Skip to content

[PR] fix(served-markers): ignore prior container generations with same PID on restart (#3923) - #5028

Closed
fathonix wants to merge 11 commits into
ranxianglei:masterfrom
fathonix:fix/container-split-brain-pid1
Closed

fathonix wants to merge 11 commits into
ranxianglei:masterfrom
fathonix:fix/container-split-brain-pid1

Conversation

@fathonix

Copy link
Copy Markdown
Contributor

Problem

In Docker/containerized environments, processes virtually always run as PID 1. When a container restarts, its prior writer marker (/writers/.json) remains in the mounted state directory with pid: 1. When the new container boots, findContestedWriters checks isPidAlive(raw.pid). Because the new container itself is running as PID 1, process.kill(1, 0) succeeds.

This causes the newly booted proxy to falsely conclude that another concurrently running instance committed the session within CONTEST_WINDOW_MS, emitting false-positive [split-brain] errors.

Fix

  1. Record startedAt (epoch ms, derived from process.uptime()) on WriterIdentity and WriterMarkerFile.
  2. In findContestedWriters, if an observed marker file shares raw.pid === process.pid but its startedAt (or file mtime) precedes this process boot time, it is recognized as a predecessor from an earlier container generation and skipped.

Tests

  • Added unit test container restart with same PID (pid 1) does not falsely contest prior generation to tests/served-markers.test.ts.
  • Verified all 12 tests pass in tests/served-markers.test.ts.

@ranxianglei ranxianglei changed the title fix(served-markers): ignore prior container generations with same PID on restart (#3923) [PR] fix(served-markers): ignore prior container generations with same PID on restart (#3923) Oct 28, 2026
@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

进度汇报(此前会话被强制重启一次,现继续):

查重: 不重复。 同家族相关项: #3923 是原始 split-brain 报告; #3952(已合并)修的是"日志点名双方"; #4989(已合并)用 stoppedAt 处理优雅停机的前代标记。本 PR 用 startedAt 处理同 PID 跨容器代的误判——机制不同(启动时间维度 vs 停止标记),互补而非重复。

标签: bug / scope/storage / p2(缺陷类;writer marker 属 state 目录持久化面;纯观测误报、可用但有噪音,非 p1)。

审核进展(已完成方向与回归审查):

  1. 问题真实存在,修法打在根因上(PID 在容器代际间复用 → 用启动时间维度消歧),不是症状遮罩。gate 只在 raw.pid === process.pid 时触发,裸金属两个存活进程不可能同 PID → 非容器路径行为不变。
  2. 发现一个实测可复现的测试缺陷:新单测把前代表记锚定在 Date.now() - 30_000,而 gate 的判定等价于要求测试进程自身 uptime < ~29s。我用探针脚本验证:真实小 uptime 下 contested=0(测试过);把进程 uptime 模拟到 60s 后 contested=1(同一断言会挂)。node --test 每文件独立子进程,通常能过,但 CI 分片进程活得久/编译慢时这是时间炸弹。
  3. 另缺负向断言(Test Blind Spot Retrospective: Reinforcing Negative Assertion Discipline, Tightening Fake Upstream Strictness, Thin Inst | 测试盲区复盘:补反向断言纪律、假上游严格化、薄安装冒烟、形态矩阵与生命周期覆盖 #3617 门禁纪律:gate 必须双向钉住):同 PID 但晚于本进程启动的对等体仍应告警——代码层面我验证了行为正确(contested=1),但 PR 没有测试钉住这一侧。
  4. 其余小项:Date.now() - Math.round(process.uptime()*1000) 在 server.ts 与 served-markers.ts 重复两处;1000 魔数无命名常量;注释引用分支名 #container-split-brain 而非 issue The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923;gate 内用 Date.now() 而函数其余部分统一用参数 now。

下一步: 按"小问题直接修"把上述修复作为标注 commit 推到 PR 分支(若 fork 不允许 maintainer 编辑则改为贴出补丁),随后给出八问结论。定向测试已在 PR head 全绿(12/12),全量交给 CI。

