Take the 2-minute tour ×
Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

I'm reading the FreeNAS userguide and came across the statement:

Note that hardware RAID configured as JBOD may still detach disks that do not respond in time; and as such may require TLER/CCTL/ERC-enabled disks to prevent drive dropouts.

I'm using a '3Ware 9550SX-8LP RAID Controller' and see quite a few stories of people successfully running raid5 on 7.2k consumer SATA drives without issue.

Are detached disks only a theoretical problem, or should I expect this to be a common occurrence?

Update:

Ok, so drives dropping out is a real problem to be concerned with.

My controller is indeed on the FreeNAS compatibility list and i'm using some cheap Toshiba OEM 500g 7.2k SATA drives. If i'm using HW Raid1 and a drive is disconnected due to seek timeout, will it pop back into raid and sync up the data without human intervention? I'm doubtful that there is a universal answer - feels like this would be controller dependent, but I haven't seen info on this for my 3ware controller, yet.

share|improve this question

3 Answers 3

up vote 2 down vote accepted

Sorry, but both of the existing answers are completely wrong; and neither address the issues you quoted from the FreeNAS website. The problems with SATA drives have nothing to do with performance or normal operation.

The problem with "consumer" SATA drives is that when they encounter a problem they sit and try to recover from it. This recovery process is generally wanted when you only have a single drive (like a typical PC), as a complete failure means the storage is lost (hope you had backups).

What happens in RAID arrays however is that you are now X times more likely to encounter a situation where a sector has a URE (unrecoverable read error, the most common critical data error); where X is the number of drives in the array. When that one drive stops and tries to recover the data for 30 seconds to 2 minutes (typical timeouts) the entire array has to wait for that drive (some controller can drop the drive after a certain amount of time, but this isn't as common as you'd think). It's pretty typical for an OS to issue several reads at a time, waiting 30s or more for each read would bring a server to it's knees.

So the solution should be obvious, a mechanism to limit the amount of time a drive will try to recover from an error. SCSI drive implement a command specifically for this. SATA do not however, so the drive has to be programmed from the get go with a short timeout. The three vendors call this "TLER", "CCTL", and "ERC". Drives with this "capability" have short timeouts, so UREs are reported back to the RAID subsystems quickly. The RAID subsystem can then figure out what to do about the failure (usually drop the drive from the array and report the failure so the drive can be replaced).

share|improve this answer
    
Fantastic information. Thank you. –  xelco52 Dec 11 '12 at 20:22

This depends on which drives you get. Since 'green' drives and power efficiency have become all the rage, they are what is primarily available to consumers. The downside is that to achieve this power efficiency, they aren't as performant as their beefier cousins. Often, they take a longer time to spin up, which can cause them to time out - Nothing is wrong with the drives, they just didn't pick up the phone quite fast enough when the controller calls.

Some controllers have a configuration option to extend the timeout so that green drives can spin up before they would otherwise get removed from the RAID array. The other option is to use drives targeted at performance rather than power efficiency, such as Western Digital's "Caviar Black" drives. Most manufacturers will have a list of NAS-friendly or 'Enterprise' drives which should work just fine with a raid controller.

share|improve this answer

If you want to be sure to have no issues you should be looking at FreeBSD raid controller support in FreeNAS docs and the drive compatibility list available for the 3ware 9550SX-8LP. Using a RAID controller in FreeNAS is dependent on the FreeBSD support for that card, moreso than the disk brand itself.

Disk manufacturers have conflicting stances on RAID support for their disks. Western Digital would say you need to be using the RE series if you want to do RAID with SATA 7.2K even though many people are successful with the black series. If there is a problem with the array an RE series should recover/rebuild quicker. Seagate will say that all of their drives are available for RAID however they too have the Constellation Enterprise series.

And as side note, one of the beauties in using FreeNAS is that using ZFS with built in disk controllers can yield excellent performance using a drive like WD Black, presuming you have installed enough RAM for it. The FreeNAS docs will recommend 8 to 12 GB and they have some detail on how RAIDZ ( the equivalent of RAID 5 ) avoids the write hole.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.