I will be testing an offsite backup strategy over the coming days where docker volumes are moved to storage offsite so in the case of total server suicide a new server can be rebuilt and totally recreated. This will incur some downtime during testing. I will comment in this thread when I intend to tun a test so you at least know when the site will be down for 15 minutes or so during a test.
This is so that a separate company to the company hosting the vps has a copy of data so that everything can be fully recreated even in a disaster where one or the other company removes or cannot retrieve data related to this instance.
Pictures which is many sizes of magnitude larger cannot be backed up at this time so if there was a major issue, while text would be restored images would be lost. This decision makes the instance a lot cheaper at this time. At some future point this could be changed so a copy of all pictures are kept but I cannot pay for 2 separate object stores per month at this time. It is also not critical data but more a nice to have if we grow large enough.
Kind regards Pete
Testing is done, Backup takes approx 15 minutes. Crontab is set for every day at 3AM EDT (Canada daylight savings). Server will be down every day during this time at around 3AM for 15 minutes as the docker containers are fully stopped and a volume backup of each container is made.
Example script: docker run --rm -v lemmy_caddy_config:/from:ro alpine ash -c ‘tar -czvf - -C /from .’ > /Lem/LemmyDeploy/Backups/$(date +%F)_caddy_config.tar.gz
There are 4 volumes - postgres, caddy data, caddy config and pictrs
Backups moved offsite during this time.