mdadm: RAID5 array problem

[darealhaze]

Hi folks,

I have a problem with a very important RAID5 in my fileserver.
The array I am talking about consists of 4 250GB HDDs. Two of those are connected to the first SATA controller (sda and sdb) and the other two connect to the second one (sdc and sdd).
Apparently the first controller crashed, kicking 2 HDDs out of the array. After replacing the controller I am little bit stuck what to do next. The HDDs are all back online, but the is an inequality of the superblocks.

When doing mdadm --examine on /dev/sda or sdb, which were disconnect by the controller crash, all 4 hdds are listed as active and sync, but if I do the same on /dev/sdc or sdd the first two device are listed as faulty and removed. All in all this behaviour seems absolutely logically and correct.

Sadly I am not sure how to reassemble the array now. However I am now very experienced with mdadm. I guess there must be an easy solution, but because of the extreme importance of the array's data I definitely do now want to try around and maybe screw up the whole array.

I would be highly pleased if anybody could give me a hint on how to fix this.


Regards

darealhaze

[Noven]

I had a similar problem with raid1, maybe the same theory applies.

I removed the faulty disk from the array { mdadm /dev/mdx --fail /dev/sdx --remove /dev/sdx }
Then run fsck on the removed drive, then re-add it { mdadm /dev/mdx --re-add /dev/sdx }
While it is removed from the array if fsck isn't enough you can run whatever operations are needed to rebuild the superblocks correctly.
_________________
- Novensiles divi Flamen

>---- Miles Militis Fons ----<

[darealhaze]

Okay, I managed to get it work again. I just added the two device once again by