RAID arrays don't initialise properly on startup

Bug #634797 reported by Delan Azabani
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
mdadm (Ubuntu)
Invalid
Undecided
Unassigned

Bug Description

Binary package hint: mdadm

I have a RAID5 array, md0, with three disks, sdb, sdc and sdd. It works, however, on startup, palimpsest reports that it is partially built, has only one, unpositioned member disk, and has no size. I can fix this by stopping and starting the array, and the array will work properly until next reboot. This is inconvenient because if I start the computer to use it as a file server, I must open palimpsest and restart the array manually.

delan@delan:~$ uname -a; apt-cache policy mdadm
Linux delan 2.6.35-20-generic #29-Ubuntu SMP Fri Sep 3 14:55:28 UTC 2010 x86_64 GNU/Linux
mdadm:
  Installed: 2.6.7.1-1ubuntu15
  Candidate: 2.6.7.1-1ubuntu15
  Version table:
 *** 2.6.7.1-1ubuntu15 0
        500 http://archive.ubuntu.com/ubuntu/ maverick/main amd64 Packages
        100 /var/lib/dpkg/status

Revision history for this message
Delan Azabani (azabani) wrote :
Revision history for this message
Delan Azabani (azabani) wrote :
Revision history for this message
Delan Azabani (azabani) wrote :
Revision history for this message
Surbhi Palande (csurbhi) wrote :

@Delan Azabani, can you please attach the initramfs in /boot corresponding to the kernel that you boot in. Most likely your initramfs is not getting updated with the array definition.
Also can you attach the output of /etc/mdadm/mdadm.conf? and the output of /usr/share/mdadm/mkconf?

Changed in mdadm (Ubuntu):
status: New → Incomplete
Revision history for this message
Delan Azabani (azabani) wrote :
Revision history for this message
Delan Azabani (azabani) wrote :
Revision history for this message
Delan Azabani (azabani) wrote :
Revision history for this message
Delan Azabani (azabani) wrote :

I've found out a lot more about this problem. The above, attached mdadm.conf is that after running mkconf to regenerate the conf file. Really, though, mdadm should do this by itself when you create an array.

The situation is much better with the "ARRAY" definition in the mdadm.conf file, however, palimpsest still isn't picking up all the devices. According to "Edit Components" there is only the middle disk, disk 1, as part of the array. Funnily enough, though, palimpsest now reports that it's 2tb, running properly, and I can also mount and use it, though I'm afraid that if I write anything the array will go out of sync.

Revision history for this message
Delan Azabani (azabani) wrote :
Revision history for this message
Delan Azabani (azabani) wrote :

The screenshot above this comment shows the (still broken) effect of palimpsest, even when an array definition is in mdadm.conf, generated by "/usr/share/mdadm/mkconf > /etc/mdadm/mdadm.conf".

Please note that this doesn't appear to be a kernel problem, as this occurs on kernels 2.6.35-13 to 2.6.35-22 as well as 2.6.36 from the kernel PPA.

Revision history for this message
Delan Azabani (azabani) wrote :

I've changed the mdadm.conf file to have a manual definition of the array including the exact device names of the three member disks, however palimpsest still exhibits this problem where only one disk shows as a component. Of course, the array still mounts fine, but I am afraid of array desynchronisation if only one disk is being used.

The boot messages are even better now:

Sep 18 13:28:26 delan kernel: [ 2.387647] md: linear personality registered for level -1
Sep 18 13:28:26 delan kernel: [ 2.403626] md: multipath personality registered for level -4
Sep 18 13:28:26 delan kernel: [ 2.413439] md: raid0 personality registered for level 0
Sep 18 13:28:26 delan kernel: [ 2.422250] md: raid1 personality registered for level 1
Sep 18 13:28:26 delan kernel: [ 2.529563] md: bind<sdc>
Sep 18 13:28:26 delan kernel: [ 2.532306] md: bind<sdd>
Sep 18 13:28:26 delan kernel: [ 2.591876] md: bind<sdb>
Sep 18 13:28:26 delan kernel: [ 4.053002] md: raid6 personality registered for level 6
Sep 18 13:28:26 delan kernel: [ 4.053054] md: raid5 personality registered for level 5
Sep 18 13:28:26 delan kernel: [ 4.053105] md: raid4 personality registered for level 4
Sep 18 13:28:26 delan kernel: [ 4.053285] md/raid:md0: device sdb operational as raid disk 0
Sep 18 13:28:26 delan kernel: [ 4.053338] md/raid:md0: device sdd operational as raid disk 2
Sep 18 13:28:26 delan kernel: [ 4.053390] md/raid:md0: device sdc operational as raid disk 1
Sep 18 13:28:26 delan kernel: [ 4.053651] md/raid:md0: allocated 3230kB
Sep 18 13:28:26 delan kernel: [ 4.053726] md/raid:md0: raid level 5 active with 3 out of 3 devices, algorithm 2
Sep 18 13:28:26 delan kernel: [ 4.053815] md0: detected capacity change from 0 to 2000409329664
Sep 18 13:28:26 delan kernel: [ 4.054578] md0: p1
Sep 18 13:28:26 delan kernel: [ 4.116269] md: raid10 personality registered for level 10

Changed in mdadm (Ubuntu):
status: Incomplete → Confirmed
Revision history for this message
Delan Azabani (azabani) wrote :

Marking as invalid as this really stems from two other bugs.

affects: mdadm (Ubuntu) → gnome-disk-utility (Ubuntu)
affects: gnome-disk-utility (Ubuntu) → mdadm (Ubuntu)
Changed in mdadm (Ubuntu):
status: Confirmed → Invalid
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.