@ranxianglei ranxianglei added bug Something isn't working p2 scope/storage labels Oct 28, 2026
@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

审核结论(完整八问)。查重: 不重复 —— #3923 是原始 split-brain 报告,#3952(已合)修日志点名,#4989(已合)用 stoppedAt 处理优雅停机的前代标记;本 PR 的 startedAt 处理的是"同 PID 跨容器代"误判,机制不同、互补。标签: bug / scope/storage / p2(缺陷类;writer marker 属 state 目录持久化面;纯观测误报,可用但有噪音,非 p1)。

修复已推送到 PR 分支(maintainer edit,两个标注 commit,API 核实 head 已更新为 cdd04204d,3 commits,mergeable):

  • 83b9b8e47 fix(served-markers): shared processStartMs helper + named margin + caller-clock gate
  • cdd04204d test(served-markers): test anchored to process boot + younger-twin negative row

另:master 在审核期间前进了 4 个 commit(全部 test/CI-only:#5017 ci-fixes、#4982 e2e pin),我已把当前 master 合入 PR 分支(84fb9684d);该 merge commit 因沙箱到 GitHub 的 git 通道瞬时连接重置暂未推上,后台重试中——它只影响 base 新鲜度,不含实质内容,不影响 CI 判定。若最终未落地我会补报。


八问:

  1. 是什么问题:Docker 里代理进程通常以 PID 1 运行;容器重启后挂载 state 目录里的旧 writer marker(pid: 1)还在,新进程扫描时把"上一代的自己"误判成另一个存活实例,发出假 [split-brain] error。
  2. 什么原因:存活判定只有 process.kill(pid, 0) 一维;PID 在容器代际间复用(恒为 1),新进程 kill(1,0) 对自己成功 → 误判并发双写。缺的不是 marker 机制,是"代际"维度。
  3. 影响范围与严重程度:仅限共享 state 卷的容器化部署(Docker restart / k8s pod 重建)且落在 10 分钟 contest 窗内;纯观测面(一行 error 日志 + onContest hook),不影响路由/serving/数据 → p2。
  4. 怎么修复的:marker 文件写入 startedAt(由 process.uptime() 派生的进程启动时刻),读取侧同 PID 且对端启动早于本进程 >1 s(旧版无 startedAt 时以 mtime 兜底)→ 判为前代跳过。我另按仓库标准对齐了实现质量:抽公共 helper processStartMs(now)(消除 server.ts/served-markers.ts 两处重复表达式,gate 从裸 Date.now() 改用调用方时钟 now,可测性更好)、魔数 1000 具名为 SAME_PID_GENERATION_MARGIN_MS、注释引用从分支名改为 The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923、测试锚点修正(见下)、补负向测试行(Test Blind Spot Retrospective: Reinforcing Negative Assertion Discipline, Tightening Fake Upstream Strictness, Thin Inst | 测试盲区复盘:补反向断言纪律、假上游严格化、薄安装冒烟、形态矩阵与生命周期覆盖 #3617 门禁双向纪律)。
  5. 是否有回归:逐项审过既有行为——不同 PID 的对等体不受影响(gate 仅在 raw.pid === process.pid 时触发,裸金属两存活进程不可能同 PID);[PR] served-markers: graceful dispose records stoppedAt — restart flows no longer raise split-brain alarms (#3923) #4989 的 stoppedAt skip 顺序不变(先于本 gate 判定);自 marker 排除、reap 逻辑未触碰;旧构建无 startedAt 的 marker 走 mtime 兜底(残余误报窗口仅 <1 s);原有 11 个测试零改动全过(共 13/13)。行为变更披露(旧→新):旧=容器重启后必然对前代表记告警(真双容器竞态时两侧都响);新=晚启动一侧不再对更早的前代表记告警,但更早一侧仍会检测到更新者(其 startedAt 更晚)→ 告警至少留一侧,方向即修复本意。
  6. 是否破坏缓存:不涉及(纯观测模块,模块头明确 "no routing, locking, or serving behavior changes",不碰前缀缓存/重试/游标/去重键)。
  7. 是否改了设计方向:不涉及(无架构/边界/数据流变更;marker 文件格式向后兼容,新增可选字段)。
  8. 是否引入新的配置:无(无新 env/config/默认值;startedAt 是 marker 文件内部字段)。

