Django Postgres关闭了与芹菜工人的连接

我有一个Django（1.8）应用程序，托管在Heroku上，使用Postgres（9.4.4）数据库。有两个pro-dyno：一个运行web服务器，一个运行芹菜（3.1）worker，芹菜配置了

在过去的几周/几个月里，这个应用程序一直在平稳运行，但有几天我偶尔会遇到几秒钟的停机时间

从日志来看，它似乎与因未知原因循环的dynos有关：

...<some successful requests>...
heroku[web.1]: State changed from up to starting
heroku[web.1]: Cycling
heroku[web.1]: Stopping all processes with SIGTERM
... Cycling in progress
heroku[web.1]: State changed from starting to up

现在这个问题每天发生15次以上

有人知道为什么dynos会收到这个信号吗？另外，为什么循环会导致“操作错误：由于管理员命令而终止连接”错误？为什么即使我们远远低于heroku设定的限制，Postgres仍在关闭连接

---

顺便说一句，应用程序在Postgres 9.3本地运行时没有任何问题。

不确定这是否为您解决了，但我非常确定错误消息表明是数据库服务器而不是worker dyno收到了信号。

heroku[worker.1]: State changed from up to starting
heroku[worker.1]: Cycling
heroku[worker.1]: Stopping all processes with SIGTERM
app[worker.1]: worker: Warm shutdown (MainProcess)
app[worker.1]: [INFO/MainProcess] beat: Shutting down...

...Celery being restarted...

app[worker.1]: [INFO/Beat] beat: Starting...
app[worker.1]: [INFO/MainProcess] Connected to amqp://my-queue-address
app[worker.1]: [INFO/Beat] Writing entries...
app[worker.1]: [INFO/MainProcess] mingle: searching for neighbors
app[worker.1]: [WARNING/MainProcess] celery@xxxxxxxxxxx ready.
app[worker.1]: [INFO/MainProcess] mingle: all alone
app[worker.1]: [INFO/Beat] Scheduler: Sending due task celery.backend_cleanup (celery.backend_cleanup)
app[worker.1]: [INFO/MainProcess] Received task: celery.backend_cleanup[13f89001-08b0-434b-bc7e-372ffcfef521]
app[worker.1]: [INFO/Beat] Writing entries...
app[worker.1]: [INFO/MainProcess] Task celery.backend_cleanup[13f89001-08b0-434b-bc7e-372ffcfef521] succeeded in 0.00162791600451s: None
app[worker.1]: [ERROR/Beat] Database gave error:
OperationalError('terminating connection due to administrator command\nSSL connection has been closed unexpectedly\n',)

...Flood of error logs...

app[worker.1]:   File "/app/.heroku/python/lib/python2.7/site-packages/django/db/backends/postgresql_psycopg2/base.py", line 218, in _set_autocommit
app[worker.1]:     six.reraise(dj_exc_type, dj_exc_value, traceback)
app[worker.1]: InterfaceError: connection already closed