Cannot contact delegate daemon. Daemon died?

[Crispig]

我按技术文档配置，调度机和客户机都各自启动了scheduler和daemon，调度机检测到新的servant。

I1110 14:19:54.431716 1780984 init.cc:114] Flare started.
I1110 14:19:54.432881 1780984 runtime.cc:425] Using fiber scheduling profile [neutral].
I1110 14:19:54.432904 1780984 runtime.cc:220] Starting 8 worker threads per group, for a total of 1 groups. The system is treated as UMA.
I1110 14:19:54.436939 1780984 init.cc:122] Flare runtime initialized.
I1110 14:19:54.702870 1780997 task_dispatcher.cc:205] [192.168.33.5:46518] Discovered new servant at [192.168.33.5:8335]

但是运行YADCC_LOG_LEVEL=0 CXX='/home/ubuntu/.yadcc/symlinks/g++' LD='/home/ubuntu/.yadcc/symlinks/g++' CC='/home/ubuntu/.yadcc/symlinks/gcc' make -j8
提示

[2022-11-10 14:20:33.068854] [TRACE] [yadcc/client/yadcc.cc:261] Started
[2022-11-10 14:20:33.068886] [DEBUG] [yadcc/client/utility.cc:92] Looking up for [gcc] in [/home/ubuntu/.yadcc/symlinks].
[2022-11-10 14:20:33.068912] [DEBUG] [yadcc/client/utility.cc:92] Looking up for [gcc] in [/usr/local/sbin].
[2022-11-10 14:20:33.068919] [DEBUG] [yadcc/client/utility.cc:92] Looking up for [gcc] in [/usr/local/bin].
[2022-11-10 14:20:33.068926] [DEBUG] [yadcc/client/utility.cc:92] Looking up for [gcc] in [/usr/sbin].
[2022-11-10 14:20:33.068932] [DEBUG] [yadcc/client/utility.cc:92] Looking up for [gcc] in [/usr/bin].
[2022-11-10 14:20:33.068942] [TRACE] [yadcc/client/utility.cc:99] Found [gcc] at [/usr/bin].
[2022-11-10 14:20:33.068947] [TRACE] [yadcc/client/yadcc.cc:133] Using compiler: /usr/bin/gcc
[2022-11-10 14:20:33.068980] [ERROR] [yadcc/client/task_quota.cc:61] Cannot contact delegate daemon. Daemon died?
[2022-11-10 14:20:34.064286] [ERROR] [yadcc/client/task_quota.cc:61] Cannot contact delegate daemon. Daemon died?
[2022-11-10 14:20:34.065163] [ERROR] [yadcc/client/task_quota.cc:61] Cannot contact delegate daemon. Daemon died?
[2022-11-10 14:20:34.068817] [ERROR] [yadcc/client/task_quota.cc:61] Cannot contact delegate daemon. Daemon died?
[2022-11-10 14:20:34.069060] [ERROR] [yadcc/client/task_quota.cc:61] Cannot contact delegate daemon. Daemon died?
[2022-11-10 14:20:35.064416] [ERROR] [yadcc/client/task_quota.cc:61] Cannot contact delegate daemon. Daemon died?
[2022-11-10 14:20:35.065259] [ERROR] [yadcc/client/task_quota.cc:61] Cannot contact delegate daemon. Daemon died?
[2022-11-10 14:20:35.068913] [ERROR] [yadcc/client/task_quota.cc:61] Cannot contact delegate daemon. Daemon died?
[2022-11-10 14:20:35.069152] [ERROR] [yadcc/client/task_quota.cc:61] Cannot contact delegate daemon. Daemon died?

请问可能是什么问题？

[Crispig]

我发现我得在调度机上也启动守护进程才能解决

[0x804d8000]

调度机应该不用，客户机需要daemon。

客户机的daemon的具体的作用可以参考 https://github.com/Tencent/yadcc/blob/master/yadcc/doc/daemon.md#处理本地请求

[Crispig]

非常感谢，但我在调度机上起编译任务，好像得在调度机上同时起守护进程

[0x804d8000]

是的，起编译任务的机器需要daemon存活。如果调度机发起编译，那么这时候调度机同时是客户机&调度机两种角色。

ps: 对于C/S方式的部署，客户机的daemon可以配置为不接受网络任务，参考https://github.com/Tencent/yadcc/blob/master/yadcc/doc/daemon.md#参数

[Crispig]

嗯嗯，已经解决了，感谢感谢！！调度机默认端口 8836在修改代码之后重新编译可以换成其他的端口吧，这里还存在其他依赖吗

[0x804d8000]

可以换，daemon的参数--scheduler_uri里面的端口号和调度机端口一致就可以

[Crispig]

已经解决，感谢！
30核的情况下CPU利用率几乎能跑满，但扩展到130核发现利用率不足，请问有什么解决方案吗，换更大的workload好像没有改善。

[0x804d8000]

先看看客户机负载，如果负载很高可能是本地预处理速度不够，可以考虑按照https://github.com/Tencent/yadcc/blob/master/yadcc/doc/client.md#配置 关闭YADCC_COMPILE_ON_CLOUD_SIZE_THRESHOLD（置为0）、启用YADCC_IGNORE_TIMESTAMP_MACROS（置为1）

之后可以参考https://github.com/Tencent/yadcc/blob/master/yadcc/doc/debugging.md#守护进程 看看distributed_task_dispatcher下面的信息，看看本地daemon收到的任务数。如果太少可能是构建系统因为各种原因运行的并发度太低（依赖关系/构建并发度/…），如果很多但是都是waiting状态可以在调度机检查是不是上报上来的编译器hash不匹配导致任务分发不出去

另外也可以看看客户机有没有其他非预期的资源瓶颈（比如并发度太高的时候可能磁盘/网络会跑满，但是这种一般是几百并发才会遇到）

[Crispig]

好的好的，我排查一下，非常感谢！