验证证据(本地定向,全量交 CI):

  • 原测试的 flake 实证:gate 触发等价于要求测试进程 uptime < ~29 s(前代表记锚定在墙钟 now−30 s)。探针脚本强制 uptime=60 s 时,原断言失败(contested=1);修复后同一场景 contested=0。
  • 修复后整份套件在强制 uptime=60 s 的 preload 下仍 13/13 全绿(确定性证明);正常环境 13/13;npm run typecheck clean;merge master 后复跑 13/13。

判定:可合(待 CI 在新 head 上绿;mergeable=True,base 仅落后 test/CI-only 提交,后台重试 merge commit 推送中)。合并动作留给你。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

重启后进度汇报:审查工作本身已完成(结论见上一条八问回复),卡在最后一步推送上——本环境到 GitHub 的 git 传输间歇性断连,之前的后台重试循环随进程重启被杀。现已直接推成功并验证:

等新 run 结果,全绿后在此补终局一行。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

终局一行:新 head 84fb9684d 全套 CI 全绿(39 success / 16 skipped / 0 failed)——此前唯一红腿 test (windows-latest, 22, 1) 转绿,继承红定性成立(#4982 套件缺陷由已合入 master 的 #5017 修复,本 PR diff 零责任)。PR 现为 mergeable、无阻塞项,审查结论维持上一条八问回复:可合,等 owner 合并。

ci-check added 2 commits October 30, 2026 02:15
Sync current master ahead of re-review of ranxianglei#5028 (ranxianglei#3923 same-PID container
generation gate in served-markers). Clean 3-way merge: no conflicts; hot-file
src/server.ts auto-merged keeping both master's new imports and the PR's
processStartMs wiring.
@ranxianglei

Copy link
Copy Markdown
Owner

(ai: vllm-qwen/qwen3.8-27b) Synced origin/master into this branch as part of the merge-guard: head is now 41f748e and fully contains the current master tip (a8c907d). Both sync merges were clean auto-merges with no conflicts — none of this PR's files (served-markers.ts, server.ts, port-owner.js) were touched by the intervening master commits (#5169/#5159/#5167/#4625 etc.). Local typecheck is green and the served-markers suite passes 13/13 (incl. the two same-PID-generation regression tests). The branch is up-to-date with master and mergeable; CI is re-running on the new head.

@ranxianglei

Copy link
Copy Markdown
Owner

(ai: vllm-qwen/qwen3.8-27b) Merge-guard re-sync: master advanced again to 44067a5 — merged it in cleanly (zero conflicts; none of this PR's files were touched by the new master commits). A parallel guard worker had separately merged origin/ci-fixes onto this same fork branch (8242ab3); instead of clobbering it I union-merged both lineages into head 30d794f, which now carries full latest master + the ci-fixes convergence + this PR's served-markers fix (the 3 PR files were byte-identical across both workers, so no semantic conflict). Local typecheck green, served-markers suite 13/13. Branch is up-to-date with master and mergeable; CI re-running on the new head.

ranxianglei pushed a commit that referenced this pull request Nov 7, 2026
… on restart

In containers processes virtually always run as PID 1; after a restart the old writer marker (pid: 1) still passes isPidAlive() against the new process, so findContestedWriters false-positives [split-brain] on every boot.

Fix: record startedAt (epoch ms) on WriterIdentity/WriterMarkerFile; a marker sharing our PID whose startedAt (or mtime for older builds without startedAt) predates this process boot by more than SAME_PID_GENERATION_MARGIN_MS is skipped as a prior generation. Integrated onto master's contestedServedAt helper (#4626); both mechanisms keep working.

