systemd 是否支持 fork 守护进程的故障重启?

Mat*_*att 8 systemd exit-status

最初发布在 StackOverflow 上,被告知来这里问。回顾一下那次谈话——

  1. 不,我还没有真正能够尝试进行更改以查看会发生什么(并非没有可能影响正在运行的生产服务)。
  2. RTFM 尽职调查完成。他们不包括这个。

我原来的问题如下:

假设我的systemd单元文件中有以下内容:

Type=forking
Restart=on-failure
Run Code Online (Sandbox Code Playgroud)

父进程以状态 0 退出(子进程成功启动)。在稍后的某个时刻,孩子以非零状态死亡。发生什么了? systemd可以跟踪子守护进程PID:

Process: 1768 ExecStart=/bin/mydaemon (code=exited, status=0/SUCCESS)
Main PID: 1770 (mydaemon)
Run Code Online (Sandbox Code Playgroud)

Restart=on-failure只盯着父退出状态,或者还有孩子呢?

the*_*rpy 7

我创建了以下单元文件:

[Unit]
Description=Something

[Service]
Type=forking
WorkingDirectory=/tmp
ExecStart=/tmp/script.sh
ExecStop=/tmp/script.sh
Restart=on-failure
Run Code Online (Sandbox Code Playgroud)

script.sh 包含以下内容:

#!/bin/sh
echo "Forking"
/tmp/myscript.sh &
Run Code Online (Sandbox Code Playgroud)

myscript.sh 包含以下内容:

#!/bin/sh
sleep 60
exit 1
Run Code Online (Sandbox Code Playgroud)

果然,每隔 60 秒,systemd 就会重启服务。

如图所示,注意不同的 PID 和开始时间,注意它会重新启动父级:

linux:~ # systemctl status myService.service
? myService.service - Something
   Loaded: loaded (/etc/systemd/system/myService.service; static; vendor preset: disabled)
   Active: active (running) since Mon 2017-07-10 20:43:29 CEST; 57s ago
  Process: 4393 ExecStart=/tmp/script.sh (code=exited, status=0/SUCCESS)
 Main PID: 4396 (script.sh)
    Tasks: 2 (limit: 512)
   CGroup: /system.slice/myService.service
           ??4396 /bin/sh /tmp/script.sh
           ??4397 sleep 60

Jul 10 20:43:29 linux.suse systemd[1]: Starting Something...
Jul 10 20:43:29 linux.suse script.sh[4393]: Forking
Jul 10 20:43:29 linux.suse systemd[1]: Started Something.
linux:~ # systemctl status myService.service
? myService.service - Something
   Loaded: loaded (/etc/systemd/system/myService.service; static; vendor preset: disabled)
   Active: active (running) since Mon 2017-07-10 20:44:29 CEST; 1s ago
  Process: 4409 ExecStop=/tmp/script.sh (code=exited, status=0/SUCCESS)
  Process: 4417 ExecStart=/tmp/script.sh (code=exited, status=0/SUCCESS)
 Main PID: 4420 (script.sh)
    Tasks: 2 (limit: 512)
   CGroup: /system.slice/myService.service
           ??4420 /bin/sh /tmp/script.sh
           ??4421 sleep 60

Jul 10 20:44:29 linux.suse systemd[1]: Starting Something...
Jul 10 20:44:29 linux.suse script.sh[4417]: Forking
Jul 10 20:44:29 linux.suse systemd[1]: Started Something.
Run Code Online (Sandbox Code Playgroud)