"Running Copilot cloud agent" workflow fails at "Processing Request (Linux)" step (exit code 1)

[plengauer]

Summary

The Running Copilot cloud agent workflow run #376 failed in the copilot job at the "Processing Request (Linux)" step, exiting with code 1.

Details

• Run URL: https://github.com/plengauer/Thoth/actions/runs/30767625849/job/91549061163
• Branch: copilot/fix-job-level-kill-switch
• Commit: da96822e62181766af68cccb8ef4f1a2f700c6e2
• Job: copilot
• Failed step: Processing Request (Linux) (step fix stderr redirect when interactive #28)
• Duration of failed step: ~6s (21:28:26Z → 21:28:32Z)
• Annotation: Process completed with exit code 1.
• All prior setup steps (MCP server startup, git-proxy download, firewall CA generation) completed successfully; the failure occurs specifically when the Copilot cloud agent begins processing the request.

Note on log detail

I was not able to retrieve the full step-by-step log output for this job — the GitHub Actions log download came back too large to fetch in full, and an unauthenticated fallback request was rate-limited. This issue is filed based on the job's annotation and step summary only. If more detail is needed, the full logs can be viewed directly in the run (link above) or downloaded via gh run view --log --job 91549061163.

Assumption

I'm assuming this failure is worth tracking as a Thoth-side issue even though the failing step is inside GitHub's own Copilot coding agent tooling (not code owned by this repo) — since it's blocking the copilot/fix-job-level-kill-switch branch. Let me know if this should instead be reported upstream to GitHub Copilot.

[github-actions]

@plengauer Thanks for filing this!

A few follow-up questions to help determine if this is a Thoth-side issue:

1. Is this reproducible? Has the copilot/fix-job-level-kill-switch branch's CI run failed consistently, or was this a one-time fluke?
2. Full logs: Can you share the relevant portion of the "Processing Request (Linux)" step output from the run? (gh run view --log --job 91549061163 as mentioned in the issue). Even the last ~50 lines would help narrow down the root cause.
3. Scope: Given your assumption note — do you believe the failure is in Thoth's instrumentation/wrapper code around the Copilot agent step, or purely inside GitHub's Copilot agent infrastructure itself? If the latter, this may be better tracked upstream.

If it turns out the exit code 1 is triggered by Thoth's instrumentation of the job step (e.g., a kill-switch or span teardown issue), that would make it clearly Thoth's to fix.

Generated by Autotriage for #3888 · sonnet46 14.8 AIC · ⌖ 4.98 AIC · ⊞ 5.2K · ◷

[plengauer]

1. it fails consistently - but i have disabled the instrumnetation for now, so on newer runs it doesnt anymore
2. logs:

RUNNER_TEMP=/home/runner/work/_temp
::stop-commands::***
Starting git-proxy...
git-proxy logs: /home/runner/work/_temp/git-proxy.log
Stopping git-proxy (pid=402803)...
::***::
git-proxy exited before listening on port 26831
::***::
rm: cannot remove '/tmp/opentelemetry_shell.398837.sdk.pipe.dmD02tiuFr': No such file or directory
Error: Process completed with exit code 1.

3. i think its an instrumenation problem, because as soon as i disabled instrumentation, it works again