Original commit c115e35 of PR #5028 (external contribution by @fathonix; later review-fix commits of the same PR condensed into this one — full credit retained via git author).
…glei#5028 conflict

3 conflicts resolved by taking master's side (master is a verified-CI-green
superset of the PR's ranxianglei#3923 fix): the same-PID generation guard is already in
master (subsumed), plus master's contestedServedAt() refactor + disposedAt
alias (ranxianglei#4626). Local verification: typecheck clean, tests/served-markers.test.ts 15/15.
@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

Conflict resolved for PR #5028 — the PR is now mergeable (no conflict).

What I did this run:

  1. Diagnosed the merge conflict: PR head fathonix:fix/container-split-brain-pid1 @ 30d794fd was mergeable_state=dirty against master @ c26c31ace4.
  2. Chose MERGE over rebase — the head carries 2427 commits (it absorbed an old master snapshot + ci-fixes via a prior converge merge), so a rebase would replay all of them. Merged current master into the head branch instead (non-destructive, no force-push needed).
  3. Resolved all 3 conflicts by taking master's side:
  4. Verified locally: npm run typecheck clean (exit 0) and tests/served-markers.test.ts 15/15 pass, including both The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923 regression tests ("container restart with same PID (pid 1) does not falsely contest prior generation" and "same-PID peer that booted after this process still contests").
  5. Committed the merge (2c6b058636, parents 30d794fd + c26c31ace4) and pushed to the fork branch (fast-forward — the remote head hadn't moved, so no force).
  6. Verified live: PR is now mergeable: true, mergeable_state: behind (no longer dirty). I locally confirmed the head contains master's tip (merge-base --is-ancestor = yes). The behind label is a GitHub fork-branch ahead/behind computation artifact — the PR has no conflict and is mergeable.

CI: re-running (triggered by the push). The mirror cache hasn't captured the new run yet (last poll 2026-10-30), and I did not gh-poll per policy. Local targeted verification (typecheck + the touched module's test) is green; the full-suite verdict is CI's.


八问(中文)

  1. 是什么问题:PR [PR] fix(served-markers): ignore prior container generations with same PID on restart (#3923) #5028 与 master 基线产生合并冲突(mergeable_state=dirty),导致 PR 无法合并。
  2. 什么原因导致的:PR 分支基于 10-30 的 master(44067a5bc)开发,期间 master 前进到 c26c31ace4,双方在 src/served-markers.ts(注释 + 逻辑)和 src/server.ts(import 块)共三处产生文本冲突。
  3. 影响范围与严重程度:仅限该 PR 的合并能力——冲突不解决则 PR 无法合并;不影响已发布版本或任何运行时行为。严重程度:阻塞 PR 合并(非运行时缺陷)。
  4. 怎么修复的:把当前 master 合并进 PR 分支(选 merge 而非 rebase——分支含 2427 个提交,rebase 会全部重放)。三处冲突全部取 master 侧:因为 The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923 的同 PID 代际守卫其实已经在 master 里(两侧守卫完全相同,已自动合入公共区),master 还额外有 contestedServedAt() 重构 + disposedAt 别名([Bug][split-brain] Restart-handover notice persists forever after the peer pid is gone (stale storage.splitBrain) #4626)。取 master 侧不丢失 PR 的任何内容。import 块改动已逐一核验:PR 独有符号(queryLogLines/entryScriptFingerprint)正文 0 次使用,master 侧符号全部被使用,无遗漏。
  5. 是否有回归问题:无。本地 npm run typecheck 通过(exit 0),tests/served-markers.test.ts 15/15 通过(含 The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923 两个回归测试)。取 master 侧 = 采用 master 已 CI 验证的超集,The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923 守卫保留在公共区,无行为漂移。
  6. 是否破坏缓存:不涉及。纯合并冲突解决,未触碰缓存/重试/增量游标/去重键逻辑,未改任何 wire 或压缩路径。
  7. 是否改了设计方向:不涉及。未改架构/边界/数据流;只是把 master 的既有实现(contestedServedAt 重构)并入 PR 分支。
  8. 是否引入了新的配置:不涉及。无新 env/config/默认值。

收工判定:冲突已解决,PR 现可合并(mergeable=true、无冲突);CI 因 push 重跑中,待其绿即可由 owner 合并。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

[ework] CI fix verified green ✓

…ranxianglei#5028 to current master

Master advanced 90 commits (c26c31a -> 259cd87) after the prior
conflict-resolution merge; none touch the PR's files (served-markers.ts /
server.ts / served-markers.test.ts), so this is a clean update that keeps the
PR validated against the master it will merge into.
@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

This run: confirmed the conflict was already resolved, then synced the PR to current master.

