Using one disk of a RAID1 volume as a "backup"

[Jens Baumann_2]

I have an idea, but I am not really sure whether it might work or not - so I would like some feedback on it.

I have a test machine, a DL380G3 with a 5302 smart array, four 72GB as a RAID5 volume and two 36GB disks as a RAID1 volume (system). OS is Windows 2000 server SP4.

What I have in mind is taking out one of the mirrored disks as a "backup" (say, in the case of a risky software install), then I do the installation with only one of the two mirrored disks. Later on let's say I find out that the install went wrong and I want to go back to the old status. So I down the server, take out the one disk which still was in the server, plug back in the "backup" disk and power on the server again.

Expectation: server accepts the old disk again and boots the server, then I can plug in the other disk of the mirror set and the controller mirrors the contents of the "backup" disk on it, after which the server is back to its old, pre-install, status.

Reality: it doesn't work. Here's what happens:

- I pull out the "backup" disk and power on the server. Everything OK, the controller announces one "failed" drive, but Windows still boots up - as expected.

- I down the server, pull out the disk which still is in the server and put back in the "backup" disk to the slot it was in. Server is powered on: array controller detects the logical drives, announces one failed disk (the one I pulled out), but does NOT boot from the backup disk I put back in (tries CD and network, then gives up).

- server is powered down. I pull out the backup disk and put in the other one, in its original slot. Same result: logical drives are recognized, server does not boot.

- all disk back in: logical drives recognized, server does not boot.

- only by re-creating the raid1 volume I can get the server to boot from the hard disks again.

What I *think* happens is that the controller "remembers" drives which failed - so when I put the failed drive in and try to boot from it (without the other, non-failed one, present), the server thinks "oh, I only have a failed drive. I cannot boot from that one". And by pulling out the second drive, too, I have two failed drives and no non-failed ones left.

Has anybody got something like this working? Is there some way to tell the array controller "hey, the drive in slot 1 isn't failed, accept it"?

[Leif Halvarsson_2 · Accepted solution]

Hi,
We are using the same hardware as you (with two mirrored,RAID 1, system disks). We have done several similar tests as you, removing one of the disks, doing some modifications on the other disk, removing it and booting on the backup disk. It has worked without problems (also after installing a different OS on one of the disks).

[Jens Baumann_2]

Have you also tried it with an additional RAID5 volume? I think this might be a crucial point, since that volume might also carry some information about the RAID configuration and it stays in all the time while the RAID1 disks are exchanged.

[Leif Halvarsson_2]

Hi,
We use two raid controllers on this box, the 5302 controller is only used for the OS. But I don't think it makes any difference (if there is no system files on the RAID5 set). The two RAID set should be independent of each other.

[Jens Baumann_2]

OK, will try again (and write down, step by step, what I am doing - if it fails again, I will post exactly what I have done, maybe you'll be able to find an error). Thanks for the info!

[Marino Meloni_1]

Hi,
I have also used this tip some time but with Netraid controller, in that case you can manualy force on-line the disk you want to use.
With SmartArray, you do not have this choise, and for what I think, he can try to boot from the only disk available in case of config mismatch (NVram says two disk failed, but one disk is marked online, as you removed it from the raid when it was on that condition) but the problem may arrise when you have other disks on that controller, in thatcase you do not have only two configuration that are mismatch (NVram and one disk) but you have also part of the configuration that match with some disks, so I suspect that the Controller keep the matching config as valid and then keep you disk (the one you pluged in) as failed.

Thanks to let us know further development of your tests

Marino

[vkaz]

Just for my own info, how do you re-create an array (RAID 1) without loosing the data? Just by putting 2 drives back (RAID 1) than going to ORCA, than delete logical drive, than create logical drive (same size, same RAID 1 level)?
Is there a way to recreate an array using a command line from DOS (maybe using SmartStart scripting tool)?
Regarding you problem. To me, if the bacup drive is not in Failed status than you should be able to boot. Please let us know after you verify this again.