RAID 5 Reported Wrong Drive Failed

[lifegard2]

I running a RAID 5 on NT4 w/ 3 IBM 10k 18GB drives and Adaptec 2100S controller. Drives are at SCSI ID's 3, 4, & 5.

Array failed and reported ID5 was to blame. Replaced ID5 with like drive and attempted to rebuild, got to about 2% on the rebuild and the new ID5 was reporting as failed.

Took all drives offline and placed on regular SCSI card to run IBM drive fitness. Turns out ID3 drive has bad sectors. When trying to run the corrupted sector repair util, it hangs on one sector. I gave up on it after 45 min. Is it normal for it to take forever at this point?

Anywho, odd thing is, I put the server back together and still seems to work OK even though I couldn't fix the bad sector. Thought I would try some fancy footwork by mirroring the RAID to an IDE drive, breaking the mirror and running off the IDE, destroy the array, replace ID3, recreate array, reestablish mirror, and then break mirror running off the array only and using the IDE as a slave.

Any thoughts on this? My only concern is that while the machine is building the mirror, I'm getting a large amount of bad sense errors on the RAID card. Any ideas, thoughts, nuggets of wisdom?

 

[corpanarchy]

I shouldn't have even posted mine, that sounds awfully similar except we're using a 2110 card in a clone running Netware 4.11.

Ours was ID5 as well, which we also replaced with a like drive.

Any resolution?

 

[lifegard2]

Not yet, the mirror finished building last night. I'm going back tomorrow to try and break the mirror and see if it will run off the IDE drive. If so, I think I'll be home free as I should then be able to kill the raid, rebuild from scratch, and then either clone or mirror back to the RAID. Hopefully this works, otherwise. . .

 

[corpanarchy]

I flashed the BIOS yesterday for my 2110 card, it still reports as dead half way through rebuild. I suspect a connection problem of sorts with the actual adapter as opposed to the array itself. The bonus is, when I rebooted the computer, it now says it's 85% rebuilt, the highest it's gone so far.

May be worth a go on your card as well?

 

[lifegard2]

I already tried the flash. The problem was that ID3 actually had bad sectors (had to run IBM's Drive Fitness Test on the drive while it was connected to a regular SCSI card).

After the mirror finished building, I was able to break it last night and boot from the IDE hard drive without the array connected. Once I could do that, I was able to delete the array, replace the bad drive (ID3), and then create a new array. After the new array was created and started bulding, I restarted the server and mirrored the IDE back to the array. This all worked without any problem.

The next step will be to break the mirror and have the array once again be the primary boot disk and the IDE back as a slave drive.

The one saving grace for me was that even with bad sectors on one drive in the array and another drive reported as failed, it would still boot up into the OS. From the OS I was able to setup the s/w mirror.

 

[corpanarchy]

Hmm, we're getting quite a few communication problems. To the point where Novell abends if it even tries to see the disks.

What card did you use to run the IBM fitness test?

 

[lifegard2]

Adaptect 29160

 

[corpanarchy]

We have a 2940 I may use, but for the time being it's back up and running... I'll be running the diags now methinks

 

[HbO]

lifegard,

where did you get the IBM Fitness Test tool?
Is it same with IBM diagnostic (F2)?

 

[rclarke250]

Can be downloaded from either ibm or hitachi...google DFT. or search the sites for DFT 3.6 is latest I think. will test other manu. drives also,but only hitachi and IBM get the advanced options.