[agent-job-health] Agent Job Health Monitor - 2026-08-28: fleet improved, 1 novel Windows Runner failure, 3 schedule blind spots #56744
Replies: 1 comment
|
Caveman say: build good. Test done. Discussion nice. Me leave mark here. Warning Firewall blocked 6 domainsThe following domains were blocked by the firewall during workflow execution:
To allow these domains, add them to the network:
allowed:
- defaults
- "accounts.google.com"
- "android.clients.google.com"
- "clients2.google.com"
- "contentautofill.googleapis.com"
- "www.google.com"
- "www.gstatic.com"See Network Configuration for more information.
|
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Summary
Note
Fleet health improved vs. the 2026-08-23 baseline: run-weighted agent-job failure rate dropped from 9.22% → 7.41%, and workflow-weighted mean failure rate dropped from 16.8% → 8.84%. No >10-point regression in either direction.
Warning
Windows Runner Integration Test fails 100% of its runs (3/3) at the
Setup Scriptsstep, and is not covered by any open issue — flagged as Novel below.Failure Rate by Step
Tracked Failures (existing open issues)
22 of 26 agent-job failures are already covered by an existing "[aw] X failed" auto-filed issue or a broader tracked cluster.
Novel Failure Clusters
2 novel clusters, 4 runs, 2 workflows — neither meets the issue-creation threshold alone
1. Windows Runner Integration Test —
Setup Scriptsstep, 3/3 runs failed (100%)Setup Scriptsstep failing consistently on the Windows runner, before any Copilot/Claude/Codex CLI step executes — likely an environment/toolchain provisioning issue specific to the Windows runner image, not an engine-side failure.2. Multi-Device Docs Tester —
Install and build docsstep, 1/1 run failedBuild documentationfailure, despite similar naming).Neither cluster independently triggers a new issue per the issue-creation criteria. Recommend filing if Windows Runner Integration Test's 100%-failure streak continues past today.
Schedule Heartbeat
3 schedule-triggered workflows have gone silent well past their expected cadence, with no existing tracking issue:
total_count:1, emptyworkflow_runs[])3 other workflows (Daily Hippo Learn, Smoke CI, GEO Optimizer Daily Audit) also show no recent runs but are
disabled_manually— explained gaps, not blind spots.One issue is being filed to cover all three genuine blind spots.
Per-Workflow Breakdown
166 distinct workflows ran in the window; 16 had at least one agent-job failure
Recommendations
Setup Scriptsfailure — 100% failure rate, no tracking issue, likely a Windows-runner provisioning regression.total_count:1, emptyworkflow_runs[]) — may indicate these daily workflows are silently stuck rather than just quiet.References:
Warning
Firewall blocked 2 domains
The following domains were blocked by the firewall during workflow execution:
api.anthropic.comapi.github.com[!TIP]
api.github.comis blocked because GitHub API access uses the built-in GitHub tools by default. Instead of addingapi.github.comtonetwork.allowed, usetools.github.mode: gh-proxyfor direct pre-authenticated GitHub CLI access without requiring network access toapi.github.com:See GitHub Tools for more information on
gh-proxymode.To allow these domains, add them to the
network.allowedlist in your workflow frontmatter:See Network Configuration for more information.
All reactions