Repository navigation
Hang and timeout while running workflow #28868
Description
I have a workflow that is hanging. Relevent snippit from logs:
2026-04-28T06:07:28.8566496Z The changelog looks correct. Now let me publish it to the wiki:
2026-04-28T06:07:28.8566894Z
2026-04-28T06:07:29.3674738Z ● Final check before publish (shell)
2026-04-28T06:07:29.3675417Z │ python3 -c "
2026-04-28T06:07:29.3676333Z │ with open('/tmp/gh-aw/agent/new_changelog.md') as f:
2026-04-28T06:07:29.3677047Z │ content = f.read()
2026-04-28T06:07:29.3677559Z │ print(len(content), 'chars')
2026-04-28T06:07:29.3678154Z │ print(content.count('\n'), 'lines')…
2026-04-28T06:07:29.3678556Z └ 4 lines...
2026-04-28T06:07:29.3678694Z
2026-04-28T06:07:34.6962383Z ● Read new_changelog.md
2026-04-28T06:07:34.6962869Z │ /tmp/gh-aw/agent/new_changelog.md
2026-04-28T06:07:34.6963286Z └ L490:530 (41 lines read)
2026-04-28T06:07:34.6963488Z
2026-04-28T06:17:41.1843543Z The changelog looks correct. Let me now publish it:
2026-04-28T06:17:41.1844296Z
2026-04-28T06:17:41.7038449Z ● Final size check (shell)
2026-04-28T06:17:41.7039011Z │ python3 -c "
2026-04-28T06:17:41.7039618Z │ with open('/tmp/gh-aw/agent/new_changelog.md') as f:
2026-04-28T06:17:41.7040318Z │ content = f.read()
2026-04-28T06:17:41.7040686Z │ # write to stdout for the tool
2026-04-28T06:17:41.7041015Z │ import sys…
2026-04-28T06:17:41.7041258Z └ 4 lines...
2026-04-28T06:17:41.7041391Z
2026-04-28T06:17:45.2672479Z ● Read new_changelog.md
2026-04-28T06:17:45.2673322Z │ /tmp/gh-aw/agent/new_changelog.md
2026-04-28T06:17:45.2674022Z └ L838:845 (8 lines read)
2026-04-28T06:17:45.2674439Z
2026-04-28T06:43:22.5405093Z ##[error]The action 'Execute GitHub Copilot CLI' has timed out after 45 minutes.
2026-04-28T06:43:22.5447543Z ##[group]Run node "${RUNNER_TEMP}/gh-aw/actions/detect_copilot_errors.cjs"
2026-04-28T06:43:22.5448095Z �[36;1mnode "${RUNNER_TEMP}/gh-aw/actions/detect_copilot_errors.cjs"�[0m
2026-04-28T06:43:22.5468765Z shell: /usr/bin/bash -e {0}
You can see there is a 25 minute hang after reading a file and then logs continue after the timeout.
Workflow run: https://github.com/microsoft/aspire/actions/runs/25036515907
Activity
JamesNK commentedon Apr 28, 2026
Again: https://github.com/microsoft/aspire/actions/runs/25038611298/job/73336099831. Big hang when content is ready but before output is published.
JamesNK commentedon Apr 28, 2026
I think I've fixed this by changing the incoming body to the output job to be a file reference.
Could the large delay be caused by reading the body content into an LLM and then passing them to the output job? A small file reference would skip that overhead.
pelikhan commentedon Apr 28, 2026
You should ask this question using an agent and our debugging prompt so that it can correlate the MCP, firewall logs to determine the root cause.
pelikhan commentedon Apr 30, 2026
Most likely feeding a massive file into the agent led to catastrophic context explosion and latency from the LLM provider. I don't think we can do ouch here.
Update community contributions in README
chore: Update Community Contributions section with latest attributions
[community] Update community contributions in README
[community] Update community contributions in README
[community] Update community contributions with Tier 3 attributions
[community] Update community contributions in README
[community] Update community contributions with Tier 3 findings