Skip to content

publish: widen the origin-verify window to 6 min (release propagation lag) - #14

Merged
coretl merged 2 commits into
mainfrom
fix/verify-propagation-window
Jun 25, 2026
Merged

coretl merged 2 commits into
mainfrom
fix/verify-propagation-window

Conversation

@coretl

@coretl coretl commented Jun 25, 2026

Copy link
Copy Markdown
Contributor

What

Widen publish.yml's origin-Verify poll from 60s → 360s (and slow the interval 1s→5s).

Why

Follow-up to #13 (the tag trampoline). On the first real release after merging #13, the trampoline worked — v0.10.0's workflow_dispatch deploy re-served and the site went live correctly — but the dispatched run's Verify step went red as a false negative.

A release deploys twice on the same SHA in quick succession:

time deploy event result
14:48:50 merge-to-main (inline) push served (10 entries)
14:50:47 tag trampoline workflow_dispatch re-served (11 entries) — but slowly

The dispatch (the newest/active deployment) won, but its origin propagation lagged behind the first deploy and only landed at 14:54:56 — ~4 min after the deployment record. Verify gave up at its 60s mark (14:52:26), so the run failed even though the site was correct moments later.

The loop still exits the instant the origin matches, so a fast / first-of-SHA deploy (main, PRs) isn't slowed at all — only a genuinely stuck origin now waits the full window before red-checking. 360s gives ~50% margin over the observed ~4 min.

Test

Next release (merge → tag): the dispatched publish.yml run should now stay green — Verify keeps polling past propagation and matches once the dispatch artifact is live.

🤖 Generated with Claude Code

… lag)

The tag trampoline (deploy via workflow_dispatch) works — v0.10.0's dispatch
deploy did re-serve — but its Verify step red-flagged a false negative. A release
deploys TWICE on the same SHA in quick succession: the merge-to-default-branch
push (inline) then the tag trampoline (dispatch). The second (dispatch) re-serve
propagates to the origin behind the first, and on 2026-06-25 took ~4 min, past
the 60s verify window — so the run failed even though the site went live correctly
moments later.

Widen the verify poll from 60s to 360s (and slow the interval 1s→5s). The loop
still exits the instant the origin matches, so a fast/first-of-SHA deploy isn't
slowed; only a genuinely stuck origin now waits the full window before going red.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@coretl
coretl merged commit d682290 into main Jun 25, 2026
3 checks passed

This branch was previously deployed

1 inactive deployment
github-pages — a833fc2f Deployed Jun 25, 2026 by coretl via publish / publish #56
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.

1 participant