Hi Wol,
You've been very helpful in trying to revive my array, but on my last
posting it has been quiet for a week now.
In that last posting I asked help from others on the aspects you wrote
"it's gotten out of my depth", but I've got no response from other list
members so far.
Am I doing something wrong?
Any case, I decided to get the OS (OpenSuse Tumbleweed) on a new 4TB
disk, so as to have enough space on the root FS to accommodate the
overlay files.
Though, as mentioned in my last posting, I don't fully grasp what is
written in the wiki on this, I decided to go ahead with the commands
given there. Alll went well,
up till assembling on the overlays (mdadm --assemble --force /dev/md1
$OVERLAYS):
cat /proc/mdstat
Personalities : [raid6] [raid5] [raid4]
md127 : active raid5 dm-0[0] dm-2[2] dm-3[4] dm-1[1]
11720649216 blocks super 1.2 level 5,
512k chunk, algorithm 2 [4/3] [UU_U]
[>....................] recovery = 0.0%
(300016/3906883072) finish=11872.9min speed=5483K/sec
bitmap: 8/30 pages [32KB], 65536KB chunk
unused devices: <none>
I then paused the rebuild, and stopped the array, as advised in the wiki:
echo 0 > /proc/sys/dev/raid/speed_limit_max
echo 0 > /proc/sys/dev/raid/speed_limit_min
mdadm --stop /dev/md127
mdadm: stopped /dev/md127
(Since I had no extra spare devices, adding that as stated in the wiki
didn't seem appropriate, but yes I issued "parallel -j1 mdadm --add
/dev/md127 ::: $OVERLAYS"; as could be expected, it failed because the
overlay drives where busy)
According the wiki, I now should do a fsck on the array, but apparently
the device ceased to exist:
fsck /dev/md127
'fsck' uit util-linux 2.32.1
e2fsck 1.44.4 (18-Aug-2018)
fsck.ext2: Bestand of map bestaat niet tijdens openen van /dev/md127
Mogelijk een niet-bestaand apparaat?
(In English: "The file or map doesn't exist while opening /dev/md127.
Possibly an non-existant device?")
And indeed, there is no /dev/md* anymore.
Trying to recreate the array on real devices fails with
mdadm --assemble --force /dev/md127 /dev/sda /dev/sdb /dev/sdc /dev/sde
mdadm: /dev/sda is busy - skipping
mdadm: /dev/sdb is busy - skipping
mdadm: /dev/sdc is busy - skipping
mdadm: /dev/sde is busy - skipping
It is not in /proc anymore too:
cat /proc/mdstat
Personalities : [raid6] [raid5] [raid4]
unused devices: <none>
Did I do something wrong? What is the best next step?
Thanks for your answer!
Cheers, Jogchum
Op 26-11-18 om 11:38 schreef Wols Lists:
On 22/11/18 10:56, Jogchum Reitsma wrote:
What is the best action here? Take /dev/sdf in the initial assembly, in
stead of /dev/sdd, and re-add /dev/sdd afterwards?
Or is there a way to tell mdadm that is should treat /dev/sdd as an
active drive?
Hmmm...
Sorry for the delay in responding, but this is getting out of my depth
:-( But yes, I think putting sdf in there should work. I don't know how
to tell mdadm that sdd should be active.
If you've got all the drives backed up, then fine. Actually, I think the
best bet from here would be to read up on overlays - it's in the wiki -
and overlay sda, sdb, sdf before re-assembling the array. That way,
you're not actually going to write to the array so if it goes
pear-shaped you haven't done any damage, and if it works you can just
tear down the overlay (reboot, say), and then assemble for real.
Cheers,
Wol
