Repository navigation
fix(ci): bound a stalled bun install and let the helm test probe wait for routing - #9409
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub. |
[Medium impact] The PR appears safe to merge; no blocking issues remain. SummaryBounds stalled dependency installs and lets Helm connectivity tests wait for Service routing.
The previous probe-duration finding is addressed. No new actionable issues were found. Reviews (2) · Last reviewed commit: "fix(ci): fit the install retry inside ev..." · Reviewed by Greptile |
There was a problem hiding this comment.
All reported issues were addressed across 4 files
Reply with feedback, questions, or to request a fix.
Turn on auto-fix | Re-trigger cubic
…probe's log from its kept pod
|
@cubic-dev-ai review this PR |
@waleedlatif1 I have started the AI code review. It will take a few minutes to complete. |
There was a problem hiding this comment.
All reported issues were addressed across 4 files
Reply with feedback, questions, or to request a fix.
Turn on auto-fix | Re-trigger cubic
Summary
bun installin the shared setup action now gets 3 minutes per attempt and up to 2 attempts (about 6.5 minutes worst case, inside the shortest caller's 10-minute job timeout). Bun already retries failed requests itself, but a registry connection that stalls without failing left the install waiting until the job timeout cancelled it. That cancelled a required integration shard and skipped a release's promote. Healthy installs finish within 146s cold (p99 94s across ~900 recent install steps). Only a stall (exit 124/137) is retried; any other failure still fails at once. This follows the bounded-attempt pattern other Bun monorepos use (kilocode, oh-my-pi, elizaOS)helm testprobe retries each URL until a 60-second deadline. A pod turns Ready before every node routes its Service to it, so a single-shot probe sent the momenthelm install --waitreturns could be refusedhelm testand adds the realtime and test-pod logs, so a failed probe says which URL refused (--logscannot be used: the hook'shook-succeededdelete policy removes a passing pod before helm reads its log)Type of Change
Testing
timeout(Linux container) against a stubbun: success (1 call), lockfile error (fails at once, 1 call), stall then success (recovers on attempt 2), permanent stall (fails after 2), stall that ignores SIGTERM (killed after the grace period)helm templateand ran the script throughsh -n, then against a stubwget: recovers after refusals; fails at the deadline with exit 1actionlinton helm.yml,helm linton the chartChecklist
test-auditauthoring gate)