Skip to content

[triage] infra_auth — dev-mesh-triage failing: Claude API overloaded (HTTP 529), 2026-07-30 (self-resolved) #853

Description

@github-actions

Triager classification

Label: infra_auth
Action: Escalate to human — do NOT re-kick, do NOT attempt a code fix.


Evidence

One dev-mesh-triage scheduled check run failed today (2026-07-30):

Run Time (UTC) Error
30520243370 06:38 HTTP 529 Claude API overloaded

The Claude Code action terminal result envelope: is_error: true, duration_ms: 35770, num_turns: 12, total_cost_usd: 0.2947491, permission_denials_count: 1.

The postrun step (assert-run-healthy.mjs) logged: 'Claude API overloaded (HTTP 529) — transient; re-run with back-off (attempt 1). Use Re-run failed jobs in the Actions UI; .github/actions/mesh-retry-backoff adds jitter delay.'

The run's own .github/actions/mesh-retry-backoff step (initial jitter) executed but was a no-op on this attempt (attempt 1), so no back-off delay preceded the overloaded call.

Classifier reasoning (src/dev-mesh/classify.js)

Precedence: infra > out-of-scope > flake > real_bug.

extractSignals matches INFRA_RE directly on the literal strings HTTP 529 / api overloaded present in the job log, giving infraError: true. classifyFailure returns infra_auth at the first precedence branch (infraError true -> INFRA_AUTH, auth/infra signature in log), before any out-of-scope/flake/real_bug check runs. This is not a code bug — there is no diff to relate it to; this is the mesh's own scheduled automation (dev-mesh-triage) hitting a transient Anthropic API capacity condition. Same class as the prior precedent issues (#385, #418, #456, #482, #816, #850).

The one permission_denials_count: 1 in the result envelope is unrelated noise from the Triager's own bounded toolset (Read,Grep,Glob,Bash(gh:*)) — not the cause of the is_error (the 529 signature is the direct cause).

Current status: self-resolved

No other scheduled dev-mesh workflow run on 2026-07-30 shows a matching failure — dev-mesh-ci-sweep (06:13), dev-mesh-intake (06:08 and 09:15), dev-mesh-security (09:16), and every other dev-mesh-* run before and after 06:38 completed with conclusion: success. Hermetic CI (ci.yml) and the nightly integration pipeline are both green on main; no open PR (#779, #772, #769, #495) has a failing check. This was an isolated single-run blip, not a cluster.

What a human should do

  1. No action required to unstick anything — surrounding scheduled runs (before and after) are green, so this did not block any other automation.
  2. If HTTP 529s recur/cluster, check the Anthropic status page for an ongoing capacity incident.
  3. Filed for the record per the infra_auth precedent set by triage-sweep: infra_auth -- Claude API 529 overloaded across dev-mesh workflows (2026-06-22 ~00:23-00:40 UTC, self-resolved) #385/infra_auth: dev-mesh-intake failing with HTTP 529 (Claude API overloaded) — 4 runs, 2026-06-22 #418/[triage] infra_auth — dev-mesh-intake failing: Claude API overloaded (HTTP 529) #456/infra_auth: persistent HTTP 529 API overload in dev-mesh-intake (8+ failures, 2026-06-22–23) #482/infra_auth: dev-mesh-triage/backlog/security failing — Claude session rate limit (429), 2026-07-05 (6 runs, self-resolved) #816/[triage] infra_auth — dev-mesh-intake failing: Claude API overloaded (HTTP 529), 2026-07-29 (self-resolved) #850 — a repeat pattern worth tracking, not a one-off worth ignoring.

Filed by the Triager (dev-mesh-triage sweep)

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions