My software RAID1 failed! Curse this wretched server!

Tips and Tutorials

There’s an excellent guide online for how to swap out a failed or degraded hard drive in a software RAID 1, particularly in CentOS (credit to Falko Timme). Unfortunately this article leaves out one very critical step of the process, so since we made some revisions for our own reference, we thought you might enjoy it if we shared our revised version with you. Yup, we’re -that- nice.

Replacing A Failed Hard Drive In A Software RAID1 Array

This guide shows how to remove a failed hard drive from a Linux RAID1 array (software RAID), and how to add a new hard disk to the RAID1 array without losing data.

Preliminary Note

In this example I have two hard drives, /dev/sda and /dev/sdb, with the partitions /dev/sda1 and /dev/sda2 as well as /dev/sdb1 and /dev/sdb2.

/dev/sda1 and /dev/sdb1 make up the RAID1 array /dev/md0.

/dev/sda2 and /dev/sdb2 make up the RAID1 array /dev/md1.

/dev/sda1 + /dev/sdb1 = /dev/md0

/dev/sda2 + /dev/sdb2 = /dev/md1

/dev/sdb has failed, and we want to replace it.

How Do I Tell If A Hard Disk Has Failed?

If a disk has failed, you will probably find a lot of error messages in the log files, e.g. /var/log/messages or /var/log/syslog.

You can also run

cat /proc/mdstat

and instead of the string [UU] you will see [U_] if you have a degraded RAID1 array.

Removing The Failed Disk

To remove /dev/sdb, we will mark /dev/sdb1 and /dev/sdb2 as failed and remove them from their respective RAID arrays (/dev/md0 and /dev/md1).

First we mark /dev/sdb1 as failed:
mdadm –manage /dev/md0 –fail /dev/sdb1

The output of
cat /proc/mdstat
should look like this:

server1:~# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid5] [raid4] [raid6] [raid10]
md0 : active raid1 sda1[0] sdb1[2](F)
      24418688 blocks [2/1] [U_]

md1 : active raid1 sda2[0] sdb2[1]
      24418688 blocks [2/2] [UU]

unused devices: <none>

 

Then we remove /dev/sdb1 from /dev/md0:
mdadm –manage /dev/md0 –remove /dev/sdb1

The output should be like this:

server1:~# mdadm --manage /dev/md0 --remove /dev/sdb1
mdadm: hot removed /dev/sdb1

 

And cat /proc/mdstat should show this:

server1:~# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid5] [raid4] [raid6] [raid10]
md0 : active raid1 sda1[0]
      24418688 blocks [2/1] [U_]

md1 : active raid1 sda2[0] sdb2[1]
      24418688 blocks [2/2] [UU]</nowiki>

unused devices: <none>

 

Now we do the same steps again for /dev/sdb2 (which is part of /dev/md1):
mdadm –manage /dev/md1 –fail /dev/sdb2

cat /proc/mdstat

server1:~# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid5] [raid4] [raid6] [raid10]
md0 : active raid1 sda1[0]
24418688 blocks [2/1] [U_]

md1 : active raid1 sda2[0] sdb2[2](F)
24418688 blocks [2/1] [U_]</nowiki>

unused devices: <none>

mdadm –manage /dev/md1 –remove /dev/sdb2

server1:~# mdadm –manage /dev/md1 –remove /dev/sdb2
mdadm: hot removed /dev/sdb2

cat /proc/mdstat

server1:~# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid5] [raid4] [raid6] [raid10]
md0 : active raid1 sda1[0]
24418688 blocks [2/1] [U_]

md1 : active raid1 sda2[0]
24418688 blocks [2/1] [U_]</nowiki>

unused devices: <none>

 

Then power down the system:
shutdown -h now
and replace the old /dev/sdb hard drive with a new one (it must have at least the same size as the old one – if it’s only a few MB smaller than the old one then rebuilding the arrays will fail).

Adding The New Hard Disk

After you have changed the hard disk /dev/sdb, boot the system.

The first thing we must do now is to create the exact same partitioning as on /dev/sda. We can do this with one simple command:
sfdisk -d /dev/sda | sfdisk /dev/sdb
You can run
fdisk -l
to check if both hard drives have the same partitioning now.

Next we add /dev/sdb1 to /dev/md0 and /dev/sdb2 to /dev/md1:
mdadm –manage /dev/md0 –add /dev/sdb1

server1:~# mdadm --manage /dev/md0 --add /dev/sdb1
mdadm: re-added /dev/sdb1

mdadm --manage /dev/md1 --add /dev/sdb2

server1:~# mdadm --manage /dev/md1 --add /dev/sdb2
mdadm: re-added /dev/sdb2

 

Now both arrays (/dev/md0 and /dev/md1) will be synchronized. Run cat /proc/mdstat to see when it’s finished.

During the synchronization the output will look like this:

server1:~# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid5] [raid4] [raid6] [raid10]
md0 : active raid1 sda1[0] sdb1[1]
      24418688 blocks [2/1] [U_]
      [=>...................]  recovery =  9.9% (2423168/24418688) finish=2.8min speed=127535K/sec

md1 : active raid1 sda2[0] sdb2[1]
      24418688 blocks [2/1] [U_]
      [=>...................]  recovery =  6.4% (1572096/24418688) finish=1.9min speed=196512K/sec

unused devices: <none>

 

When the synchronization is finished, the output will look like this:
mdstat-fix

server1:~# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid5] [raid4] [raid6] [raid10]
md0 : active raid1 sda1[0] sdb1[1]
      24418688 blocks [2/2] [UU]

md1 : active raid1 sda2[0] sdb2[1]
      24418688 blocks [2/2] [UU]

unused devices: <none>

 

Last (and one of the most important) step, Install grub on new drive and reinstall on existing drive:
server1:~# grub-install –recheck /dev/sdb

Probing devices to guess BIOS drives. This may take a long time.
Installation finished. No error reported.
This is the contents of the device map /boot/grub/device.map.
Check if this is correct or not. If any of the lines is incorrect,
fix it and re-run the script `grub-install'.

(hd0)   /dev/sda
(hd1)   /dev/sdb
server1:~# grub-install /dev/sda
Installation finished. No error reported.