RAID 1 failure

[VectraMan]

My desktop has two 1TB Seagate drives connected to an Intel IHC7R onboard RAID controller as RAID 1 (i.e. mirrored). But now on startup one drive shows as "Error", and the RAID status as "degraded".

I'm not sure how to procede. Is this likely to mean one drive is dead, or did it just notice one parity error somewhere and decide to break the RAID? I suppose I could try disconnecting the good drive, and seeing if I can boot off the dodgy one, and run some kind of checker on it. If that works, the options on the controller can break the RAID and rebuild the RAID - does it do this retaining all the data?

I guess with hindsight buying two identical drives was a bit silly, as maybe the good one is also about to fail.

[Sysman]

You need to be very careful with diagnosis here. My very first attempt with RAID 1 ended up with the out of date copy overwriting the up to date copy That was actually a software RAID solution and on each mount it incremented a generation number, so repeatedly mounting the bad copy meant it ended up with a higher generation number than the good copy...

Take backups of the good copy first.

I guess with hindsight buying two identical drives was a bit silly, as maybe the good one is also about to fail.

It's likely that they came from the same batch. Does your controller support mixed disks?

[Sockpuppet]

Go into the RAID Bios (assuming its hardware raid) and see which drive has failed. Buy drive of same spec (search for model number) and put it into the drive and select rebuild drive.

I only generally use 3ware so not sure how the Intel stuff compares but should identify port number of the affected drive.

[VectraMan]

I found the manual on the Dell site, and that does suggest that any drive can be used as long as it's SATA. I'll probably just try to rebuild it with the existing drive first, and see if it goes wrong again, assuming it hasn't died completely.

Now doing a monster backup to an external drive, just in case the other two backup schemes don't work.

[VectraMan]

All the options in the BIOS say all the data will be destroyed! I have it backed up, but obviously I'd prefer not to do that.

However, having disabled the dodgy drive in the BIOS and re-enabling it, the status has changed to "rebuilding" (array will be rebuilt in the Operating System). Back into Windows, the hard disk light is now permanently on, so I guess it's working and the driver is rebuilding the mirror.

Maybe it's me, but with this sort of thing I want to know exactly what it's doing and why and find anything that sorts itself out deeply scary.

[doodab]

Take a backup

Verify the backup

Break the mirror

Scan the defective drive and replace if necessary

Resilver the mirror

If it works you are golden, if not you need to replace the dead drive, recreate the mirror and restore your backup

The resilvering involves copying all the data from the good drive to the bad one, it will take a while.

[VectraMan]

Breaking the mirror deletes all the data, at least according to the warning. I guess it could be lying (I can't see why it would do that for RAID1; RAID0 maybe).

It says normal again now. I'll see if it happens again.

[stek]

It's all a bit redundant anyway (or non-redundant) since you have a SPOF anyway - only one controller so I'd just enjoy the 2TB's....

In Solaris we'd do metareplace or metadetach then metattach, under windows you've no idea what it's doing.

AIX LVM is even easier, who says Unix is hard?

[Sysman]

Breaking the mirror deletes all the data, at least according to the warning. I guess it could be lying (I can't see why it would do that for RAID1; RAID0 maybe).

It says normal again now. I'll see if it happens again.

Hardware engineers think "differently". It doesn't always translate easily to what software people want, or expect.

It's all a bit redundant anyway (or non-redundant) since you have a SPOF anyway - only one controller so I'd just enjoy the 2TB's....

In Solaris we'd do metareplace or metadetach then metattach, under windows you've no idea what it's doing.

AIX LVM is even easier, who says Unix is hard?

I have successfully run test machines with RAID 1 and only one controller per 24 disks. With that ratio RAID 1 was definitely worth it.

I prefer something controlled by software, and software that I can control. You can also write scripts to minimise the damage caused by finger trouble or brain fade (if the hardware offers a CLI you can write scripts for that, but they can get ugly).

[escapeUK]

On a recent new build, I tried RAID1 with two new 1 TB Drives. I couldnt believe how god damn slow it was. I think it was something like RAID 1 = 55 Mb a second, a single drive, 110 Mb a second, RAID 0 180 Mb.

Needless to say I wiped and reinstalled with RAID 0.