Skip to content

CI: remove Retry Probe fixture; fix canary dry_run flag that was always live - #490

Merged
Szer merged 1 commit into
mainfrom
remove-retry-probe
Sep 14, 2026
Merged

Szer merged 1 commit into
mainfrom
remove-retry-probe

Conversation

@Szer

@Szer Szer commented Sep 14, 2026

Copy link
Copy Markdown
Owner

Follow-up to #489.

Removes

Fixes

  • workflow-canary.yml: DRY_RUN: ${{ inputs.dry_run == 'true' }} compared a boolean input to a string. GitHub coerces mismatched types to numbers (true → 1, 'true' → NaN), so the expression was always false and every "dry run" dispatch sent real PagerDuty events. That is how a Retry Probe (temporary): stalled — no run in 26h warning reached the owner at 10:25Z today while I was testing; it was resolved via the Events API from the throwaway branch. Now inputs.dry_run == true. Verified on branch canary-probe-test: the run prints [dry-run] would trigger … lines and makes no HTTP calls.

Verified before removal (all on main, run ids in #489 comments)

  • recover mode: attempt 1 leg cancelled by timeout → Agent Retry fired 10 s later via rerun-failed-jobs → attempt 2 re-ran only leg + finalize, both green → second Agent Retry firing skipped by the run_attempt == 1 guard.
  • timeout-both mode: exactly one retry, run ends cancelled on attempt 2, no attempt 3.
  • canary (real dry-run from the throwaway branch): the timeout-both run is classified failed with conclusion timed_out; the recovered run logs the needed attempt 2 notice.

🤖 Generated with Claude Code

https://claude.ai/code/session_01V4AMHh8hMfZh53Tnj3TSSn

…ys live

The temporary _retry-probe.yml from #489 has done its job (see #489
comments): attempt 1 killed by timeout-minutes, Agent Retry fired within
10 s and called rerun-failed-jobs, attempt 2 re-ran only the dead leg,
the completion of attempt 2 was ignored by the run_attempt guard, and in
timeout-both mode exactly one retry happened (no attempt 3).

Canary: `inputs.dry_run == 'true'` compares a boolean input to a string;
GitHub coerces both to numbers (1 vs NaN) so it was always false and
every dry-run dispatch sent real PagerDuty events — which is how a
"Retry Probe (temporary): stalled" warning reached the owner at 10:25Z
today (resolved manually). `inputs.dry_run == true` fixes it; verified
on branch canary-probe-test run 34833513756's successor: "[dry-run]
would trigger" lines, no HTTP calls.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01V4AMHh8hMfZh53Tnj3TSSn
@Szer
Szer merged commit 0d51a95 into main Sep 14, 2026
8 checks passed
@Szer
Szer deleted the remove-retry-probe branch September 14, 2026 10:32
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