I have been testing the Openshift Online platform for several months now (using the free 3 gears option). I setup several python applications and also the very basic python application (just using the Openshift python template) and outside monitoring, which also kept the applications alive (= non idling).
My findings are that the applications are restarted every cca 14 days. And basically not available for cca 5-20 mins. In order to have 0 downtime the HA scalable setup is advised.
This is typically declared as 2+ gears for application server (e.g. python ) with HA proxy setup + 1 gear for the database (PostgreSQL in my case).
I wonder whether the DB gear would not be also restarted from time to time as it is in non-scaled application case?
Uptime Logs for python application:
Latest Events (up, down, start, pause) Export Logs
Event Date-Time Reason Duration
Up 04-07-2014 19:03:10 OK 53 hrs, 11 mins
Down 04-07-2014 18:54:48 --- 0 hrs, 8 mins
Up 04-07-2014 18:48:51 OK 0 hrs, 5 mins
Down 04-07-2014 18:45:29 --- 0 hrs, 3 mins
Up 02-07-2014 22:52:56 OK 43 hrs, 52 mins
Down 02-07-2014 22:32:38 --- 0 hrs, 20 mins
Up 23-06-2014 11:41:43 OK 226 hrs, 50 mins
Down 23-06-2014 11:27:05 --- 0 hrs, 14 mins
Up 03-06-2014 04:00:05 OK 487 hrs, 27 mins
Down 03-06-2014 03:50:22 --- 0 hrs, 9 mins
Currently our database gears aren't scalable. This is something that will be resolved in the near future. However for now, its best to use a DBaaS as a work around. Thankfully we have several options on marketplace.openshift.com and you can easily tie those into your existing applications.