3 ms·
"Well let's put aside the disk failure issue. Modern 4-drive NAS systems can tolerate a lost drive and alert you promptly to replace it. You'd have to have pret
by rsync 11y ago
"Well let's put aside the disk failure issue. Modern 4-drive NAS systems can tolerate a lost drive and alert you promptly to replace it. You'd have to have pretty bad luck to lose more than one drive at the exact same time."
Very interesting and well thought out posting, but the above quotation represents a very, very naive understanding of how these arrays work with large, multi-TB hard drives.
In fact, the reality is exactly backwards to what he has written here: with multiple terabytes of data on the array, a single drive failure results in a long, intensive rebuild process that can serve to hasten the failure of the remaining drives.
I am not anti-NAS - I use them myself for critical data - but with 3 and 4TB hard drive, I would only use raidz3 (or equivalent) at this point (and preferably with 12 or fewer drives in the array).
- ValentineC 11y ago> In fact, the reality is exactly backwards to what he has written here: with multiple terabytes of data on the array, a single drive failure results in a long, intensive rebuild process that can serve to hasten the failure of the remaining drives. Are you talking about Unrecoverable Read Errors? [1] [1] https://news.ycombinator.com/item?id=8306499 https://news.ycombinator.com/item?id=8306499
- rsync 11y agoI'm talking about any problem that a hard drive could have. One drive dies, and the response to that drive dying is to go beat the shit out of the remaining drives, and given their large size, that beating goes on for days (possibly). Yes, an unrecoverable read can be one failure mode, although not necessarily the scariest - you can often tell your raid hardware to ignore UREs and just rebuild anyway, hoping that the missing bits aren't crucial (which in the case of photo storage isn't that bad of a bet...).
- slantyyz 11y agoIs a RAID really necessary though? I know the RAID units today are pretty good, but I've had some failures in the past with older systems, and I just stay away from them, because I have less than 1TB of data that I can't afford to lose, and the mirroring doesn't really add a lot of value for my needs. To keep things simple, I just make redundant copies by pointing Crashplan to multiple targets - the cloud, a single-drive NAS, and an internal hard drive in my desktop dedicated for backups. I also burn periodic copies of my most critical data to Blu-Rays. Super critical data gets burned to M-Disc Blu-Rays.
- prapam2 11y agoAre Blue-Rays/M-Disc reliable for long term storage?
- slantyyz 11y agoI guess that's unproven. M-Disc is supposed to be very reliable. In any case, it's moot for me. I am regularly rewriting the -entire data set-. I am not doing incrementals, so it is pretty unlikely that I'll ever use my oldest burned discs for recovery.
- lh7777 11y agoI think the key point here is raidz3 -- exactly why I've been running a FreeNAS server for the past 4 years. That and backing it up to Crashplan in case of a multi-drive failure.