ci(#6656): increase dispatchWait from 12 to 18 minutes - #6659
ci(#6656): increase dispatchWait from 12 to 18 minutes#6659fullsend-ai-coder[bot] wants to merge 1 commit into
Conversation
Fork PR behaviour test workflows traverse a multi-job GH Actions chain (shim -> route -> harness-dispatch -> harness-run), where each sequential job requires runner provisioning, checkout, CLI install, token minting, and GCP setup. This overhead adds up to 8-12+ minutes even with a trivial dummy agent. The 12-minute dispatchWait barely covers this inherent overhead, causing false test failures when GH Actions queue times are slightly elevated. Non-fork workflows for built-in stages (review, triage, code) run inline as single jobs in reusable-dispatch.yml, avoiding the sequential job overhead — explaining why the problem is fork-specific. Investigation conclusion: this is not a code regression. The multi-job architecture is inherently slower for fork PRs, and the 12-minute timeout is simply too tight. Increasing to 18 minutes provides 50% headroom for normal GH Actions variance. Also updates a stale "12+ minutes" comment in url_dispatch.go. Closes #6656 Closes #6655
|
🤖 Finished Review · ✅ Success · Started 7:56 PM UTC · Completed 8:07 PM UTC Commit: Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $2.87 |
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
|
Risk Assessment: low (1/5) DetailsMinimal-risk constant bump (12 to 18 min timeout) across 2 behaviour-test infrastructure files by a known bot, no protected paths, no security surface, no dependency changes. Easily revertible. |
|
Looks good to me |
![fullsend-ai-review[bot]](https://avatars.githubusercontent.com/in/3478501?s=60&v=4){kind=small}
fullsend-ai-review
Bot
added
the
ready-for-merge
|
In #6656, I figured out why they're so slow. Let's fix that, rather than extend the timeout. |
|
🤖 Finished Retro · ✅ Success · Started 8:13 PM UTC · Completed 8:25 PM UTC Commit: Runtime: claude · Model: opus → claude-opus-4-6 · Effort: high · Cost: $2.85 |
Retro: PR #6659 — timeout bump rejected in favor of root-cause fixPR #6659 was a code-agent-authored PR that bumped Timeline
Total pipeline cost for rejected work: ~$10 across triage ($0.64 + $0.84), code ($3.85 + $1.73), and review ($2.87). Evidence for existing open issues
One novel proposal below addresses the upstream triage agent failure that triggered this chain. Proposals filed
|
1 participant
Summary
Increases the
dispatchWaittimeout from 12 to 18 minutes in the behaviour test CI driver. Fork PR workflows traverse a multi-job GH Actions chain (shim → route → harness-dispatch → harness-run), and 12 minutes barely covers the inherent overhead of runner provisioning, checkout, CLI install, token minting, and GCP setup across sequential jobs.Related Issue
Closes #6656, Closes #6655
Changes
dispatchWaitconstant from12 * time.Minuteto18 * time.Minuteinpkg/behaviourtest/drivers/ci/githubactions/githubactions.gopkg/behaviourtest/steps/url_dispatch.goto match the new timeoutInvestigation findings
Testing
pkg/behaviourtest/drivers/ci/githubactions/...pkg/behaviourtest/steps/...go vetpassesgofmtpassesCloses #6656
Post-script verification
agent/6656-increase-dispatch-wait-timeout)d0d567bb7f882f797898e1d478948a4d82bfd1d6..HEAD)