One drive failed in my RAID 1 array, am I safe to replace it without losing data?

Question

We have a server with three disks. Two of them are in RAID 1 defined in BIOS. One of these is showing an error at the bootup screen with the corresponding disk serial code in red.

I intend to replace it but I'm very wary of doing something wrong.

1. Is it just a matter of turning the server off, replacing the disk, and turning it on then waiting for the disks to synchronize?

2. We haven't bought the replacement disk yet. The defective one is a 1.8 TB (2 TB?) Hitachi hard disk. Is there a way to see the RPM (7200 or 5400) in order to buy the correct one without having to turn the server off? (I have no problem in turning it off, it would just be more convenient that way).

3. I'm assuming a missing disk won't affect the RAID 1 config settings (the server will only lose its mirror disk until a replacement is put in the slot) but out of curiosity I'd like to be sure. Will the RAID 1 settings be kept unchanged?

Solution

1. Check whether your server supports hot plug - this is the easiest method: pull out the failed HDD, plug in the new one. Without hot plug, you need to shut down the server, replace the failed disk (check the IDs at this point), power up, and possibly tell the BIOS which disk is to be mirrored where (mind the IDs). Rebuild/remirror should start.

2. You should replace the disk with a type that matches the remaining disk. Mixing different speeds is not a very good idea. If the disks are not vendor-specific you should have the exact type documented. Additionally, the RAID software should show the exact disk types. Make sure the new disk is at least the size of the remaining one.

3. You should only replace the disk once replacement has arrived. Depending on the actual problem, the defective drive might still provide some support for the other disk. Additionally, with hot-plug slots, you shouldn't leave a slot open for an extended time (the cooling might fail).

Comments

[HBruijn]

That mainly depends on whether or not you have a “real” hardware raid controller of a “fake” one that depends largely on the software driver. 1) usually server hardware should even allow hot swap of a failed drive 2) depends on ”how much dead” drive is, but typically you can determine the properties of the remaining drive in the array. 3) yes that is the theory

[Piovezan]

It's probably a hardware controller; I have pulled the non-raid disk with the server working and put it back without significant consequences to it. But if there are other ways I can check that, I'll be glad to hear.

[symcbean]

"without significant consequences" - how would you know if you had shortened its life or corrupted data you can't verify? NEVER DO THIS UNLESS YOU KNOW YOUR HARDWARE AND OS SUPPORT HOST SWAP. Whether its a hardware or software controller is irrelevant to a rebuild of the array on the original host hardware. But don't plug in / unplug stuff while the machine is on power.

[Piovezan]

I know; the unplug was accidental (the slot appeared to be empty).