The default of 120 seconds is far more than most of our apps need, resulting in sitting around mid-deployment waiting for the old unicorns to die. The tests are dark and full of mocks, but no more so than those that were there before.
Changes - Minor change for unicorn_rails - no longer passing through the -P argument to set the PIDFILE as this is not supported by unicorn_rails. User much manually match up PIDFILE argument to unicornherder and the pidfile config option in configuration file used by unicorn_rails
Unicorn_rails does not use the -P flag to specify the location of its pid file (it uses the accompanying configuration file). Unfortunately, including the -P flag was causing `ENV['RAILS_RELATIVE_URL_ROOT']` to be overwritten with the path to the unicorn pid (which is nonsensical).
Since we have now moved this repository from Alphagov to GDS Operations, we should change the old URL to the new one, despite 301s existing, and ensure we conform to our own guidelines for open-source projects (published at http://gds-operations.github.io/guidelines/)
Unicorn temporarily pushes the pidfile to pidfile + '.oldbin', without checking for the existence of this file, it is very easy for unicornherder to miss the boat on an otherwise uneventful hot reload
this is a commit i accidentally omitted which breaks the -b and -g flags