October 16, 2019
The mail servers are scheduled for migration to reverse proxied local containers during the night between Saturday and Sunday this weekend. An extensive testing phase has finally returned good results. There are 2 old mail servers and can be migrated to 1 or 2 new ones. This means that it could very well be possible for 100% availability to remain, even while migration is ongoing. However, this isn't sure yet, hence this has been scheduled during a time in which the mailers don't see much traffic. Hopefully by next week this mess will finally be cleaned up.
After the migration, this will also mean that the redundancy on both the mailers and the website will be upgraded to n+2 (previously n+1 and n+0 respectively), since all the edge servers will be forwarding all the services. Additionally, since the edge servers won't be doing a whole lot of work anymore after this, it might become possible to do caching there to mitigate the impact on network resources.
October 19, 2019
Migration has been successfully completed without downtime. Oyasumi.. *yawn*
October 31, 2019
2 occurences of a few seconds of downtime as the underlying webserver had to be moved to somewhere else for maintenance.
November 2, 2019
Scheduled rebuild of the edge server that's forwarding this one. Expected downtime of up to a few hours. The rebuild will be done at night (CET).
Migration has completed, we're back online. It took roughly 1 hour. Perhaps it wouldn't have been a bad idea to hang the webserver on another edge for the time being after all... Meh. No use in crying over spilled milk. It's done now.