Re: 4 disks outage in RAID6

[Phil Turmel <philip@turmel.org>]

Good morning Mark,

On 12/20/2014 12:03 PM, Mark Kolama wrote:
> Dear List,
> 
> due to a controller failure, a raid6 with 16 drives
> lost 4 drives at once. The failure was noticed a few days later.
> 
> a examine output of all 16 drive is listed at
> http://pastebin.com/4WH9xp7K

Ok.  In the future, paste these in your email.  kernel.org has a
generous size limit and this sort of stuff should stay in the archives.
 As long as posters trim replies appropriately, it's not a problem.

> As you can see the event count differs on 4 drives with
> about 150 comparing to the other 12 drives.
> 
> I have already tried:
> mdadm --assemble --scan:
> assembled from 12 drives - not enough to start the array.
> 
> Then i tried:
> mdadm --assemble /dev/md0 /dev/sda1 /dev/sdb1 /dev/sdc1 /dev/sdd1
> /dev/sde1 /dev/sdf1 /dev/sdg1 /dev/sdh1 /dev/sdi1 /dev/sdj1 /dev/sdk1
> /dev/sdl1 /dev/sdm1 /dev/sdn1 /dev/sdo1 /dev/sdp1 --force

This should have worked, unless the array wasn't stopped first.  You
didn't show the actual response from mdadm, so we don't know.

There have also been bugs in various assembly features, so a report of
your kernel version and your mdadm version would be appropriate.

> /proc/mdstat after that:
> 
> Personalities : [raid6] [raid5] [raid4]
> md0 : inactive sda1[0](S) sdp1[15](S) sdo1[14](S) sdn1[13](S)
> sdm1[12](S) sdl1[11](S) sdk1[10](S) sdj1[9](S) sdi1[8](S) sdh1[7](S)
> sdg1[6](S) sdf1[5](S) sde1[4](S) sdd1[3](S) sdc1[2](S) sdb1[1](S)
>       62353932288 blocks super 1.2
> 
> No success either.

Try --force again, like so:

mdadm --stop /dev/md0

mdadm -Avf /dev/md0 /dev/sd[a-p]1

Show all of the output.  Also show the tail of dmesg where this
operation happens.

> So the next try would be recreating the array ?

Absolutely not.

Phil