背景
这个问题其实之前遇到的还挺多的, 简单来说, 就是某个A服务启动B服务, 但是B服务又必须等待另一个服务启动完毕(比如A服务), 产生了一定的死锁. 而在ps的进程表中看到的就是systemd-tty-ask-password-agent这样一个agent进程卡住. 一般查一下当前卡住的service和他依赖的after,before的几个服务, 就知道了, 一般就是循环死锁了.
具体理论这块还没去查源码, 暂时只能给个结论.有大佬看过的话, 希望也能分享下~
样例
``` bash # new2.service [Unit]
Description=System Logging Service
Documentation=man:rsyslogd(8)
Documentation=https://www.rsyslog.com/doc/
[Service]
Type=forking
EnvironmentFile=-/etc/sysconfig/rsyslog
ExecStart=/root/new2.sh
Increase the default a bit in order to allow many simultaneous
files to be monitored, we might need a lot of fds.
[Install]
WantedBy=multi-user.target
new2.sh
#!/bin/bash echo "heelo" sleep 1000 & systemctl restart new1
new1.service
[Unit]
Description=System Logging Service
Documentation=man:rsyslogd(8)
Documentation=https://www.rsyslog.com/doc/
After=new2.service
[Service]
Type=forking
EnvironmentFile=-/etc/sysconfig/rsyslog
ExecStart=/root/new1.sh
Increase the default a bit in order to allow many simultaneous
files to be monitored, we might need a lot of fds.
[Install]
WantedBy=multi-user.target
new.sh
#!/bin/bash echo "heelo" sleep 1000 & ``` bash