RAIDZ1 Risks Practical Home User Views

[rliver73]

Hi all,

First post. I'm enjoying FreeNAS and feel comfortable with my 3x 1TB RAIDZ1 array. I've really played about with FreeNAS on a test machine, a Dell R320 ; pulling drives hot and cold, replacing etc. to practice how to recover from a degraded array; which went smoothly. I've read a lot about unrecoverable rebuild read errors and read users calculations on how likely this is.

My question is what is the reality of this and what happens if one gets one. My FreeNAS usage is for home. I've photos, movies, music etc. on my FreeNAS.

If I lost all my data I'd be upset.

If I lost one or two of say my honeymoon photos I'd be slightly miffed.

If I lost a few movies / music then that would just be a hassle to rip the movies again.

I've enable SMART monitoring and in the event of it reporting a ton or errors I'd network copy the data onto a different computer before going ahead with physically replacing the drive.

As an aside I've accepted that as NAS is not a backup and will soon sign up to BackBlaze B2 for a proper external backup solution.

 

[danb35]

My question is what is the reality of this

Yes, it can happen. I don't think it's as likely as some of the "gloom and doom" types would have you believe, and it wouldn't be a great concern for me with a 3 x 1 TB array, but single parity is still a bit less than I'd like to have.

what happens if one gets one.

It depends on where, but it's unlikely that a single URE would result in complete loss of the pool (as it would with traditional RAID5). If there's a read error somewhere in the data, that data will be corrupted--you could lose some files. Maybe they'd be movies, maybe they'd be honeymoon photos, maybe both, there's no way to say. If the URE's somewhere in the metadata, there's the potential for a broader impact--but to mitigate that, ZFS stores multiple copies of all metadata, and all of it is checksummed. If a metadata block fails to read, or reads incorrectly, there's at least one other copy available.

Now, if a second disk outright fails, your pool is gone. Much is made of the "highly stressful" resilver operation and the chance that it could cause a marginal disk to fail--and I guess it's possible, but it's the same stress you'd be putting on a disk with a scrub, which is a routine maintenance operation for your pool. Thus, I don't believe the risk of a second disk failure is any greater during the resilver than at any other time--but it can, of course, still happen.

 

[rliver73]

Thanks, thanks really helpful. I'll sleep better knowing that I'm in a significantly better place than I was before FreeNAS and will continue with my BackBlaze B2 additions.