Timeouts after server running for some time

[ypadlyak]

"upstream timed out (110: Connection timed out) while reading response header from upstream"

The Puma server stopped responding with fore-mentioned timeouts in Nginx logs. If I restart, it works correctly for some time, then it repeats. How do I debug this king of problems?

[pchaganti]

Hmm. Seeing the same thing. In our case, it seems to timeout for a bit, and then start responding again. The time varies between 1-5 min. This server not really under a lot of load either.

[ypadlyak]

Yes, the same! But it is processing the request or two and stops again

[ypadlyak]

I've now rebooted it and the log looks strange (these TERMs):
[2419] - Starting phased worker restart, phase: 5
[2419] + Changing to /home/ubuntu/tickets/current
[2419] - Stopping 3229 for phased upgrade...
[2419] - TERM sent to 3229...
[2419] - TERM sent to 3229...
[2419] - TERM sent to 3229...
[2419] - TERM sent to 3229...
[2419] - TERM sent to 3229...
[2419] - TERM sent to 3229...
[2419] - KILL sent to 3229...
[4812] + Gemfile in context: //20170419101552/Gemfile
I, [2017-04-19T10:19:18.119862 #4812] INFO -- sentry: ** [Raven] Raven 2.4.0 ready to catch errors
[2419] - Worker 1 (pid: 4812) booted, phase: 5
[2419] - Stopping 4356 for phased upgrade...
[2419] - TERM sent to 4356...
[4839] + Gemfile in context: //20170419101552/Gemfile
I, [2017-04-19T10:19:27.204147 #4839] INFO -- sentry: ** [Raven] Raven 2.4.0 ready to catch errors
[2419] - Worker 0 (pid: 4839) booted, phase: 5

[ypadlyak]

I have now tried this suggestion: rails/rails#27455 (comment) . I'll report if it happens again

[nateberkopec]

This particular issue looks like the above ^^^ If config.eager_load = true doesn't fix it for you, please open a new issue.