BUILD Sanity check for new NAS

[no_connection]

Data planning is fun =)
Spending is not =(

Is a 3 4TB drive Z1 a bad option for decent performance and stability?

 

[ZFS Noob]

Data planning is fun =)
Spending is not =(

Is a 3 4TB drive Z1 a bad option for decent performance and stability?

Yes, it's a bad option. Those drives are too big for them to successfully rebuild the RAIDZ1 pool after a drive fails without data loss. Go RAIDZ2 or mirrors instead.

 

[warri]

Drive size does not influence ZFS' ability to rebuild a pool. It will just increase the time needed for the rebuild and hence increase the chance that a second drive fails during the resilver operation. If you perform regular backups a 3-drive RAID-Z1 is fine.

 

[ZFS Noob]

He asked for a "sanity check." Resilvering an 8TB RAIDZ1 mean you're looking at a 71% chance of data loss due to an unrecoverable read error (assuming lots of stuff.) Apparently that doesn't kill the pool, and the data you lose can be readily identified by ZFS, which is a huge plus.

And yes, if you're performing regular backups you'll be fine. Of course, regular backups + RAID0 is fine too. Doesn't mean it'll pass a "sanity check" though...

 

[no_connection]

A Z1 would obviously not guarantee data integrity during a rebuild or in a degraded state. A Z2 would.
But the chance of a disk crashing completely is not too important to me(and you usually get some sort of warning if it's mechanical). Knowing my data is not corrupting itself is.
And assuming a second disk does not crash during rebuild, I would be notified of any corrupt data during rebuild, and it will complete?
A 6 drive Z2 might still be open but it is a lot harder to upgrade unless I replace all 6 drives with larger, not to mention having higher initial cost.

I might as well throw out a non freenas question related to backup:

Is there any practical system for check-summing backups just so I know if any backup files have silently been corrupted during storage?
Preferably a "index entire desk then tell me next time if anything is corrupt". The benefit would be that it reads the entire drive preforming a "resilvering" of weak areas that should be done once in a while anyway for disk being in storage.

 

[jgreco]

People are getting a little extreme with the anti-Z1 stuff; Z1 is perfectly capable of not losing your data as long as the remainder of the disks do not err. The risks outlined in the oft-cited "RAID is dead" paper are based on assumptions not quite suitable to FreeNAS and the definition of "failure" is very pessimistic.

There are a variety of ways to checksum files and trees. For single files, md5 is fast and reliable.