Farjad Adamjee <[email protected]> wrote: > Hello, > > I am not sure if this is the correct place for this, I have googled > around to see if anyone else has encountered such an issue, but I > did not find anything.
This is the correct place :) Plain-text email is the only way I handle support for ease-of-archival and distribution. > I am having this issue where after a period of time of no use (on > beta [production] environment), maybe 12 hours or so, the unicorn > process sleeps. When I go and re-request the page, it returns blank. This sounds like a database (or any other persistent connection backend) connection expiring due to an idle timeout. This could also be a firewall/router timeout problem between the server running unicorn and the database host. Tony experienced this last year: http://bogomips.org/unicorn-public/m/cakm1spnrses6h6byk6bo9djwa8wvyv6hj-reahopruybvff...@mail.gmail.com You can confirm by running lsof on a worker processes to look for open sockets. In case the connection loss isn't noticed by the OS, use strace (or any other similar syscall tracer for your OS) instead and watch connection activity. With strace (or similar) knock yourself down to one worker process via SIGTTOU to the master to ensure you're always tracing the correct one. > I am using Apache proxy to Unicorn. By the way, this is not recommended for slow clients. nginx (Open Source edition) is currently the only supported proxy. > # Using this method we get 0 downtime deploys. > defined?(ActiveRecord::Base) and > ActiveRecord::Base.connection.disconnect! > defined?(OathKeeper) and OathKeeper.disconnect! I suggest disabling any idle timeout you might have, enabling TCP keepalive and enabling reconnects (see AR and your database driver documentation). Worst case (firewall/router problem you can't fix via configuration), is to send a request to unicorn every few minutes/hours to keep things going. Perhaps OathKeeper has the same problem, too... I've never heard of it until now.
