How often should a RAID consistency check be done?

I have a LaCie 60TB RAID array. I've had it for a couple of years, but I am wondering how often I should do a "Consistency Check" on it. I don't know a whole lot about RAIDs. It's formatted HFS+, so I think this is a different than defragmenting. LaCie don't mention doing this as a routine in the section of the manual regarding the subject.

Solution 1:

A RAID consistency check will verify all of the blocks in the volume so there will be a significant hit to operational read/write performance. That means you want to run this test ideally during off hours.

That said, how often you run it will depend on its use. Typically, in an IT department with a typical file server and lots of reads and writes, a consistency check every week is advised. If this is for archival access - more reads than writes - a less frequent consistency check is fine. On the other side of that coin, if it's a file server with more writes, you'll want to check it more often.

Three file system type really plays no bearing on it because the consistency is for the RAID array to ensure that it's consistent across the whole volume.

Solution 2:

You should balance running the consistency check with your requirements for uptime for the RAID-array. Essentially running the consistency check makes the RAID-volume slower for a long period of time, so you want to avoid doing that when you need the full performance of the volume.

On the other hand, not running it often enough could result in errors going undetected for a longer period of time, eventually leading to a situation where you'll fail to be able to read the data even though it is on a RAID (assuming we're talking about RAID-5, RAID-6 or similar with redundancy). Running a consistency check earlier on could have detected the error and corrected it, perhaps kicking in a hot spare or having you eventually replace the drive.

This means that not running the consistency check often enough increases the risk of you having to take the whole volume down in order to restore a backup, which is often a quite time consuming task. So it's a balance.

Contrary to popular advice, the interval between consistency checks should not be chosen simply based on how many writes you have on the volumes or hold how the hard drives are. I.e. for example the argument that an archive that it is only written to very seldom and mostly endures reads will need only a less frequency consistency check is actually not a good idea. Similarly thinking that new hard drives require fewer consistency checks is also not a good idea.

Instead you need to take into account the actual size of the individual drives. The larger the drives, the higher the risk that errors occur undetected on parts of the drive that are not accessed. If you have very large individual drives, you want to run consistency checks very often. It is for this reason that RAID-5 is really not advisable with very large drive sizes.

Note that if your RAID consists of SSDs things are slightly different.