I have another container running the proxmox backup client that does a filesystem dump daily… Not bad, but I think postgres may not recover from that…
Right now 2 weeks of daily replicated snapshots are probably as good as I’m going to get…
Comment on Docker, Truenas, and self-hosting...
frongt@lemmy.zip 5 days ago
I use a volume bind mount for the container’s config directory, then back that up. And the docker-compose file.
ShredderFeeder@shredderfood.net 5 days ago Jason2357@lemmy.ca 4 days ago
Works fine… Most of the time. As long as the file gets saved at an instant in time (e.g. zfs snapshot), postgres sees it like a power failure and restores. You want to have several recent backups if you do this because it will occasionally fail.
- ShredderFeeder@shredderfood.net 4 days ago
Saw an article that says proxmox backup client can be set up to be ZFS aware…will have to do more reading… (for backups)
Replication is always a point in time, and I keep two weeks worth of dailies.
frongt@lemmy.zip 4 days ago
Test it, find out.
- ShredderFeeder@shredderfood.net 4 days ago
Well that worked… Amaze! I copied the lemmy directory to my user directory, made the necessary changes in docker-compose, and boom, it came up. If I were to script the changes using sed, I could even have a full DR failover.
- ShredderFeeder@shredderfood.net 4 days ago
Not a bad idea…I’ll have to make some network changes in the docker compose and build a new macvlan for it…but it should be easily addressable by IP address. I did to a test startup of my mastodon DB and that did seem to work…
And I should probably block outbound traffic just to be sure it’s not trying to update the fediverse…
Cool, now I know what I’m playing with tonight. :)
Jason2357@lemmy.ca 4 days ago
Yeah, I bind mount the config inside a config directory just below the compose file, and the state (databases, files, etc) to a seperate storage location that makes sense. The compose and config gets added to git and pushed to a forge, while the storage directories are backed up like normal (in my case, zfs snapshots pushed to another computer and then restic backs up those cold files to b2. The volumes are ephemeral.