Difference between revisions of "Convert a single drive system to RAID"

From ArchWiki
Jump to: navigation, search
(Partition Original Disk)
(Add Disk to Array)
Line 270: Line 270:
  [root@arch ~]# fdisk -l
  [root@arch ~]# fdisk -l
=== Add Disk to Array ===
=== Add Disk Partition to Array ===
  [root@svn ~]# mdadm /dev/md0 -a /dev/sda2
  [root@svn ~]# mdadm /dev/md0 -a /dev/sda2
  mdadm: hot added /dev/sda2
  mdadm: hot added /dev/sda2

Revision as of 07:11, 28 August 2009

You already have a fully functional system setup on a single drive, but you would like to add some redundancy to the setup by using RAID-1 to mirror your data across 2 drives. This guide follows the following steps to make the required changes, without losing data.

  • Create a single-disk RAID-1 array with our new disk
  • Move all your data from the old-disk to the new RAID-1 array
  • Verify the data move was successful
  • Wipe the old disk and add it to the new RAID-1 array
Warning: Make a backup first. Even though our aim is to convert to a RAID setup without losing data, there's no guarantees the process will be perfect, and there is a high risk of accidents happening.


  • I will assume for the sake of the guide that the disk currently in your system is Template:Filename and your new disk is Template:Filename.
  • We will create the following configuration:
    • 1 x RAID-1 array for the file-system (using 2 x partitions, 1 on each disk)
    • 2 x Swap Partitions using 1 partition on each disk.

The swap partitions will not be in a RAID array as having swap on RAID serves no purpose. Refer to this article for reasons why.

  • To minimize the risk of Data on Disk (DoD) changing in the middle of our changes, I suggest you drop to single user mode before you start by using the Template:Codeline command.
  • You will need to be the root user for the entire process.

Create New RAID Array

First we need to create a single-disk RAID array using the new disk.

Partition the Disk

Use fdisk or your partitioning program of choice to setup 2 primary partitions on your new disk. Make the swap partition half the size of the total swap you want (the other half will go on the other disk).

Drop to single user mode:

init 1

To see the current partitions:

fdisk -l

To partition the new disk

fdisk /dev/sdb

Then the fdisk commands to partition the new disk. Note that everything after the "#" is an explanation of what the command is doing:

n # new
p # primary
1 # first partition
1 # start at first cylinder
101 # end cylinder, 0.1% of the disk. Note: update this number as appropriate for your disk.
n # new
p # primary
2 # second partition
press enter # Uses the default start from the end of the first partition
press enter # Uses the default of using all the remain space on the disk.
t # set the partition type
1 # for partition number 1
82 # ... and set it to be swap
t # set the partition type
2 # for partition number 2 ...
fd # ... and set it to be "linux raid auto"
p # print what the partition table will look like
w # now write all of the above changes to disk

At the end of partitioning, your partitions should look something like this:

[root@arch ~]# fdisk -l /dev/sdb
Disk /dev/sdb: 80.0 GB, 80025280000 bytes
255 heads, 63 sectors/track, 9729 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Disk identifier: 0x00000000

   Device Boot      Start         End      Blocks   Id  System
/dev/sdb1               1          66      530113+  82  Linux swap / Solaris
/dev/sdb2              67        9729    77618047+  fd  Linux raid autodetect

Make sure your your partition types are set correctly. Template:Codeline is type Template:Codeline and Template:Codeline is type Template:Codeline.

Create the RAID Device

Next, create the single-disk RAID-1 array. Note the Template:Codeline keyword is specified as one of our devices.

[root@arch ~]# mdadm --create /dev/md0 --level=1 --raid-devices=2 missing /dev/sdb2
mdadm: array /dev/md0 started.

Note: if the above command causes mdadm to say "no such device /dev/sdb2", then reboot, and run the command again.

Make sure the array has been created correctly by checking Template:Filename:

[root@arch ~]# cat /proc/mdstat
Personalities : [linear] [raid0] [raid1] [raid5] [multipath] [raid6] [raid10]
md0 : active raid1 sdb2[1]
     40064 blocks [2/1] [_U]

unused devices: <none>

The devices are intact, however in a degraded state. (Because it's missing half the array!)

Make File Systems

Use whatever filesystem is your preference here. I'll use ext3 for this guide.

