服务器虚拟环境下supervisor启动django-celery

最近在弄django+python+uwsgi的微信部署, 因为要给python加入定时任务, 用以定期改变数据库数据, 所以加入了supervisor.

supervisor安装参考文章,  3.3.2版本我是用的默认配置的supervisord.conf.

重点内容:

1. 项目文件建立在/etc/supervisor/conf.d/ 目录下, touch yourfilename.conf
   
   
2. 我的项目文件celery_worker.conf
   

   # supervisor的进程名称为celery_worker
   [program:celery_worker]
   # 进入工作目录
   directory=/home/bosheng/work/projects/rxiiodev
   # 执行celery指令
   command=python manage.py celery worker -l info
   autorestart=true
   loglevel=info
   redirect_stderr=true
   stdout_logfile=/var/log/supervisor/celery_worker.log# 在项目在虚拟环境下, 需要配置虚拟环境库所在的位置, 如下:environment=PYTHONPATH="$PYTHONPATH:/home/bosheng/work/vrxiio/lib/python2.7/site-packages"
   

   
   
3. celery_beat.conf同上, 区别如下:
   

   [program:celery_beat]
   command=python manage.py celery beat
   stdout_logfile=/var/log/supervisor/celery_beat.log

   
   
4. 需要用supervisor运行celery的worker进程, 还需要在tasks.py 文件里加入:
   

   from celery import platforms
   platforms.C_FORCE_ROOT = True

   
   我的celery tasks.py:
   

   #!/usr/bin/env python
   # -*- coding: utf-8 -*-
   
   from celery import task, platforms
   from wauth.models import wxsetting, order, product,booking,slot,contactinfo,userprofile,vouch,vouchdef, pagecontent
   from datetime import datetime, timedelta
   
   platforms.C_FORCE_ROOT = True
   
   
   @task(name='wauth.tasks.order_timeout')
   def order_timeout():
       yesterday = (datetime.now()-timedelta(days=1)).date()
       qry = order.objects.filter(place_date__gt=yesterday, active=True)
       for p in qry:
           c = p.bookings.filter(active=True)
           for q in c:
               t = q.service_date.strftime('%Y-%m-%d')
               # 处理订单时间
               if q.connected_order.id == p.id:
                   tmp_second = (
                       datetime.now() - p.place_date.replace(tzinfo=None) - timedelta(hours=8)
                   ).total_seconds()
                   remain_time = 900 - tmp_second
                   if remain_time <= 0 and not q.connected_order.ispaid:
                       if q.active is True and p.active is True:
                           q.active = False
                           q.save()
                           p.active = False
                           p.save()
                       else:
                           pass
       return 'timeout order clear'

   
   
5. 启动supervisor:
   

   sudo supervisord -c /etc/supervisor/supervisord.conf
   sudo supervisorctl -c /etc/supervisor/supervisord.conf

   
   
   
6. sudo supervisorctl
   进入supervisor交互指令, 参考:
   

   # 查看进程状态
   supervisorctl status
   # 停止某一个进程，program_name 为 [program:x] 里的 x
   supervisorctl stop program_name
   # 启动某个进程
   supervisorctl start program_name
   # 重启某个进程
   supervisorctl restart program_name
   # 结束所有属于名为 groupworker 这个分组的进程 (start，restart 同理)
   supervisorctl stop groupworker:
   # 结束 groupworker:name1 这个进程 (start，restart 同理)
   supervisorctl stop groupworker:name1
   # 停止全部进程，注：start、restart、stop 都不会载入最新的配置文件
   supervisorctl stop all
   # 载入最新的配置文件，停止原有进程并按新的配置启动、管理所有进程
   supervisorctl reload
   # 根据最新的配置文件，启动新配置或有改动的进程，配置没有改动的进程不会受影响而重启
   supervisorctl update

   
   
7. 重新进入supervisorctl交互:
   

   sudo supervisorctl -c /etc/supervisor/supervisord.conf

   
   
8. 一些疑问:
   (1) 为什么要用sudo来启动supervisor?
   (2) 怎样输出celery worker 和beat的日志文件?
   解决: supervisor和独立进程的日志所在在文件夹需要手动建立, 如在log下建立supervisor文件夹: /var/log/supervisor/
   (3) 加入C_FORCE_ROOT = True 会带来什么影响?
   疑似解决: 允许root权限运行celery