[Cherry-Pick][KVCache][BugFix] Buffer early layer0 cache signals(#7872)

[kevincheng2]

Motivation

Cherry-pick #7872 to release/2.6.

Layer0 cache completion signals can arrive before the cache task is registered. Without buffering, the prefill cache send thread may miss these early layer0 signals.

Modifications

• Add pending layer0 signal buffering in CacheMessager.
• Recover buffered layer0 signals when cache task info is registered.
• Drop stale pending signals when cache tasks finish.
• Add unit tests for early layer0 signal buffering and recovery.

Usage or Command

No usage change.

Targeted test attempted:

python -m pytest tests/cache_manager/test_cache_messager.py -q
/root/paddlejob/inference-public/chengyanfu/.venv/py310/bin/python -m pytest tests/cache_manager/test_cache_messager.py -q

Both commands were stopped by the current shell environment with exit code 147 before test completion.

Accuracy Tests

Not applicable. This PR does not change model outputs.

Checklist

• Add at least a tag in the PR title.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[paddle-bot]

Thanks for your contribution!

[PaddlePaddle-bot]

🤖 Paddle-CI-Agent | pr_review | 2026-05-22 14:45:56

📋 Review 摘要

PR 概述：Cherry-pick #7872，修复 CacheMessagerV1 中 layer0 缓存完成信号早于任务注册到达时被丢失的问题
变更范围：fastdeploy/cache_manager/cache_messager.py、tests/cache_manager/test_cache_messager.py
影响面 Tag：[KVCache]

问题

级别	文件	概述
📝 PR 规范	标题	Cherry-Pick 标题含两个官方 Tag [KVCache][BugFix]，规范要求仅含一个

仅 PR 规范问题，无阻塞性 Bug。

📝 PR 规范检查

标题 [Cherry-Pick][KVCache][BugFix] Buffer early layer0 cache signals(#7872) 含两个官方 Tag，而 checklist §D1 规定 Cherry-Pick PR 标题格式为 [Cherry-Pick][Tag] 标题描述(#原PR号)，仅允许一个功能 Tag。

标题建议（可直接复制）：

• [Cherry-Pick][BugFix] Buffer early layer0 cache signals(#7872)

PR 描述结构完整，所有 §D2 必填段落（## Motivation、## Modifications、## Usage or Command、## Accuracy Tests、## Checklist）均有实质内容，Checklist 勾选合理，无需修改。

总体评价

实现思路清晰：通过 pending_layer0_signals 字典缓冲早到的 layer0 信号，任务注册时原子恢复并进行 token 数量合法性校验，任务完成时清理 stale 信号；锁顺序（engine_cache_task_thread_lock → pending_layer0_signal_lock）在所有三处代码路径中保持一致，无死锁风险。新增三个单元测试覆盖缓冲恢复、无效信号丢弃、stale 清理场景，代码质量良好，仅标题含双 Tag 需按规范修改，不影响合入。

[PaddlePaddle-bot]

🤖 Paddle-CI-Agent | ci_status_monitor | 2026-05-23 01:53:08

CI报告基于以下代码生成（30分钟更新一次）:

• PR commit: 198426c
• Merge base: b562b8d (branch: release/2.6)
• 查看完整 Diff
• CI 详情

---

1 任务总览

所有 required 任务已通过，建议通过；当前存在 3 个 optional 任务失败，不阻塞合并，仅供参考。

总执行（rerun次数）	总任务	✅ 通过	❌ 失败	⏳ 运行中	⏸️ 等待中	跳过
36(0)	36	33	3	0	0	0

2 任务状态汇总

日志列说明：失败任务直接使用 CI 日志链接，运行中任务手动拼接 Job 链接。

2.1 Required任务 : 10/10 通过

必选任务阻塞合并，失败需优先处理。

状态	任务	耗时	根因	修复建议	日志	重跑
✅	其余 10 个必选任务通过	-	-	-	-	-

2.2 可选任务 — 23/26 通过

可选任务不阻塞合并，失败仅供参考。

状态	任务	耗时	日志	重跑
❌	Run iluvatar Tests / run_iluvatar_cases	15m40s	Job	-
❌	Trigger Jenkins for PR	1m1s	Job	-
❌	CI_HPU	1h6m	Job	-
✅	其余 23 个可选任务通过	-	-	-

---

3 失败详情（仅 required）

无 required 失败任务。

[codecov-commenter]

Codecov Report

❌ Patch coverage is 97.50000% with 1 line in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (release/2.6@b562b8d). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
fastdeploy/cache_manager/cache_messager.py	97.50%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@              Coverage Diff               @@
##             release/2.6    #7896   +/-   ##
==============================================
  Coverage               ?   72.39%           
==============================================
  Files                  ?      381           
  Lines                  ?    54252           
  Branches               ?     8480           
==============================================
  Hits                   ?    39276           
  Misses                 ?    12216           
  Partials               ?     2760           

Flag	Coverage Δ
GPU	72.39% <97.50%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 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.