[root@arch ~]# mkfs -t ext3 -j -L RAID-ONE /dev/md0
mke2fs 1.38 (30-Jun-2005)
Filesystem label=
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
10027008 inodes, 20027008 blocks
1001350 blocks (5.00%) reserved for the super user
First data block=0
612 block groups
32768 blocks per group, 32768 fragments per group
16384 inodes per group
Superblock backups stored on blocks:
        32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,
        4096000, 7962624, 11239424

Writing inode tables: done
Creating journal (32768 blocks): done
Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 25 mounts or
180 days, whichever comes first.  Use tune2fs -c or -i to override.

Make a file system on the swap partition:

[root@arch ~]# mkswap /dev/sdb1 -L NEW-SWAP
Setting up swapspace version 1, size = 271314 kB
no label, UUID=9d746813-2d6b-4706-a56a-ecfd108f3fe9

Copy Data

The new RAID-1 array is ready to start accepting data! So now we need to mount the array, and copy everything from the old system to the new system

Mount the Array

[root@arch ~]# mkdir /mnt/new-raid
[root@arch ~]# mount /dev/md0 /mnt/new-raid

Copy the Data

[root@arch ~]# cd /mnt/new-raid
[root@arch mnt]# rsync -avH --delete --progress -x / /mnt/new-raid

Alternatively, you can use tar if you prefer instead of the above rsync command. However rsync will be quicker if you are only copying over changes. The tar command is: tar -C / -clspf - . | tar -xlspvf -

Update GRUB

Use your preferred text editor to open Template:Filename.

--- SNIP ---
default   0
color light-blue/black light-cyan/blue

## fallback
fallback 1

# (0) Arch Linux
title  Arch Linux - Original Disc
root   (hd0,0)
kernel /vmlinuz26 root=/dev/sda1

# (1) Arch Linux
title  Arch Linux - New RAID
root   (hd1,0)
#kernel /vmlinuz26 root=/dev/sda1 ro
kernel /vmlinuz26 root=/dev/md0
--- SNIP ---

Notice we added the Template:Codeline line and duplicated the Arch Linux entry with a different Template:Codeline directive on the kernel line.

Also update the "kopt" and "groot" sections, as shown below, if they are in your Template:Filename file, because it will make applying distribution kernel updates easier:

- # kopt=root=UUID=fbafab1a-18f5-4bb9-9e66-a71c1b00977e ro
+ # kopt=root=/dev/md0 ro

## default grub root device
## e.g. groot=(hd0,0)
- # groot=(hd0,0)
+ # groot=(hd0,1)

Alter fstab

You need to tell fstab on the new disk where to find the new devices. It's better to use UUID codes here, which should not change, even if our partition detection order changes or a drive gets removed.

To find the UUID to use:

[root@arch ~]# blkid
/dev/sda1: TYPE="swap" UUID="34656682b-34ad-8ed5-9233-dfab42272212" 
/dev/sdb1: UUID="9ff5682b-d5a1-4ed5-8d63-d1df911e0142" TYPE="swap" LABEL="NEW-SWAP" 
/dev/md0: UUID="6f2ea3d3-d7be-4c9d-adfa-dbeeedaf128e" SEC_TYPE="ext2" TYPE="ext3" LABEL="RAID-ONE" 
/dev/sda2: UUID="13dd2227-6592-403b-931a-7f3e14a23e1f" TYPE="ext2" 
/dev/sdb2: UUID="b28813e7-15fc-d4aa-dc8a-e2c1de641df1" TYPE="mdraid" 

Look for the partition labeled "NEW-SWAP", on /dev/sdb1, that we created above. Copy your swap partition's UUID into the new fstab, as shown below. Of course we also add /dev/md0, as our root mount point.

[root@arch ~]# cat /mnt/new-raid/etc/fstab
/dev/md0    /    ext3     defaults   0 1
UUID=9ff5682b-d5a1-4ed5-8d63-d1df911e0142 none swap sw 0 0

Rebuild initcpio or initramfs

[root@arch ~]# mount --bind /sys /mnt/new-raid/sys
[root@arch ~]# mount --bind /proc /mnt/new-raid/proc
[root@arch ~]# mount --bind /dev /mnt/new-raid/dev
[root@arch ~]# chroot /mnt/new-raid/
[root ~]# 

You are now chrooted in what will become the root of your RAID-1 system. Complete the appropriate section below for your distribution (almost every other step is identical, irrespective of the linux variant).

For Arch linux: Rebuild initcpio

Edit Template:Filename to include Template:Codeline in the HOOKS array. Place it before Template:Codeline but after Template:Codeline, Template:Codeline and Template:Codeline (whichever is appropriate for your hardware).

