使用Supervisor守护Python进程
1、需求现在有一个进程需要每时每刻不断的跑，但是这个进程又有可能由于各种原因有可能中断。当进程中断的时候我希望能自动重新启动它，此时，就需要使用到了Supervisor。Supervisor起到守护进程的作用。
2、安装
https://pypi.python.org/pypi/supervisor
tar zxvf supervisor-3.0.tar.gz
cd supervisor-3.0
python2.7 setup.py build
python2.7 setup.py install
需要依赖于meld3-1.0.0.tar.gz，联网安装，由于所在机器无法上网，故进行手动安装meld。
参照上面方法再进行安装meld，安装完成之后再次进行install即可。
测试安装是否成功：echo_supervisord_conf
配置文件：echo_supervisord_conf > /etc/supervisord.conf
3、Supervisor相关
supervisord ： supervisor的服务器端部分，启动supervisor就是运行这个命令
supervisorctl：启动supervisor的命令行窗口
4、Demo1测试
需求是对本地的API进程进行守护，要求这个服务能在意外停止后自动重启。
API的执行命令为：python2.6 /data1/guosong/mysqlapi/main/api_main.py --port=8005
修改配置文件：
command = python2.6 /data1/guosong/mysqlapi/main/api_main.py --port=8005
autostart = true
autorestart = true
startsecs = 3
stdout_logfile = /data1/guosong/mysqlapi/log/mysqlapi_demo1.log
启动supervisord：
# ps aux |grep super
root 153770.00.061268 788 pts/3 R+ 16:23 0:00 grep super
# supervisord
Unlinking stale socket /var/tmp/supervisor.sock
# ps aux |grep super
root 154580.00.0 1471487976 ? Ss 16:23 0:00 /usr/bin/python /usr/bin/supervisord
root 155330.00.061268 792 pts/3 S+ 16:23 0:00 grep super
查看进程状态：
# supervisorctl
mysqlapi RUNNING pid 15460, uptime 0:01:19
supervisor> status
mysqlapi RUNNING pid 15460, uptime 0:01:21
supervisor> exit
# ps aux |grep 8005
root 154600.20.0 208436 15484 ? S 16:23 0:00 python2.6 /data1/guosong/mysqlapi/main/api_main.py --port=8005
root 163140.00.061268 788 pts/3 S+ 16:24 0:00 grep 8005
#
kill测试：
# ps aux |grep 8005
root 154600.20.0 208436 15484 ? S 16:23 0:00 python2.6 /data1/guosong/mysqlapi/main/api_main.py --port=8005
root 163140.00.061268 788 pts/3 S+ 16:24 0:00 grep 8005
#
# kill 15460
# ps aux |grep 8005
root 17431 21.00.0 208436 15484 ? S 16:25 0:00 python2.6 /data1/guosong/mysqlapi/main/api_main.py --port=8005
root 174370.00.061268 772 pts/3 R+ 16:25 0:00 grep 8005
# kill -9 17431
# ps aux |grep 8005 |grep -v grep
root 175407.00.0 208440 15488 ? S 16:25 0:00 python2.6 /data1/guosong/mysqlapi/main/api_main.py --port=8005
#
使用kill以及kill -9 进程的id都发生变化
使用supervisorctl进行程序重启：
# supervisorctl stop all
mysqlapi: stopped
# ps aux |grep 8005 |grep -v grep
# supervisorctl start all
mysqlapi: started
# ps aux |grep 8005 |grep -v grep
root 196493.00.0 208440 15488 ? S 16:28 0:00 python2.6 /data1/guosong/mysqlapi/main/api_main.py --port=8005
#
5、Demo2测试
守护多个进程，修改/etc/supervisord.conf配置
command=python26/data1/guosong/mysqlapi/main/api_main.py --port=140%(process_num)02d
process_name=%(program_name)s_%(process_num)02d ; process_name expr (default %(program_name)s)
numprocs=8
numprocs_start=81
startretries=3
stopwaitsecs=10
autorstart=true
log_stdout=true
log_stderr=true
logfile=/data1/guosong/mysqlapi/log/mysql_api_demo.log
超过2位数如何表示？
command=python26/data1/guosong/mysqlapi/main/api_main.py --port=140%(process_num)02d
结果演示：
6、问题集锦
问题1-mysqlapi entered FATAL state, too many start retries too quickly:
2014-05-15 17:26:30,898 INFO exited: mysqlapi (exit status 1; not expected)
2014-05-15 17:26:30,899 INFO received SIGCLD indicating a child quit
2014-05-15 17:26:31,901 INFO gave up: mysqlapi entered FATAL state, too many start retries too quickly
原因在于版本导致，api的脚本版本为python26，故使用supervisord也需要使用python26进行安装，查看进程信息如下：
root 240580.00.0 151420 10204 ? Ss 18:28 0:00 /usr/bin/python26 /usr/bin/supervisord -c /etc/supervisord.conf
问题2- http://127.0.0.1:9001 refused connection:
# supervisorctl
http://10.75.xxx.xxx:9001 refused connection
原因在于没有开启inet_http_server，配置文件默认使用分号注释每行
将inet_http_server前面的分号去除后，重新开启supervisord
# /etc/init.d/supervisord stop
Stopping supervisord:
# supervisord -c /etc/supervisord.conf
# supervisorctl
mysqlapi:mysqlapi_81 RUNNING pid 32513, uptime 0:00:02
mysqlapi:mysqlapi_82 RUNNING pid 32511, uptime 0:00:02
mysqlapi:mysqlapi_83 RUNNING pid 32512, uptime 0:00:02
mysqlapi:mysqlapi_84 RUNNING pid 32509, uptime 0:00:02
mysqlapi:mysqlapi_85 RUNNING pid 32510, uptime 0:00:02
mysqlapi:mysqlapi_86 RUNNING pid 32507, uptime 0:00:02
mysqlapi:mysqlapi_87 RUNNING pid 32508, uptime 0:00:02
mysqlapi:mysqlapi_88 RUNNING pid 32514, uptime 0:00:02
访问界面：
页:
[1]