  1. Verified live state (status-freshness policy): the merge conflict from the prior run is holding — the PR was mergeable: true, mergeable_state: behind (no longer dirty). This re-trigger was based on stale dirty state.
  2. Found master had advanced c26c31ace4 → 259cd87674 (90 commits / 136 files) after the prior conflict-resolution merge — that's why the PR showed behind. None of those 90 commits touch the PR's files (src/served-markers.ts, src/server.ts, tests/served-markers.test.ts).
  3. Merged the latest master into the PR branch — clean, no conflicts (as predicted from the file-overlap check). This keeps the PR validated against the master it will merge into and clears the behind state.
  4. Verified locally: npm run typecheck clean (exit 0) + tests/served-markers.test.ts 15/15 pass (incl. both The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923 regression tests). (The workdir's node_modules didn't persist across the run boundary, so I re-ran npm install before verifying.)
  5. Pushed the master-sync merge (9b637a8f4a, parents 2c6b058636 + 259cd87674) to the fork branch — fast-forward, no force (the remote head hadn't moved).
  6. Live state now: mergeable: true, mergeable_state: blocked, base 259cd87674 (head contains it). blocked = the new CI run (triggered by the push) is still in progress — not a conflict and not a failure. The prior run's checks were green (18 success, 12 skipped, 0 failed per the mirror cache); this merge only adds master's green commits + the unchanged, verified PR files, so the new run is expected green. I did not gh-poll per policy.

Net result: conflict resolved, PR in sync with master, mergeable, CI re-running (expected green).


八问(中文)

  1. 是什么问题:PR [PR] fix(served-markers): ignore prior container generations with same PID on restart (#3923) #5028 此前与 master 有合并冲突(dirty),上一 run 已解决;但 master 随后前进了 90 个提交,PR 变成 behind(落后于基线,仍可合并)。
  2. 什么原因导致的:上一 run 解决冲突时合并的是当时的 master(c26c31ace4);之后 master 继续前进到 259cd87674(90 提交/136 文件),PR 分支没跟上,故显示 behind。本次 re-trigger 基于过期的 dirty 状态。
  3. 影响范围与严重程度:仅限该 PR 的合并就绪度——behind 不是冲突、可合并,但落后于最新 master。严重程度:就绪度问题(非缺陷、非阻塞)。
  4. 怎么修复的:把最新 master(259cd87674)合并进 PR 分支。先核验这 90 个提交不触碰 PR 的三个文件(served-markers.ts/server.ts/served-markers.test.ts),故为干净合并(无冲突)。合并后 PR 与最新 master 对齐。
  5. 是否有回归问题:无。本地 npm run typecheck 通过(exit 0),tests/served-markers.test.ts 15/15 通过(含 The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923 两个回归测试)。本次合并只并入 master 已 CI 验证的提交 + 未改动的 PR 文件,无行为漂移。
  6. 是否破坏缓存:不涉及。纯 master 同步合并,未触碰缓存/重试/增量游标/去重键逻辑。
  7. 是否改了设计方向:不涉及。未改架构/边界/数据流;只是把 PR 分支同步到最新 master。
  8. 是否引入了新的配置:不涉及。无新 env/config/默认值。

收工判定:冲突已解决,PR 已对齐最新 master(mergeable=true、无冲突);新 CI run 因 push 重跑中(blocked=检查进行中,非失败),预期绿;待其绿即可由 owner 合并。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

Status after restart — the work is complete and holding (nothing to re-do).

