Bot cannot sustain an active connection to IRC when running on heroku #1
Description
Dumping the heroku logs shows:
2013-02-28T18:54:11+00:00 app[bot.1]: [2013/02/28 18:54:11.098] << PING 0
2013-02-28T18:54:11+00:00 app[bot.1]: [2013/02/28 18:54:11.190] >> :niven.freenode.net PONG niven.freenode.net :0
2013-02-28T18:54:19+00:00 app[bot.1]: [2013/02/28 18:54:19.123] >> :[email protected] JOIN #arlingtonruby
2013-02-28T18:56:11+00:00 app[bot.1]: [2013/02/28 18:56:11.108] << PING 0
2013-02-28T18:56:11+00:00 app[bot.1]: [2013/02/28 18:56:11.200] >> :niven.freenode.net PONG niven.freenode.net :0
2013-02-28T18:56:26+00:00 heroku[bot.1]: Idling
2013-02-28T18:56:29+00:00 heroku[bot.1]: Stopping all processes with SIGTERM
2013-02-28T18:56:31+00:00 heroku[bot.1]: Process exited with status 143
2013-02-28T18:56:31+00:00 heroku[bot.1]: State changed from up to down
It seems that even though the bot received a ping, Heroku still thought it was idle for too long.
Per the Herkou docs on dyno idling
Apps that have scaled the number of web dynos (dynos running the web process type) so that only a single web dyno is running, will have that web dyno idled out after one hour of inactivity. When this happens, you’ll see the following in your logs:
2011-05-30T19:11:09+00:00 heroku[web.1]: Idling 2011-05-30T19:11:17+00:00 heroku[web.1]: Stopping process with SIGTERM
Normally, Heroku will spin the app back up when a request is made to it. However, in this case, when Heroku spins down the app, it leaves IRC. Thus it never receives a wake-up request.
Something else to consider regarding Heroku dynos:
If your app has only a single web dyno running, it will idle out - irrespective of the number of worker dynos. You have to have more than one web dyno to prevent idling.