Skip to content

restore: recover GPU host metrics, infra alerting, and Kimi retry clobbered by #111's stale merge - #135

Merged
Thangnguyenvn98 merged 1 commit into
mainfrom
restore/pr-111-collateral
Sep 3, 2026
Merged

Thangnguyenvn98 merged 1 commit into
mainfrom
restore/pr-111-collateral

Conversation

@Thangnguyenvn98

Copy link
Copy Markdown
Collaborator

What happened

PR #111 (builds: fix OTel blocked status and restore missing group columns) was merged from a branch cut from an old main (before #113 landed) and never rebased. Its diff deleted or reverted ~8.5k lines across 77 files of unrelated landed work, which is why ci.vllm.ai/gpu lost the CPU/RAM/Disk columns and drill-down.

What this restores (state at 7202e78^, includes #122)

Also removes the stale pre-#113 scripts/gpu-reporter.py/.service/.timer duplicates the old branch re-added (nothing references them).

What is deliberately kept

Verification

  • pytest alerting: 199 passed · deploy/aws: 23 · migrations: 27 · scripts/gpu-reporter: 29
  • ruff clean; mypy clean except 5 errors that reproduce identically on pristine 7202e78^ (local venv: missing boto3, stricter mypy)
  • npm test 120/120, tsc clean, eslint clean, next build success

cc @khluu — your #111 builds fix is preserved; if it needs follow-ups please branch fresh from main rather than re-merging the old branch.

Signed-off-by: Thang Nguyen thang.nguyen@inferact.ai

…bbered by #111

PR #111 was merged from a stale branch predating #113-#122, so its diff
deleted or reverted ~8.5k lines of unrelated landed work:

- GPU host-metrics feature (#113, #117, #118, #121): gpu-host-table,
  gpu-host-data, gpu-host-view, gpu-report, migration 0018, k8s reporter
  scripts, /api/cron/retention
- Infra alerting module (#119): alerting/infra.py, migration 0019,
  alerts-infra, infra-alerts, /api/alerts/infra, systemd units
- Kimi transient-request retry logic (#120) in alerting/kimi.py
- Per-path Slack channel routing (#112) in src/lib/slack.ts
- Related docs, tests, CI workflow, .env.local.example, vercel.json cron

This restores those files to their state at 7202e78^ (which includes
#122) and removes the stale pre-#113 scripts/gpu-reporter.py/.service/
.timer duplicates the old branch re-added.

Kevin's intended #111 work is kept intact: otel-ci.ts blocked->passed
mapping, buildkite-build-jobs.ts REST roster, and the builds groups/jobs
routes. #110, #123, #124 are preserved; #124's deadline-bounded timeouts
are re-applied on top of the restored retry-aware kimi.py.

Signed-off-by: Thang Nguyen <thang.nguyen@inferact.ai>
@vercel

vercel Bot commented Sep 3, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
vllm-ci-dashboard Ready Ready Preview Sep 3, 2026 12:50pm UTC

Request Review

@Thangnguyenvn98
Thangnguyenvn98 merged commit 6a56883 into main Sep 3, 2026
3 of 4 checks passed
khluu added a commit that referenced this pull request Sep 4, 2026
Resolve conflicts with the GPU host-health restructure (#135):
- gpu-dashboard.tsx: keep upstream's GpuHostTable/buildHostRows and pass
  the /api/gpu/agents map through as an optional prop instead of the old
  inline Host Summary table.
- gpu-host-table.tsx: optional agents prop renders 'queue . job' with a
  Buildkite link under the host name.
- poll-metrics: keep agent sampling; move buildkite_agent_snapshots
  retention into the dedicated /api/cron/retention task per upstream's
  decoupling of raw-snapshot cleanup.
- Renumber the agent-snapshots migration to 0020 (upstream took 0018 and
  0019).

This branch was successfully deployed

1 active deployment
Preview — 6dd94a02 Deployed Sep 3, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant