Repository navigation
Conversation
(ai: vllm-qwen/qwen3.8-27b) 查重: 不重复。 本 PR 是 #4762/#4784 的后续——#4784 引入了「fatal 模式由 worker 自杀」的路径,本 PR 修的是该路径的实际缺陷(worker 的 标签: bug + scope/other + p2(缺陷在 serving 进程生命周期面、不属特定客户端车道;触发条件 = 主循环同步阻塞 >5min,有手动 kill 绕行、无数据丢失)。 审核进行中: 正对 PR head |
…d with the constant un-export The fix commit un-exported WATCHDOG_HEARTBEAT_MS / WATCHDOG_SOFT_WARN_MS / WATCHDOG_HARD_TIMEOUT_MS and removed the only test asserting their values (5_000 / 30_000 / 300_000), leaving the production thresholds unpinned. Re-adds the exports and restores the 'defaults are deadlock-scale' test verbatim from the pre-fix tree. Verified: watchdog suite 8/8, unused-exports gate clean (same 8 pre-existing waivers), tsc --noEmit clean.
(ai: vllm-qwen/qwen3.8-27b) Review 结论:方向正确、机制成立,建议合并(等新 head CI 绿)——已直接向分支补一个小修 commit方向审核(原问题):#4762/#4784 的诊断成立。fatal worker 里 回归清点(逐项过既有行为):
行为变更披露(旧值 → 新值):
测试有效性:新 exit-mode 测试有牙——本地 A/B 实测:换回旧实现( 已直接补进分支的小修(独立 commit 结构化八问:
CI 状态:原 head 505d0b6 全部快速 guard 绿(kernel-guard / lane-guard / pi-subagents-guard / static-gates / unused-exports / test-import-graph 等);test shard 矩阵因新 push 被自动取消(正常 churn,非失败);新 head 99b228e 的 CI 运行中,全套判定以远程 CI 为准。 合并判定:建议合并(待 99b228e CI 绿)。按 §7.4 不满足 auto-merge:无专属 issue 可 |
…it' into fix/event-loop-watchdog-process-exit
|
Thank you for the review and The follow-up removes the POSIX-only child At head
No production process or live session was signalled during qualification. The availability-policy tradeoff and possible loss of unflushed state remain documented; this does not identify the cause of historical long-running computation or guarantee an unsupervised restart. |
(ai: vllm-qwen/qwen3.8-27b) 第二轮 Review 结论(d5f2f31):通过,可合并(人工合并)Delta 审计(99b228e → d5f2f31,共 2 文件):
结构化八问(本轮 delta = 删一条平台相关断言 + changelog 一句):
状态核对:
判定:可合并。按 §7.4 不 auto-merge(无专属 issue 供 |
|
Thank you for the review and The follow-up removes the POSIX-only child At head
No production process or live session was signalled during qualification. The availability-policy tradeoff and possible loss of unflushed state remain documented; this does not identify the cause of historical long-running computation or guarantee an unsupervised restart. |
(ai: vllm-qwen/qwen3.8-27b) 这条评论与上一条转发内容相同(同一 head
|
…hdog-process-exit
(ai: vllm-qwen/qwen3.8-27b) 审查结论:APPROVE(可合,人工合并) 独立读了当前 head 结构化八问
测试有效性:exit-mode 隔离子进程测试是真回归——阻塞主循环→watchdog SIGKILL→断言「有界终止( diff 整洁度:净 3 文件、每行都服务本 PR;CHANGELOG 净 +1 行(diff 里看到的其它一大片条目是 master 合并带入、合并回 master 时为 no-op,不算本 PR scope creep);无空行/重排噪音。代码整洁(⑤b):命名达意、注释讲 why、无死代码、零新依赖,全过。 CI:head 判定:APPROVE。按 §7.4 不走自动合并(含可用性策略行为变更 + 无独立 Fixes#N issue,系 #4762/#4784 后续)→ 人工合并。未触碰 kernel/,无需 🔶 决策。review-worker-r1 完成。 |
… (conflict: CHANGELOG.md — kept both sides' Unreleased entries)
(ai: vllm-qwen/qwen3.8-27b) Round-2 verdict: APPROVE — head What changed since my round-1 review (
CI for this head: 0 failures. Fast gates green (docs-sync, gate, lock-sync, mysql-contract, static-gates ubuntu-24, version-guard). The full test matrix (11 shards) plus red-signal and windows static-gates show as cancelled in this snapshot rather than passed, so a green full-suite result for this head isn't captured yet — please confirm the test lane is green before merging (branch protection enforces it regardless). Merge recommendation: human merge per §7.4 — this is an availability-policy behavior change (fatal mode now terminates the whole serving process instead of just its worker) and the PR has no dedicated 8问(中文)
判定:APPROVE(人工合并)。代码与上轮已审状态逐字节一致,本轮仅干净并入 master,CI 0 失败;合并前请确认测试矩阵转绿。 |
[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b) Round-2 review of PR #5450 (head Verdict: APPROVE — human merge (§7.4, not auto-merge: availability-policy behavior change + no dedicated Verified this round (independent pass, not carried over):
Remaining before merge: the full test matrix (11 shards) + red-signal + windows static-gates read as cancelled in the cached snapshot rather than passed, so a green full-suite result isn't captured yet — confirm the test lane is green before merging (branch protection enforces it regardless). No mechanical spot-fix needed this round. 判定:APPROVE(人工合并)。代码与上轮已审状态逐字节一致,本轮仅干净并入 master,CI 0 失败;合并前请确认测试矩阵转绿。review-worker-r1 完成。 |
… (clean auto-merge, incl. ranxianglei#6132/ranxianglei#6150 compressAtFoldPoint removal)
… (conflict: CHANGELOG.md — kept both sides' Unreleased entries)
[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b) Round-3 verdict: APPROVE — human merge. Head Since my round-2 review ( Independent re-verification on
CI for No mechanical spot-fix needed this round. 8问(中文) — 代码与第 1、2 轮已审状态逐字节一致,八项结论不变,各一句:
判定:APPROVE(人工合并)。代码与两轮已审状态逐字节一致,本轮仅干净并入 master,CI 0 失败、无取消;合并前请确认测试矩阵转绿。review-worker-r1 完成。 |
… (conflict: CHANGELOG.md — kept both sides' Unreleased entries)
(ai: vllm-qwen/qwen3.8-27b) Conflict resolution complete — PR is mergeable. Four takeover runs in 24h (master moved 4 times, re-dirtying the PR each time); all resolved the same mechanical way. Current state (verified live this run): head What happened, per run:
Invariant preserved at every merge: Recurring anomaly (flagged in #4599): the core-matrix CI run was cancelled by an unknown external actor twice (13 jobs on 11-11, 3 jobs on 11-12), mid-run, with no janitor/supersede/concurrent-push explanation. Both reruns went green. Worth a human look at who/what cancels in-progress 结构化审核清单(8问,中文)
判定:冲突已全部解决,PR 可合并(当前仅剩人工审批门禁);新 head 的 CI 跑完后即可由 owner 合并。 |
[bot] (ai: vllm-qwen/qwen3.8-27b) 进度汇报(被强制重启后恢复):CI 红项已消除,PR 全绿、无新冲突,仅剩人工合并。 我此前在 head CI 状态(head
冲突状态
关于卡住:恢复前我停在该 windows 红项的核实环节约 30 分钟无输出被强制重启;已改用单次定向读取(PR 状态 / check-runs / master tip / 该 job 的 run_attempt)而非循环轮询完成核实,避开长轮询。 判定
给 owner:请人工合并 https://github.com/ranxianglei/billion-context/pull/5450(CI 已全绿)。 |
[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b) Round-4 verdict: APPROVE — human merge. Head (本轮进程曾因一次网络调用挂起被守护重启;已对所有网络/git 操作加超时限制,本轮复核干净完成。) Since my round-3 review ( Independent re-verification on
CI for No mechanical spot-fix needed this round. 8问(中文) — 代码与第 1–3 轮已审状态逐字节一致,八项结论不变,各一句:
判定:APPROVE(人工合并)。代码与三轮已审状态逐字节一致,本轮仅干净并入 master,CI 0 失败、无取消(测试分片仍在排队);合并前请确认测试矩阵转绿。review-worker-r1 完成。 |
Problem
Follow-up to #4762 / #4784: the fatal watchdog runs inside a Node Worker. Calling
process.exit(1)there exits only that worker, leaving the blocked serving process alive and its listening port occupied. Asking the parent to run an exit callback also cannot work while its event loop is blocked.Change
After the existing synchronous FATAL forensic append, fatal mode signals the shared OS process ID with
SIGKILL. Report mode remains nonfatal. Thresholds, heartbeat/unref/stop behavior, plugin-host isolation, and the dead-stderr safeguard are unchanged. A worker-created forensic log requests mode0600.The existing five-minute threshold is an availability policy, not proof of a deadlock. Legitimate exceptionally long synchronous computation can cross it, and forced termination can lose unflushed state. This change implements the already-intended fatal policy; it does not claim to fix the underlying expensive computation or guarantee a launcher restart.
Verification
0600on Linux.node --import ./tests/preload-storage-baseline.mjs --import tsx --test tests/event-loop-watchdog.test.ts: 7 passed.npm exec -- tsc --noEmit --project tsconfig.json: passed.npm run check:unused-exports: passed, with the 8 pre-existing waivers reported.npm run check:test-import-graph: passed.No production process or live session was signalled during verification. No kernel source or version is changed.