[GH-ISSUE #4069] Nightly build/test failure - 2026-09-12 #1035

New issue

Open

opened 2026-10-29 21:22:11 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-29 21:22:11 +00:00

Owner

Originally created by @github-actions[bot] on GitHub (Sep 12, 2026).
Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/4069

Nightly Failure Report

Run: https://github.com/kvcache-ai/Mooncake/actions/runs/34620633139
Branch: refs/heads/main
Timestamp: 2026-09-12T16:11:14.165Z

Please investigate the failed jobs in the workflow run linked above.

Originally created by @github-actions[bot] on GitHub (Sep 12, 2026). Original GitHub issue: https://github.com/kvcache-ai/Mooncake/issues/4069 ## Nightly Failure Report **Run**: https://github.com/kvcache-ai/Mooncake/actions/runs/34620633139 **Branch**: refs/heads/main **Timestamp**: 2026-09-12T16:11:14.165Z Please investigate the failed jobs in the workflow run linked above.

bsakaguchi added the

nightly-failure

label

2026-10-29 21:22:11 +00:00

bsakaguchi commented

2026-10-29 21:22:12 +00:00

Author

Owner

@he-yufeng commented on GitHub (Sep 13, 2026):

Broke the three failures down against last night's run:

publish-testpypi: first occurrence. The wheel uploads to 100% and TestPyPI answers 400 after the body, with --skip-existing already in place, so this is not the usual "file already exists" case. One-off TestPyPI-side rejection is the most likely read; if tonight's run 400s the same way, the nightly version stamp's metadata is the next suspect.
tone-sglang-integration / test-sglang-integration: second night in a row, and it fails before any test runs — the job cannot fetch the Build & Test artifact for the SHA ("Failed to fetch artifacts for workflow run ... after $max_attempts attempts"). Recurring, so this is a real breakage in the artifact handoff, not a flake; worth a look at whatever changed in the upstream run naming/retention.
build-musa (09-11 only, gone tonight): transient.

No code regression in either night that I can see from the logs.

@he-yufeng commented on GitHub (Sep 13, 2026): Broke the three failures down against last night's run: - `publish-testpypi`: first occurrence. The wheel uploads to 100% and TestPyPI answers 400 after the body, with `--skip-existing` already in place, so this is not the usual "file already exists" case. One-off TestPyPI-side rejection is the most likely read; if tonight's run 400s the same way, the nightly version stamp's metadata is the next suspect. - `tone-sglang-integration / test-sglang-integration`: second night in a row, and it fails before any test runs — the job cannot fetch the Build & Test artifact for the SHA ("Failed to fetch artifacts for workflow run ... after $max_attempts attempts"). Recurring, so this is a real breakage in the artifact handoff, not a flake; worth a look at whatever changed in the upstream run naming/retention. - `build-musa` (09-11 only, gone tonight): transient. No code regression in either night that I can see from the logs.

bsakaguchi referenced this issue

2026-10-29 21:22:45 +00:00

[GH-ISSUE #4337] [RFC][Store] Offline timestamped master RPC trace replay #1088