Heroku - dyno gives error H12-503 for no clear reason

Question

I have a node.js application running on heroku with 2 web dynos. I use express as web server. After running a few hours, one dyno or both of them stops responding and I see in the log messages like this:

013-11-24T14:14:48.697766+00:00 heroku[router]: at=error code=H12 desc="Request timeout" method=GET path=... host=... fwd=... dyno=web.2 connect=7ms service=30003ms status=503 bytes=0

Restarting the application solves the issue, but it will probably happen again. There was no peak of activity before the crash, and there is no hint in the logs that something bad is going on. I have newrelic for this app, but there's no hint there what causes the problem.

What would be the recommended course of action to fix this situation? Is it possible/advisable to have heroku restart dynos that stop responding?

Answer 1

You seem to have a long running process that is taking more than 30 seconds (Heroku's limit) to return any data to the client.

Here are some options for dealing with it.