Help, my RAID won't start

Tue Apr 4 15:48:28 UTC 2017

On Tue, Apr 4, 2017 at 8:18 AM, Kevin O'Gorman <kogorman at gmail.com> wrote:

>
>
> On Tue, Apr 4, 2017 at 6:50 AM, Rashkae <ubuntu at tigershaunt.com> wrote:
>
>> On 17-04-04 12:17 AM, Kevin O'Gorman wrote:
>>
>>> I've been fooling with this for a while.  Some things are fixed, but the
>>> darned thing still isn't right.
>>>
>>>
>> It looks like you've partitioned your RAID device itself.  I'm not too
>> sure how that's supposed to look, as I've never set them up that way.
>> (I partition the physical devices, as you've done, then create a raid
>> array, and create my FS directly on that array.)
>>
>> That's not to say what you have here is wrong, just that I'm not sure how
>> to interpret the blkid output.
>>
>> But actually fixing your current situation is probably going to be
>> simple.. *Don't* use device names in your fstab.. put in the UUID
>>
>> Example:
>>
>>
>> UUID=a6470f33-4709-4c96-8cf4-2fc0a4c2cdf9 /raid3        ext4   defaults
>> 0    2
>>
>> (Note the removal of nofail, and changing the last field to 2)
>>
>>
>> You might also want to update your mdadm.conf
>>
>> Check the device name of your running array with cat /proc/mdstat
>>
>> Assuming it's the same as you have in the original message, md127
>>
>> mdadm --detail --scan /dev/md127
>>
>> Use the output of that command to replace the ARRAY /dev/md3 metadata=1.2
>> name=camelot-x:3
>> UUID=07c3d55f:2c9b1534:6b0d5b78:dd9856ed
>>
>
> Thanks...
>
> I tried your entry in my /etc/fstab, and "mount /raid3", and while the
> mount finished without reporting an error, there
> was no filesystem on /raid3, and /var/log/syslog now ends with
>
> Apr  4 07:37:14 camelot-x wpa_supplicant[1553]: wlp7s0: WPA: Group
> rekeying completed with b4:75:0e:fc:84:9a [GTK=TKIP]
> Apr  4 07:46:46 camelot-x kernel: [38933.017667] EXT4-fs (md127p1):
> mounted filesystem with ordered data mode. Opts: (null)
> Apr  4 07:46:46 camelot-x systemd[1]: raid3.mount: Unit is bound to
> inactive unit dev-md3p1.device. Stopping, too.
> Apr  4 07:46:46 camelot-x systemd[1]: Unmounting /raid3...
> Apr  4 07:46:46 camelot-x systemd[1]: Unmounted /raid3.
>
> All of this in spite of /proc/mdstat reporting an active RAID array....
>
> So I tried putting the filesystem directly on the RAID, as you suggested,
> with
> dd if=/dev/zero of=/dev/md127 bs=1024 count=1024
> mdadm --stop /dev/md127
> mdadm --remove /dev/md127
> mdadm -A --scan /dev/md3
> mkfs.ext4 -m .01 -b 4096 -E stride=32,stripe-width=64 /dev/md3
>
> And was still unable to mount /raid3, but now I can see the syslog is
> complaining that /raid3 is still bound to the
> non-existent partition /dev/md3p1.  So I mounted it on a new directory
> /raid3x.  It seems okay, in that it has
> lost+found and lots of space.  But it did so last time too.
>
> I'm going to stick with my /etc/fstab entry for now, because I want to be
> sure I get a warning if it ever gets assembled
> as anything except /dev/md3.  The mdadm.conf entry is still correct, so
> I'm going to reboot to make sure everything gets
> assembled correctly.  Back in a few.....
>
> --
> Kevin O'Gorman
> #define QUESTION ((bb) || (!bb))   /* Shakespeare */
>
> Please consider the environment before printing this email.
>
That did NOT work.  Not exactly, anyway.

There was nothing mounted on /raid3 after the reboot, and there was a
/dev/md127 that I cannot get rid of in spite of multiple tries at
mdadm --stop /dev/md127
mdadm --remove /dev/md127
But it did not show up in /proc/mdstat any more so I reassembled it
mdadm -A --scan /dev/md3
And it was not only assembled but mounted too.

lsblk shows
sdd       8:48   0   3.7T  0 disk
└─sdd1    8:49   0   3.7T  0 part
  └─md3   9:3    0   7.3T  0 raid5 /raid3
sde       8:64   0   3.7T  0 disk
└─sde1    8:65   0   3.7T  0 part
  └─md3   9:3    0   7.3T  0 raid5 /raid3
sdf       8:80   0   3.7T  0 disk
└─sdf1    8:81   0   3.7T  0 part
  └─md3   9:3    0   7.3T  0 raid5 /raid3

blkid shows
/dev/sdd1: UUID="07c3d55f-2c9b-1534-6b0d-5b78dd9856ed"
UUID_SUB="24275237-b493-169e-9e88-df0debaae551" LABEL="camelot-x:3"
TYPE="linux_raid_member" PARTLABEL="cam-raid5-a"
PARTUUID="729558f8-fbcf-45c1-9d9b-c6bfe6ec4bf6"
/dev/sde1: UUID="07c3d55f-2c9b-1534-6b0d-5b78dd9856ed"
UUID_SUB="cf36e69a-c110-b0ae-b504-fcfeb2dea024" LABEL="camelot-x:3"
TYPE="linux_raid_member" PARTLABEL="cam-raid5-c"
PARTUUID="654fc581-f6ef-43a4-b583-6f3cd7d52e61"
/dev/sdf1: UUID="07c3d55f-2c9b-1534-6b0d-5b78dd9856ed"
UUID_SUB="54fbf294-766f-c8eb-9663-02a39534003f" LABEL="camelot-x:3"
TYPE="linux_raid_member" PARTLABEL="cam-raid5-b"
PARTUUID="f6a7ba77-278e-4b64-8f7e-16041b85f097"
and
/dev/md3: UUID="97309dc0-0f54-45d8-853f-af99dead3ed2" TYPE="ext4"

df -h shows
/dev/md3        7.3T   51M  7.3T   1% /raid3

QUESIONS:

So why is there still a /dev/md127, and why does the array not assemble as
/dev/md3?
/etc/mdadm/mdadm.conf contains just the one array:
ARRAY /dev/md3 metadata=1.2 name=camelot-x:3
UUID=07c3d55f:2c9b1534:6b0d5b78:dd9856ed
which is still what I get from
mdadm --detail --scan

mount and umount commands work as expected, both with my original
/etc/fstab entry or with yours.  I'll leave yours in place and try a
reboot....

-- 
Kevin O'Gorman
#define QUESTION ((bb) || (!bb))   /* Shakespeare */

Please consider the environment before printing this email.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/ubuntu-users/attachments/20170404/46555a57/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 441 bytes
Desc: not available
URL: <https://lists.ubuntu.com/archives/ubuntu-users/attachments/20170404/46555a57/attachment.gif>