Skip to content

CI: exclude bad Phoenix L40S node atl1-1-03-007-31-0 (uncorrectable ECC) - #1750

Merged
sbryngelson merged 1 commit into
masterfrom
ci/exclude-phoenix-007-31-0
Aug 23, 2026
Merged

CI: exclude bad Phoenix L40S node atl1-1-03-007-31-0 (uncorrectable ECC)#1750
sbryngelson merged 1 commit into
masterfrom
ci/exclude-phoenix-007-31-0

Conversation

@sbryngelson

Copy link
Copy Markdown
Member

What

Adds atl1-1-03-007-31-0 to the Phoenix GPU --exclude list. Net exclude is now atl1-1-03-007-29-0,atl1-1-03-007-31-0.

Why

A second L40S node in the same chassis as the already-excluded 007-29-0 (both atl1-1-03-007-*) is failing every GPU job with the same hardware fault:

Accelerator Fatal Error: call to cuCtxCreate returned error 214
  (CUDA_ERROR_ECC_UNCORRECTABLE): uncorrectable ECC error encountered

Confirmed across a CI sweep — 007-31-0 is the ECC source in the most recent master run (32601124944, 105 ECC hits) and a PR run, exactly as 007-29-0 was in the prior batch. Reruns don't help: SLURM has no GPU-health awareness and keeps rescheduling onto the same bad node.

Note: the PMIX ERROR: PMIX_ERR_NO_PERMISSIONS ... dstore_base.c lines in these logs are benign (they appear on passing syscheck/pre_process steps too, from the hpcx/2.19-cuda PMIx dstore); they are not the failure cause. The failure is the ECC fault above.

Both bad GPUs are being reported to PACE for repair/drain; this exclude is the CI unblock in the meantime.

Copilot AI lite review requested due to automatic review settings August 23, 2026 02:22
@sbryngelson
sbryngelson enabled auto-merge (squash) August 23, 2026 02:22
@sbryngelson
sbryngelson disabled auto-merge August 23, 2026 02:27
@sbryngelson
sbryngelson merged commit e556597 into master Aug 23, 2026
28 checks passed
@sbryngelson
sbryngelson deleted the ci/exclude-phoenix-007-31-0 branch August 23, 2026 02:27

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR updates the Phoenix GPU CI SLURM submission script to avoid scheduling jobs onto a known-bad L40S node that is consistently failing GPU runs with uncorrectable ECC errors, helping keep the GPU CI lane reliable until the hardware is drained/repaired.

Changes:

  • Extend the Phoenix GPU #SBATCH --exclude list to include atl1-1-03-007-31-0.
  • Keep the existing exclusion of atl1-1-03-007-29-0, resulting in a two-node exclusion list for Phoenix GPU jobs.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

@codecov

codecov Bot commented Aug 23, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 61.61%. Comparing base (d2ed786) to head (8391edd).
⚠️ Report is 3 commits behind head on master.

Additional details and impacted files
@@           Coverage Diff           @@
##           master    #1750   +/-   ##
=======================================
  Coverage   61.61%   61.61%           
=======================================
  Files          84       84           
  Lines       21520    21520           
  Branches     3180     3180           
=======================================
  Hits        13259    13259           
  Misses       6082     6082           
  Partials     2179     2179           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

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

Labels

None yet

Development

Successfully merging this pull request may close these issues.

2 participants