Skip to content

PR Review Agent — failures detected 2026-07-15 #1262

Closed

Description

@github-actions

1. Executive Summary

Status: WARNING
Period: 2026-07-15T03:39:54Z – 2026-07-15T08:32:07Z
Result: 2 of 100 runs failed (2.0%) | 9 cancelled (9.0%) | 19 skipped (19.0%)

Key findings:

  • No log content was returned for either failed run (#56122, #56072) — root cause cannot be confirmed from provided data
  • Each failure is immediately followed by a skip at the same timestamp (failure→skip pairs), suggesting branch-filter or path-filter logic is misfiring on the event that triggers the failure
  • Cancellations cluster in bursts (e.g., #56108–#56110 at 06:02:31–06:02:39Z; #56034/#56036 at 03:42:05–03:42:37Z), consistent with a missing concurrency group on the caller stub, causing rapid push events to race

Action required: Retrieve and inspect full step logs for runs #56122 and #56072; add a concurrency key to pr-review-trigger.yml to prevent burst cancellations.


2. Failure Breakdown

Failure Category Affected Runs Example Error Message
Unknown — logs not returned #56122, #56072 (no log content provided in analysis window)
Concurrency burst / implicit cancellation #56108, #56109, #56110, #56034, #56036, #56047, #56065, #56087, #56124 (cancelled by GitHub runner — no error message)
Skip / filter mismatch (correlated with failures) #56123 (→ #56122), #56073 (→ #56072) (conclusion inferred from identical timestamps)

3. Error Patterns

Pattern A — Failure → Skip co-occurrence

Run #56122 (failure) and #56123 (skip) share timestamp 06:14:27Z. Run #56072 (failure) and #56073 (skip) share 04:17:13Z/04:17:14Z. Both pairs are also followed within seconds by a cancellation (#56124 at 06:14:30Z, #56065 at 04:06:31Z in the prior burst). This pattern is consistent with one of:

  • The reusable workflow at pr-review.yml@pr-review/next exiting non-zero in a step that is conditional on the same predicate that causes the sibling run to be skipped — e.g., a check_suite event fires the caller, the caller job fails, and a concurrent pull_request event fires a second invocation which then short-circuits via a skip guard.
  • A transient infrastructure error in the uses: reusable resolution for @pr-review/next at the moment the tag was in motion (e.g., mid-cut-release.sh tag advance).

Without actual log lines, these remain hypotheses; no confirmed error message is available.

Pattern B — Cancellation bursts

Nine cancellations occur in three identifiable clusters:

  • 03:42:05–03:42:37Z (#56034, #56036)
  • 06:02:31–06:02:39Z (#56108, #56109, #56110)
  • Isolated: #56047, #56065, #56087, #56124

This is characteristic of GitHub's implicit run cancellation when a newer run supersedes an in-flight one and no concurrency group is configured on the caller. The stub (pr-review-trigger.yml) has no concurrency: key, so every new pull_request or check_suite event spawns a new run without cancelling the previous one explicitly — GitHub's UI then shows the earlier runs as "cancelled" when the newer run completes or when resources are reclaimed.


4. Token Scope Analysis

Scopes granted in workflow source (permissions: block):

  • contents: read
  • pull-requests: write
  • checks: read

secrets: inherit passes the calling repo's secrets (including GITHUB_TOKEN and any org secrets) to the reusable.

Assessment based on available data:

Scope Status Notes
contents: read Present Required to checkout scripts at agent_ref
pull-requests: write Present Required to post review comments
checks: read Present Required for check_suite event handling
statuses: read Not declared May be needed if the reusable reads commit statuses (common in CI-gate checks); absence would silently return empty results, not a hard failure
read:org Not declarable via permissions: If the reusable calls any org-level API (team membership, org PRs), this requires a PAT with read:org — GITHUB_TOKEN cannot carry this scope

Recommendations:

  • If gh api /orgs/... calls exist in engine.sh or review-one-pr.sh, ensure a PAT with read:org is available as a named secret and explicitly passed via secrets: rather than relying on secrets: inherit from GITHUB_TOKEN.
  • No scope errors are confirmed from logs — this is a precautionary audit.

5. Recommendations

  1. [CRITICAL — if confirmed] Retrieve step-level logs for #56122 and #56072

    • What: Run gh run view 56122 --log-failed and gh run view 56072 --log-failed in the petry-projects/.github-private repo.
    • Why: The failure root cause is entirely unknown; all other recommendations are blocked on this.
    • Expected impact: Identifies whether failures are transient, a tag-advance race, a script error, or a permissions issue.
    • Urgency: CRITICAL — no remediation is possible without this data.
  2. [HIGH] Add a concurrency group to pr-review-trigger.yml

    • What: Add to the workflow top-level (before jobs:):
      concurrency:
        group: pr-review-${{ github.event.pull_request.number || github.event.check_suite.id || github.run_id }}
        cancel-in-progress: true
    • Why: The current lack of a concurrency group causes burst events (rapid pushes, check suite completions) to spawn redundant runs that GitHub then cancels ad hoc. Explicit cancellation is safer and preserves the most recent run.
    • Expected impact: Eliminates the cancellation bursts; reduces wasted runner minutes; makes the run history easier to audit.
    • Urgency: HIGH
  3. [MEDIUM] Investigate the failure→skip co-occurrence pattern

    • What: After retrieving logs (recommendation 1), check whether the skip guard in pr-review.yml (or a if: condition on the review job) evaluates github.event_name or PR state in a way that races with the failure condition.
    • Why: Simultaneous failure and skip on distinct runs triggered by the same event burst suggests either a shared predicate bug or a tag-move race during cut-release.sh.
    • Expected impact: Prevents correlated failures from being masked by the surrounding skip/cancel noise.
    • Urgency: MEDIUM
  4. [LOW] Add workflow_run or on.workflow_call health annotations

    • What: Consider adding a brief run_name: field to pr-review-trigger.yml (e.g., run_name: "PR Review — ${{ github.event.pull_request.number || github.event_name }}") so individual runs are identifiable without opening logs.
    • Why: The current run list is opaque — all 100 runs have the same display name, making triage harder.
    • Expected impact: Faster triage in future health reports.
    • Urgency: LOW

6. Health Score

Health: 8/10 — Failure rate is low (2%) but root cause is unconfirmed due to missing log content; cancellation bursts are wasteful and should be addressed with a concurrency group.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    automated-reportCreated by automated workflowhealth-checkAutomated health check report

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions