I have a TrueNAS SCALE NAS installed on server-grade hardware with ECC RAM. I’d like to back up to another ZFS storage box (possibly another TrueNAS SCALE appliance, but I might just use debian or something else).
I’d like to build a very low power and low-heat NAS with only 2-3 spinning hard drives in it, which I’d place inside my gun safe. Obviously a gun safe isn’t the ideal place to store a running computer by way of heat, however I did run a J1900 SOC board in a small enclosure with a couple drive inside for a week or so as a test, and the CPU never realy got that hot. The setup also has the added bonus of preventing rust from accummulating on the guns inside the safe.
I’m curious as to what the implications are of using non-ECC memory as a backup target, and if there is any way to mitigate them.
I run my weekly replication and plug the drives and they remain cold on top of the case. All that could happen to the data gets checked by a scrub. You just need drives for backup, not a running system. Unless you’re using Ceph in which case you need an entire additional cluster running (but that’s a “me” problem and offtopic here).
If you just want something for backup, get hotswap bays or external disks with enclossure and just lock the drives into your safe. That way you only need one server. Separate bare-metal backup server isn’t necessary unless you want to also do other things with that machine.
I’m of the mind that at least one of each system’s backups should run completely automatically without any action being required by any human. Reason being is that backup tasks that require even a small amount of physical effort tend to end up being neglected at least some of the time; I am certainly guilty of that.
run a bunch of memtest / stressapptest when the system is new.
perhaps repeat that intermittently to check degradation.
don’t use caches in memory (ZFS is not the best option for that AFAIK)
limit the runtime of the server to guard against bitflips (power off & use wake-on-lan in the backup script?)
verify the integrity of your backups (scrub?)
verify integrity of backup against source files (e.g. a script that compares checksums on both ends - I don’t know if ZFS can do this with replication natively?)
If you deal with an FS with checksumming (ZFS/BTRFS/…), once the data is on disc it should stay fine. The concern would be during transfer/writing. So I think you should be safe if you compare checksums on both ends after the backup.