[Fleet Monitor] petry-projects/google-app-scripts — pr-auto-review.yml

[don-petry]

🟡 WARNING: petry-projects/google-app-scripts / pr-auto-review.yml

Metric	Value
Failure Rate	14.3% (1 / 7 runs)
Status	WARNING
p50 Duration	9s
p95 Duration	11s
Successful Runs	6
Cancelled Runs	0

Workflow: pr-auto-review.yml

Last updated by Fleet Monitor run on 2026-06-05.
Threshold: failure rate > 10% over the monitored window.

[don-petry]

Dev-Lead Implementation Plan

Issue: #356 — [Fleet Monitor] petry-projects/google-app-scripts — pr-auto-review.yml

Root Cause

All 12 failures share the same error: GraphQL: API rate limit already exceeded for user ID 36422719.

The workflow fires on four trigger types simultaneously (workflow_run, check_suite, pull_request, pull_request_review). When multiple events land at once (e.g., multiple check suites completing after a push), concurrent runs all make GitHub REST + GraphQL API calls at the same time, exhausting the PAT's 5,000 req/hr rate limit and causing exit code 1.

Scope

Add a concurrency group to .github/workflows/pr-auto-review.yml so at most one readiness-check run executes per commit SHA, cancelling stale in-progress runs when a newer trigger fires. This serializes API calls and eliminates the rate-limit race condition.

Implementation Checklist

• Add concurrency: block to .github/workflows/pr-auto-review.yml keyed on the head SHA (consistent across all four event types)

Tests to Write

• Exception: Pure YAML configuration change with no observable JS/application behavior to unit-test. No new test files required. Validation is by running npm run check (lint/format) and npm test (existing test suite) to confirm nothing regressed.

Test Command

npm test

Notes

• The concurrency.group expression uses || to extract the head SHA across all four event types:
  github.event.pull_request.head.sha || github.event.workflow_run.head_sha || github.event.check_suite.head_sha
• With cancel-in-progress: true, a new trigger for the same commit cancels any in-progress run — correct behaviour for a readiness gate (we only care about the latest state).
• Per workflow AGENTS note, only allowed changes are: version tag in uses:, workflow_run.workflows names, and non-prohibited additions. concurrency: is not a trigger event type or job-level permissions: block, so it is permitted.

[don-petry]

Dev-Lead: Implementation Complete

Plan Execution

• Identified root cause: all 12 failures share GraphQL: API rate limit already exceeded for user ID 36422719
• Confirmed no v3 of the reusable workflow exists (only v1/v2 tags), so version bump is not the fix
• Added concurrency: block to .github/workflows/pr-auto-review.yml keyed on commit head SHA

Root Cause

Multiple triggers (workflow_run + check_suite) fire in rapid succession for the same PR/commit. All concurrent runs simultaneously make GraphQL API calls to check review threads, exhausting the PAT's 5,000 req/hr rate limit and causing exit code 1.

Fix

Added a concurrency group that serializes runs per commit SHA and cancels stale in-progress runs when a newer trigger fires:

concurrency:
  group: pr-auto-review-${{ github.event.pull_request.head.sha || github.event.workflow_run.head_sha || github.event.check_suite.head_sha }}
  cancel-in-progress: true

• The SHA expression resolves correctly for all four event types the workflow handles
• cancel-in-progress: true ensures only the latest trigger proceeds, making stale readiness checks obsolete
• This does not change trigger event types or the job-level permissions: block (per AGENTS note constraints)

Test Results

Test Suites: 12 passed, 12 total
Tests:       301 passed, 301 total
Snapshots:   0 total
Time:        4.192 s
Ran all test suites.

npm run check passed (prettier + eslint, 0 errors, 22 pre-existing warnings unchanged).

Files Changed

• .github/workflows/pr-auto-review.yml: Added concurrency: group after permissions: {}, before jobs:

Notes

TDD exception applied: this is a pure YAML workflow configuration change with no JS application behavior to unit-test. Correctness is validated by the existing test suite passing and by the expected GitHub Actions runtime behavior of the concurrency key.

[don-petry]

Dev-Lead Implementation Plan

Issue: #356 — [Fleet Monitor] petry-projects/google-app-scripts — pr-auto-review.yml

Root Cause

All 6 hard failures (2026-06-01) share the same error:

GraphQL: API rate limit already exceeded for user ID 36422719.
##[error]Process completed with exit code 1.

The reusable workflow (pr-auto-review-reusable.yml@v2) step 4 (unresolved review threads check) runs:

UNRESOLVED=$(gh api graphql ... --jq "...")

Under set -euo pipefail, when gh api graphql exits non-zero due to a rate limit, the variable assignment aborts the step — causing a workflow failure instead of a graceful skip.

Scope

Two files change across two repos:

1. petry-projects/.github — fix the reusable workflow to handle GraphQL rate-limit errors gracefully, then cut v3 tag
2. petry-projects/google-app-scripts — bump the caller stub from @v2 → @v3

Implementation Checklist

• Patch petry-projects/.github/.github/workflows/pr-auto-review-reusable.yml: wrap the GraphQL call in an if ! guard so a non-zero exit (rate limit, transient error) sets ready=false and exits 0 instead of failing the workflow
• Commit the patch to petry-projects/.github main
• Create v3 tag on that commit
• Update .github/workflows/pr-auto-review.yml in this repo: @v2 → @v3

Tests to Write

This is a pure CI workflow fix with no application code or testable logic. No Jest tests exist or are applicable.

TDD exception: Workflow YAML changes cannot be unit-tested with Jest. The only verification is running the workflow in CI or inspecting the changed logic.

Test Command

npm test (no changes to application logic; will confirm no regressions in existing suite)

[don-petry]

✅ Auto-resolved by Fleet Monitor. This workflow's failure rate has been below the 10% threshold (or the workflow was removed) since 2026-06-05 (40 days) — the monitor stopped flagging it. Closing. It will reopen automatically if the workflow crosses the threshold again.