[root ~]# mkinitcpio -g /boot/kernel26.img
[root ~]# exit

For Ubuntu or Debian: Rebuild initramfs

nano /etc/mdadm/mdadm.conf

... and change the "MAILADDR" line to be your email address, if you want emailed alerts of problems with the RAID-1.

Then save the array configuration with UUIDs to make it easier for the system to find /dev/md0 on bootup. If you don't do this, you can get an "ALERT! /dev/md0 does not exist" error when booting :

mdadm --detail --scan >> /etc/mdadm/mdadm.conf 

Then rebuild initramfs, incorporating the above two changes:

update-initramfs -k `uname -r` -c -t

Install GRUB on the RAID Array

Start grub:

[root@arch /]# grub --no-floppy

Then we find our two partitions - the current one (hd0,0) (I.e. first disk, first partition), and (hd1,1) (i.e. the partition we just added above, on the second partition of the second drive). Check you get two results here:

grub> find /boot/grub/stage1

Then we tell grub to assume the new second drive is (hd0), i.e. the first disk in the system (when it is not currently the case). If your first disk fails, however, and you remove it, or you change the order disks are detected in the BIOS so that you can boot from your second disk, then your second disk will become the first disk in the system. The MBR will then be correct, your new second drive will have become your first drive, and you will be able to boot from this disk.

grub> device (hd0) /dev/sdb

Then we install GRUB onto the MBR of our new second drive. Check that the "partition type" is detected as "0xfd", as shown below, to make sure you have the right partition:

grub> root (hd0,1)
 Filesystem type is ext2fs, partition type 0xfd
grub> setup (hd0)
 Checking if "/boot/grub/stage1" exists... yes
 Checking if "/boot/grub/stage2" exists... yes
 Checking if "/boot/grub/e2fs_stage1_5" exists... yes
 Running "embed /boot/grub/e2fs_stage1_5 (hd0)"...  16 sectors are embedded. succeeded
 Running "install /boot/grub/stage1 (hd0) (hd0)1+16 p (hd0,1)/boot/grub/stage2 /boot/grub/grub.conf"... succeeded
grub> quit

Verify Success

Reboot your computer, making sure it boots from the new raid disk (Template:Filename) and not the original disk (Template:Filename). You may need to change the boot device priorities in your BIOS to do this.

Once the GRUB on the new disk loads, make sure you select to boot the new entry you created in Template:Filename earlier.

Verify you have booted from the RAID array by looking at the output of mount. you should have a line similar to the following in the output:

[root@arch ~]# mount
/dev/md0 on / type ext3 (rw)

Also Template:Codeline:

[root@arch ~]# swapon -s
Filename                Type           Size    Used    Priority
/dev/sdb1               partition      4000144 16      -1

Note it is the swap partition on Template:Filename that is in use, nothing from Template:Filename.

If system boots fine, and the output of the above commands is correct, then congratulations! You're now running off the degraded RAID array. We can add the original disk to the array now to bring it up to full performance.

Add Original Disk to Array

Partition Original Disk

Take the output of Template:Codeline on your new disk, and make the partitions on your original disk look the same.

[root@arch ~]# fdisk -l /dev/sda
Disk /dev/sda: 80.0 GB, 80025280000 bytes
255 heads, 63 sectors/track, 9729 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Disk identifier: 0x00000000

   Device Boot      Start         End      Blocks   Id  System
/dev/sda1               1          66      530113+  82  Linux swap / Solaris
/dev/sda2              67        9729    77618047+  fd  Linux raid autodetect

Another way - this will output /dev/sda partition layout to a file, then it's used as input for partioning /dev/sdb. Use the --force if needed...

[root@arch ~]# sfdisk -d /dev/sda > raidinfo-partitions.sda
[root@arch ~]# sfdisk /dev/sdb < raidinfo-partitions.sda
[root@arch ~]# sfdisk --force /dev/sdb < raidinfo-partitions.sda

To verify that the partitioning is identical:

[root@arch ~]# fdisk -l

Add Disk Partition to Array

[root@svn ~]# mdadm /dev/md0 -a /dev/sda2
mdadm: hot added /dev/sda2

Verify that the RAID array is being rebuilt.

[root@arch ~]# cat /proc/mdstat
Personalities : [linear] [raid0] [raid1] [raid5] [multipath] [raid6] [raid10]
md0 : active raid1 sda2[2] sdb2[1]
      80108032 blocks [2/1] [_U]
      [>....................]  recovery =  1.2% (1002176/80108032) finish=42.0min speed=31318K/sec

unused devices: <none>