Comments on: Moving your Linux root partition to software RAID

By: tyler

tyler — Thu, 29 Oct 2009 22:42:17 +0000

Hi Mudgen,

What’s the exact command with mkinitrd? Any switches? It would be good to have it here.

By: Mudgen

Mudgen — Thu, 29 Oct 2009 17:36:33 +0000

What a great post! I’m actually familiar with all the pieces, but it’s terrific to have them put together in a road map.
Surprised the original Centos poster did not call out that the update-initramfs equivalent on Redhat/Centos/Fedora would be mkinitrd.

By: Help setting up LinuxRAID-1 in opensuse 11.1 - Page 2 - openSUSE Forums

Help setting up LinuxRAID-1 in opensuse 11.1 - Page 2 - openSUSE Forums — Sun, 05 Jul 2009 15:50:14 +0000

[...] if this go well. I have find this next 2 pages: Fundaci

By: tyler

tyler — Mon, 02 Feb 2009 14:38:26 +0000

Thanks to several of you for pointing out the need to create filesystems on the new RAID partitions. I’ve edited the article to reflect that in step 2.

By: wil

wil — Mon, 02 Feb 2009 09:51:53 +0000

Hi,

Thanks for this document. I agree with Wes that it is best to format your disk right after step 2 if it already has data on it. I neglected to do that and on reboot it gave me this error:

The filesystem size (according to the superblock) is nnn blocks
The physical size of the device is yyy blocks
Either the superblock or the partition table is likely to be corrupt!
Abort?

Of course this makes perfect sense as now both the original partition and the raid disk had their own superblocks on there. Formatting the md(x) partitions and rerunning the rsync solved that.

By: tyler

tyler — Sun, 11 Jan 2009 20:31:22 +0000

Correction: the list of devices in the superblock can not match whenever the partitions are not actually in the array. I had a machine showing this problem until recently, but I can’t say how it came about. I just re-added the affected partition to the array and it resolved itself.

Compare:

mdadm –detail /dev/md6
mdadm -E /dev/sda9
mdadm -E /dev/sdb9

If sdb9 still shows bad superblock data, I’d suggest removing it from the array and re-adding it.

By: tyler

tyler — Sun, 11 Jan 2009 20:23:50 +0000

Yikes! I’ve never seen two parts of a RAID array show different results! If you solve this, please post it here.

By: Ioannis

Ioannis — Sun, 11 Jan 2009 19:04:26 +0000

Marvelous article. Compact and with rich information.
Unfortunately I have the most weird problem (in 8.04 server):
After creating the raid arrays, rebooting successfully growing the raid (1 for me) arrays to include the second (sdb1, sdb2 etc…) devices, the system cannot boot (md devices are stopped during boot). It might have something to do with the order the modules are loaded (the sda and sdb sit in different controllers with different drivers), but it might have something to do with a mix in md arrays and sd devices. For example I have sda9-sdb9 that belong to md6 but the mdadm reports for the sda member:
mdadm -E /dev/sda9
/dev/sda9:
Magic : a92b4efc
Version : 00.90.00
UUID : 61b6692d:3900807a:5e76b0e2:85a6fbc1 (local to host TransNetStreamer)
Creation Time : Sat Jan 10 17:13:51 2009
Raid Level : raid1
Used Dev Size : 9735232 (9.28 GiB 9.97 GB)
Array Size : 9735232 (9.28 GiB 9.97 GB)
Raid Devices : 2
Total Devices : 1
Preferred Minor : 6

Update Time : Sun Jan 11 20:15:26 2009
State : clean
Active Devices : 1
Working Devices : 1
Failed Devices : 1
Spare Devices : 0
Checksum : bb676a86 – correct
Events : 0.760

Number Major Minor RaidDevice State
this 0 8 9 0 active sync /dev/sda9

0 0 8 9 0 active sync /dev/sda9
1 1 0 0 1 faulty removed
while for the sdb member:
mdadm -E /dev/sdb9
/dev/sdb9:
Magic : a92b4efc
Version : 00.90.00
UUID : 61b6692d:3900807a:5e76b0e2:85a6fbc1 (local to host TransNetStreamer)
Creation Time : Sat Jan 10 17:13:51 2009
Raid Level : raid1
Used Dev Size : 9735232 (9.28 GiB 9.97 GB)
Array Size : 9735232 (9.28 GiB 9.97 GB)
Raid Devices : 2
Total Devices : 2
Preferred Minor : 6

Update Time : Sun Jan 11 15:22:22 2009
State : clean
Active Devices : 2
Working Devices : 2
Failed Devices : 0
Spare Devices : 0
Checksum : bb67209c – correct
Events : 0.66

Number Major Minor RaidDevice State
this 1 8 25 1 active sync /dev/sdb9

0 0 8 9 0 active sync /dev/sda9
1 1 8 25 1 active sync /dev/sdb9

different things. HOW IS THAT possible ?

Again thanks for your time

By: maddog737

maddog737 — Sat, 10 Jan 2009 04:35:49 +0000

Same setup (Ubuntu 8.04), but my problem for the past few days was 2-fold:

1st at the end of step 4, grub-install didn’t work (hung), so I manually ran grub and entered:
device (hd1) /dev/sdb
root (hd1,0)
setup (hd1)
quit

worked, but then on reboot at step 7 and selecting the second drive as the boot drive through the BIOS, the boot process would hang, dropping me to initramfs. At this point, I could mount /dev/md0 by

mdadm -A /dev/md0 /dev/sdb1

for each RAID partition, then exit from initramfs. Everything would boot then. After too many attempts at resolving this, I found that the problem was resolved by allowing the boot sequence to drop to initramfs, and enter

mdadm -As –homehost=” –auto=yes –auto-update-homehost

exit initramfs and allow the boot to complete. Reboot once (or twice), and everything works. From what I understand (noob here), when the RAID1 was created, the homehost= in mdadm.conf used by mdadm during the boot process didn’t match what was in the properties written to the RAID partition when it was created resulting in arrays not being mounted automatically. Several days of debugging were summarized in that simple line…

That, and to read about the problem with degraded RAID arrays here:
https://help.ubuntu.com/community/DegradedRAID

By: tyler

tyler — Wed, 12 Nov 2008 14:39:23 +0000

Fixed the sfdisk line, Wes. Thanks.

Regarding the cp, that assumes that the directories used for mounting like /dev, /mnt, /var/lock, etc are tmpfs. That is true for Ubuntu but not all distributions. Just running the cp -ax like that may copy the contents of those directories. Hence the cautious rsync approach.