RAIDz1 - Failure points

[nyquist]

Hey!

I got confused with some RAIDz1 statements in the FreeNAS v8.3.1 User Guide.

The manual says nothing about RAIDz1 array failures (at least in the RAID Overview - Page 15), but it clarifies how RAIDz2 in theory works.

Some FreeNAS guide quotes:

RAIDZ1: ZFS software solution that is equivalent to RAID5. Its advantage over RAID 5 is that it
avoids the write-hole and does not require any special hardware, meaning it can be used on commodity
disks. If your FreeNAS® system will be used for steady writes, RAIDZ is a poor choice due to the
slow write speed. (until here, nothing is mentioned about failure points).
RAIDZ2: double-parity ZFS software solution that is similar to RAID-6. Its advantage over RAID 5 is
that it also avoids the write-hole and does not require any special hardware, meaning it can be used on
commodity disks. RAIDZ2 allows you to lose one drive without any degradation as it basically
becomes a RAIDZ1 until you replace the failed drive and restripe. At this time, RAIDZ2 on FreeBSD
is slower than RAIDZ1.

As per what I found on Google, RAIDz1 supports up to 1 disk to go down without data loss (in theory), and RAIDz2 supports up to 2 disks to go down within the array without data loss.

Which one is wrong?
Could any of you confirm if I'm on the right path to work with a 3 disks RAIDz1 array without possibilities to loose data if a disk dies?

Thanks in advance!

 

[cyberjock]

Both are correct. It's technicalities in the wording.

RAIDZ1 protects you from 1 disk failure, and RAIDZ2 protects you from 2 disk failures.. in a perfect world. The truth is that drives are getting more and more error-prone and so a failure of a single disk in a RAIDZ1 is likely to result in loss of some data to some extent. Big picture, with a RAIDZ2 and one disk being resilvered you still have the ability to recover from small errors that statistically DO exist on multi-TB arrays. With a RAIDZ1, you are screwed. How much you are screwed depends on where the corruption is and what bits are wrong. You could have anything from a single file corrupted to a zpool that is unmountable(aka loss of all data).

Here's a slashdot article from 2008 explaining that RAID5 basically "died" in 2009 in terms of reliability.

 

[gpsguy]

Neither.

Which one is wrong?

Let me start with this premise - "RAIDz1 supports up to 1 disk to go down without data loss (in theory), and RAIDz2 supports up to 2 disks to go down within the array without data loss."

If you have RAIDz1 and a disk fails (or RAIDz2 and two disks fail), the array will continue to function, but it will be running in a "degraded" mode. You'd probably experience bad performance, until the drives are replaced and resilvered.

The other statement "RAIDZ2 allows you to lose one drive without any degradation" should be self-explanatory. The array will continue to function, but won't suffer from the condition described above.

That being said, as we now see drives sizes in the 3-4Tb range, I'd suggest RAIDz2. Even if you only have a single drive failure, the resilvering process (after replacing a drive) puts a heavy load on your system. Users have been known to loose another drive, while the resilvering process is taking place.

Having RAID is not a substitute for good backups. Stuff happens!

ps, thanks for reading the documentation, etc. before posting your message. So many users, jump right in and put their data in jeopardy (or suffer loss), because they didn't read the manuals, etc. first.

 

[nyquist]

Ok, thank you both for the quick responses.

It was clear to me that the fact of loosing a hard drive at any time and loosing a second one during the replacement procedure was possible, but I was not aware this was so common.

Since I'm preparing this system to serve as a virtual environment storage (implementing iSCSI), I was trying to get any possible additional latency factors off my system, so I was trying to avoid RAIDz2.

Beyond that, I think I will be getting a few more hard drives in order to build a RAIDz2 configuration instead.. just because I would like to keep my brain's calm level as it is right now :)

ps, I've just finished with cyberjoks' slideshow about zpools, vDevs, etc.. Nice one and thanks!