We set up real-time back-up for the user database (as opposed to hourly backup, which we already had) and we discovered a bug in our deployment script in the process, which meant the services did not come back up the same in production as they had in testing. This caused the deployment to triplepat.com to fail, which halted the deployment with a non-working triplepat.com, while all the mirrors kept working fine.

The bug is fixed, the user-service was down briefly, and the checkin-service kept chugging along just fine. As always, we need a completely red column to run the risk of losing user data, so everything is okay.

triplepat.com and triplepat.net recovered.

triplepat.net went down.

triplepat.com went down.