3 ms·
This is all great info. The only thing I shudder at is that there is one huge single point of failure. I've learned one thing when building huge dumb storage de
by post_break 11y ago
This is all great info. The only thing I shudder at is that there is one huge single point of failure. I've learned one thing when building huge dumb storage devices, build two and mirror them. I've got 32TB of storage mirrored so if one hits the fan I've got an exact copy.
- beachstartup 11y agoi would recommend you also do snapshots. mirroring will just replicate a bad fuckup. as for OP, i share your concerns - i'm skeptical of the reliability of consumer gear in an application like this, especially in the absence of an actual backup solution (maybe he doesn't care). he calls it a backup server, but it's actively serving files.
- pfarnsworth 11y agoYes, I agree, mirroring will help when a disk goes back, but it won't help if you accidentally delete data. This is the one that people tend to forget a lot, which is ironically the original reason for having backups.
- foobarian 11y agoI have two sets of drives on my home server and keep them mirrored using a nightly rsync. It's infrequent enough that I can recover from accidents.
- XorNot 11y agoThis is why people should use ZFS (or BTRFS) - snapshots. My home server runs a minutely, hourly, daily and monthly snapshot job. Cryptolocker could have a field day and the most likely scenario is at worst I'd lose a day of data - more likely maybe 15 minutes worth.
- post_break 11y agoI'd love to but my OS limits me from running ZFS.