  • Push intact: the PR branch head is still 9b637a8f4a (the master-sync merge I pushed before the restart).
  • PR state (live): mergeable: true, mergeable_state: blocked, base = latest master 259cd87674 (head contains it). blocked = the CI run is in progress — not a conflict and not a failure.
  • CI (fresh mirror cache): 30 checks → 7 success, 11 skipped, 8 queued, 4 in_progress, 0 failed. The run is still going; no failures so far, expected green.

The conflict was already resolved and the PR is in sync with master (full 八问 in my previous comment). I did not gh-poll per policy — the mirror cache will reflect the final CI result as the run completes.

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

[ework] CI fix verified green ✓

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

Review verdict: do NOT merge — this PR is a no-op. Its fix is already in master. Please close it as superseded by 7a923f8bb7.

Critical finding (verified against the pushed head 9b637a8f4 and the GitHub API, not the PR description):

The PR's entire contribution — the #3923 same-PID container-restart guard — is already merged into master via commit 7a923f8bb7 (2026-11-07). That commit's message says it explicitly: "Original commit c115e35 of PR #5028 (external contribution by @fathonix; later review-fix commits of the same PR condensed into this one — full credit retained via git author)."

After that, the merge-guard / ci-engineer syncs repeatedly merged master back into this branch, so the two trees converged. The PR now has zero net changes:

Verdict: Merging this PR changes nothing (head tree == master tree), and the fix is already shipped in master with the author's credit retained. The correct action is to close the PR as superseded by 7a923f8bb7, not to merge it. I don't close PRs (owner-only) — please close it. No code changes were needed; there is nothing to review or fix in the head.


八问(中文)

  1. 是什么问题:这个 PR 已没有需要审查的增量。它要修的 The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923(容器重启后同 PID 误报 split-brain)修复已经在 master 里了——经 7a923f8bb7(2026-11-07),该 commit 明确把本 PR 的原始提交 c115e351 + 后续 review 修复压缩成一个,并保留 @fathonix 的 authorship。PR head 的 tree 与 master 的 tree 逐字节相同,net diff 为空。
  2. 什么原因导致的:修复先经本 PR 的 review(本线程 floor 12 的八问)后被单独合入 master(7a923f8bb7);之后 merge-guard/ci-engineer 反复把 master 合回 PR 分支,两侧收敛到同一 tree,PR 相对 master 不再有任何增量。
  3. 影响范围与严重程度:零运行时影响。PR 若被合并不改变任何代码/行为(与 master 完全相同);唯一风险是"无意义合并"污染 merge 历史,属流程问题而非缺陷。
  4. 怎么修复的:不需要改代码。正确处置 = 关闭本 PR,标注 superseded by 7a923f8bb7(修复已在 master、作者 credit 已保留)。我不自行关闭(owner-only),请 owner 关闭。
  5. 是否有回归问题:无。tree 与 master 相同,合并 = 无操作,不可能引入回归;master 侧的 The same sessionId is served by two bili processes simultaneously: interleaved logs from global + pi-native dual instanc | 同一 sessionId 被两个 bili 进程同时服务:全局 + pi-native 双实例日志交织、重复 cold-load(split-brain) #3923 修复(含两个回归测试)已过 CI。
  6. 是否破坏缓存:不涉及(无代码变更)。
  7. 是否改了设计方向:不涉及(无代码变更)。
  8. 是否引入新的配置:不涉及(无代码变更)。

判定:不可合(合并 = 无操作)。建议 owner 关闭本 PR,标注 superseded by 7a923f8bb7。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

PR #5028 review complete — verdict: do NOT merge (no-op); close as superseded by 7a923f8bb7.

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

[ework] CI fix verified green ✓

@ranxianglei

Copy link
Copy Markdown
Owner

Closing: the branch is fully in sync with master — the three-dot diff against origin/master is empty; the container-generation guard it carries is already on master (src/served-markers.ts:286).

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

Labels

bug Something isn't working p2 scope/storage

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants