3

We have a Lenovo RD430 with a RAID 500 controller.

Installed in the server are 6x 4TB drives configured as 2x RAID0 arrays of 3x 4TB each, however, we also have another 3x identical 4TB drives and would like to be able to rotate one of the arrays on a weekly basis (to allow a set to be taken offsite).

When I try (powering down &) rotating one of the arrays with the spare drives I get an error saying that one of the VD's is missing and that continuing will remove it from the configuration.

I'm assuming this is because some part of the RAID configuration is written to the disks themselves and not just stored on the RAID controller? I guess in my naivety I just assumed I could use the Lenovo easy startup software to replicate the RAID configuration on the spare disks and then swap them out at will.

So firstly, Is this even possible? And if so how would I go about it?

EDIT: I probably should have said earlier that this whole server is in fact a backup server - we have an entirely separate RD430 running RAID10 acting as the main file server and the file server backs up to this server nightly. I simply wanted to hedge my bets by introducing offsite. The only reason the disks are in RAID0 is convenience (1x 12TB instead of 3x 4TB).

UPDATE: In response to comments... The primary data is stored on a RAID10 array which is rsync'd nightly to a second server (RAID0 array) and then archived off to third array (again RAID0). It is this third copy (forth if you count the fact the primary is RAID10) that I am looking to rotate. Additionally, the drives are labeled (SET x DISK x of x) and when out of the server will be stored in a Peli case.

What I would like to know is how to configure the drives such that the rotation is possible?

Alex.K.
  • 167
  • 2
  • 2
  • 8
JimmyJ
  • 157
  • 3
  • 2
    First of all: RAID isn't backup! If you want to have your data offsite, you should use some backup procedure to write thata data to an external disk or tape or whatever. Secondly: On every disk in a RAID there is written some meta-information (depending on the controller used). So you can't just swap disks, you have to initialize them for this specific RAID on this specific controller (well, mostly, some are compatible with others). – Lenniey Apr 23 '15 at 11:54
  • Nowhere does he say he's taking them off-site as backups. I'm not saying he's not, but I'm wiling to give him the benefit of the doubt that he has a valid reason for doing so.. – GregL Apr 23 '15 at 12:00
  • _"to allow a set to be taken offsite"_ Seems it. – ewwhite Apr 23 '15 at 12:01
  • 1
    Ok, *now* the "RAID isn't backups" barrage can begin. – GregL Apr 23 '15 at 12:18
  • 1
    @GregL That only applies to situations where a single RAID array is treated as a "backup" of itself, though. Having a whole separate copy of your data is a perfectly good backup, even if it happens to be stored on a RAID array. (It just needs to be *separate* from the data that it's a backup of.) –  Apr 23 '15 at 18:34
  • 1
    @duskwuff, We're starting to get a little off-topic here, but I'd agree with you, *if* the seperate copy was stored on RAID5 or 6. I don't think storing it on a RAID0 array is terribly sound; you loose one disk, you lost it all. Unless of course you're doing Disk-Disk-Tape and the RAID0 copy is transient. – GregL Apr 23 '15 at 19:11

1 Answers1

9

On modern hardware RAID controllers, RAID metadata is written to the disks, so they are somewhat portable, with restrictions on drive ordering; they have to be moved together.

However, this is the wrong way to handle backups.

Intentionally failing (groups of) disks and forcing a degraded array is fraught and a bad approach to accomplishing offsite backup.

In particular, doing this with a RAID0 set would require you to turn the server off or offline the array from the controller level as you make the physical switch.

ewwhite
  • 194,921
  • 91
  • 434
  • 799