2

I've recently moved house and have had some power problems after switching on my file server that's using a mdadm RAID5 set up. The file system I'm using is ext4.

It seems that at some point two (of five) disks fell out of the array almost simultaneously after a quick look with --examine. I figured that there was unlikely to be any writes to any of the disks to throw them out of parity since it's not used as the mount point for the root or home directories and wasn't being used at the time (although it may have been mounted).

So I figured this was a good time to stop the array and try

mdadm --assemble --scan --force -v

It will accept the 2nd disk to be kicked out as part of the array but will not accept the first. There doesn't appear to be anything terribly wrong with it other than the last sync date is "too far" away. To make matters worse, even though it can start the array with 4 of 5 disks the array doesn't appear to assemble correctly.

Another thing that might be worth noting is that the disks in the array appear to have changed drive letters after I booted them up again. Possibly something is wrong with the SATA controller?

I'm going to put in an order for some replacement disks to dd the current contents across just in case anything else I try ruins things further.

Fortunately I have backups for most of the contents however they're a little bit old and I'd like to avoid resorting to using them if possible.

If there's no way around this does anybody have a suggestion for a file recovery tool that would deal with this sort of scenario? I know how that should be impossible, but it can't hurt to ask right?

1 Answer 1

3

(This is probably too late, but in case anyone else stumbles across this...)

Something similar happened to me recently on a RAID10 array. A drive was accidentally removed for a few weeks (don't ask...). When I re-inserted it, mdadm --add and mdadm --re-add both failed. I had to do mdadm --zero-superblock on the drive, then mdadm --add worked fine.

If you have 4 out of 5 working disks and are using persistent superblocks the array should start in degraded mode; I'm not sure exactly what you mean by "doesn't appear to assemble correctly". If the shutdown was unclean then --force would probably be necessary, which you are already using.

There's also a bunch of information about RAID recovery at https://raid.wiki.kernel.org/index.php/RAID_Recovery.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .