSDC setup composite action and use it across CI workflows#932
Add retrying OSDC setup composite action and use it across CI workflows#932jeanschmidt wants to merge 1 commit into
Conversation
tofu plan — meta-prod-aws-ue2✅ Plan succeeded · commit Plan output |
tofu plan — meta-prod-aws-uw1✅ Plan succeeded · commit Plan output |
tofu plan — meta-prod-aws-ue1✅ Plan succeeded · commit Plan output |
tofu plan — lf-prod-aws-ue2✅ Plan succeeded · commit Plan output |
|
@claude Do a sanity check to confirm if all the steps we need to setup OSDC are now added to the new composite GHA |
|
Claude finished @huydhn's task in 1m 6s —— View job Sanity check: OSDC composite setup action
Verdict: ✅ Yes — all OSDC setup steps are faithfully captured in Parity checkThe original inline setup (identical across
One correct subtlety worth calling out: the original Retry coverageEvery setup step is wrapped in the retry-by-duplication pattern (
The QEMU step ( Intentional exclusions (correct)
Notes / nits
Bottom line: the composite action is a complete and correct extraction of the OSDC toolchain setup. Nothing is missing. |
Stack from ghstack (oldest at bottom):
Impact: CI only — OSDC deploy, drain, undrain, and pre-merge workflows
Risk: low
What
Extracts the repeated "install just + mise +
uv sync" setup into a new local composite action (.github/actions/osdc-setup) and swaps the four OSDC workflows over to it. Each setup step (plus the QEMU setup in_osdc-deploy.yml) is now retried on transient failure via retry-by-duplication.Why
The toolchain install steps were copy-pasted across
_osdc-deploy.yml,osdc-drain.yml, andosdc-undrain.yml, and occasionally failed on transient network hiccups — flaking deploy and merge-queue runs. Consolidating into one action removes the duplication and adds retries in a single place.The retry is done by duplicating each step (
continue-on-erroron the first attempt + a copy gated on its failure) rather than pulling in a third-party retry action, so no unpinnable code runs on the deploy / merge-queue path.actions/checkoutandconfigure-aws-credentialsare left alone since they retry internally.Notes
The SKILL.md doc update also corrects a stale reference: the
_osdc-deploy.ymlskip input isskip_lint_test, notskip-checks.Signed-off-by: Jean Schmidt contato@jschmidt.me