https://wiki.archlinux.org/api.php?action=feedcontributions&user=Mcarreira&feedformat=atomArchWiki - User contributions [en]2024-03-29T02:39:22ZUser contributionsMediaWiki 1.41.0https://wiki.archlinux.org/index.php?title=Convert_a_single_drive_system_to_RAID&diff=188070Convert a single drive system to RAID2012-03-06T23:11:55Z<p>Mcarreira: Include information about raider</p>
<hr />
<div>[[Category:File systems (English)]]<br />
{{i18n|Convert a single drive system to RAID}}<br />
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.<br />
* Create a single-disk RAID-1 array with our new disk<br />
* Move all your data from the old-disk to the new RAID-1 array<br />
* Verify the data move was successful<br />
* Wipe the old disk and add it to the new RAID-1 array<br />
<br />
{{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.}}<br />
<br />
<br />
== Assumptions ==<br />
* I will assume for the sake of the guide that the disk currently in your system is {{ic|/dev/sda}} and your new disk is {{ic|/dev/sdb}}.<br />
* We will create the following configuration:<br />
** 1 x RAID-1 array for the file-system (using 2 x partitions, 1 on each disk)<br />
** 2 x Swap Partitions using 1 partition on each disk.<br />
The swap partitions will not be in a RAID array as having swap on RAID serves no purpose. Refer to [http://unthought.net/Software-RAID.HOWTO/Software-RAID.HOWTO-2.html#ss2.3 this article] for reasons why.<br />
<br />
* 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 {{Ic|telinit 1}} command.<br />
<br />
* You will need to be the root user for the entire process.<br />
<br />
== Create new RAID array ==<br />
First we need to create a single-disk RAID array using the new disk.<br />
=== Partition the Disk ===<br />
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).<br />
<br />
Drop to single user mode:<br />
telinit 1<br />
<br />
To see the current partitions:<br />
fdisk -l<br />
<br />
To partition the new disk<br />
fdisk /dev/sdb<br />
<br />
Then the fdisk commands to partition the new disk. Note that everything after the "#" is an explanation of what the command is doing:<br />
c # Turn off DOS compatibility (optional).<br />
n # new<br />
p # primary<br />
1 # first partition<br />
1 # start at first cylinder<br />
101 # end cylinder, 0.1% of the disk. Note: update this number as appropriate for your disk.<br />
n # new<br />
p # primary<br />
2 # second partition<br />
press enter # Uses the default start from the end of the first partition<br />
press enter # Uses the default of using all the remain space on the disk.<br />
t # set the partition type<br />
1 # for partition number 1<br />
82 # ... and set it to be swap<br />
t # set the partition type<br />
2 # for partition number 2 ...<br />
fd # ... and set it to be "linux raid auto"<br />
a # Toggle the bootable flag to be "on"<br />
2 # for partition number 2.<br />
p # print what the partition table will look like<br />
w # now write all of the above changes to disk<br />
<br />
At the end of partitioning, your partitions should look something like this:<br />
[root@arch ~]# fdisk -l /dev/sdb<br />
Disk /dev/sdb: 80.0 GB, 80025280000 bytes<br />
255 heads, 63 sectors/track, 9729 cylinders<br />
Units = cylinders of 16065 * 512 = 8225280 bytes<br />
Disk identifier: 0x00000000<br />
<br />
Device Boot Start End Blocks Id System<br />
/dev/sdb1 1 66 530113+ 82 Linux swap / Solaris<br />
/dev/sdb2 67 9729 77618047+ fd Linux raid autodetect<br />
<br />
Make sure your your partition types are set correctly. {{Ic|"Linux Swap"}} is type {{Ic|82}} and {{Ic|"Linux raid autodetect"}} is type {{Ic|FD}}.<br />
<br />
=== Create the RAID device ===<br />
Next, create the single-disk RAID-1 array. Note the {{Ic|"missing"}} keyword is specified as one of our devices. We are going to fill this missing device later.<br />
[root@arch ~]# mdadm --create /dev/md0 --level=1 --raid-devices=2 missing /dev/sdb2<br />
mdadm: array /dev/md0 started.<br />
<br />
Note: If the above command causes mdadm to say "no such device /dev/sdb2", then reboot, and run the command again.<br />
<br />
If you want to use Grub 0.97 (default in the Arch Linux 2010.05 release) on RAID 1, you need to specify an older version of metadata than the default. Add the option "--metadata=0.90" to the above command. Otherwise Grub will respond with "Filesystem type unknown, partition type 0xfd" and refuse to install. This is supposedly not necessary with Grub 2.<br />
[root@arch ~]# mdadm --create /dev/md0 --metadata=0.90 --level=1 --raid-devices=2 missing /dev/sdb2<br />
mdadm: array /dev/md0 started.<br />
<br />
Make sure the array has been created correctly by checking {{ic|/proc/mdstat}}:<br />
[root@arch ~]# cat /proc/mdstat<br />
Personalities : [linear] [raid0] [raid1] [raid5] [multipath] [raid6] [raid10]<br />
md0 : active raid1 sdb2[1]<br />
40064 blocks [2/1] [_U]<br />
<br />
unused devices: <none><br />
<br />
The devices are intact, however in a degraded state. (Because it's missing half the array!)<br />
<br />
=== Make file systems ===<br />
Use the file system of your preference here. I'll use ext3 for this guide.<br />
[root@arch ~]# mkfs -t ext3 -j -L RAID-ONE /dev/md0<br />
mke2fs 1.38 (30-Jun-2005)<br />
Filesystem label=<br />
OS type: Linux<br />
Block size=4096 (log=2)<br />
Fragment size=4096 (log=2)<br />
10027008 inodes, 20027008 blocks<br />
1001350 blocks (5.00%) reserved for the super user<br />
First data block=0<br />
612 block groups<br />
32768 blocks per group, 32768 fragments per group<br />
16384 inodes per group<br />
Superblock backups stored on blocks:<br />
32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,<br />
4096000, 7962624, 11239424<br />
<br />
Writing inode tables: done<br />
Creating journal (32768 blocks): done<br />
Writing superblocks and filesystem accounting information: done<br />
<br />
This filesystem will be automatically checked every 25 mounts or<br />
180 days, whichever comes first. Use tune2fs -c or -i to override.<br />
<br />
Make a file system on the swap partition:<br />
[root@arch ~]# mkswap -L NEW-SWAP /dev/sdb1<br />
Setting up swapspace version 1, size = 271314 kB<br />
LABEL=NEW-SWAP, UUID=9d746813-2d6b-4706-a56a-ecfd108f3fe9<br />
<br />
== Copy data ==<br />
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<br />
<br />
=== Mount the array ===<br />
[root@arch ~]# mkdir /mnt/new-raid<br />
[root@arch ~]# mount /dev/md0 /mnt/new-raid<br />
<br />
=== Copy the data ===<br />
[root@arch ~]# rsync -avxHAXS --delete --progress / /mnt/new-raid<br />
<br />
Note that by using the '''-x''' option you are limiting rsync to a single file system. If you have a more traditional file system layout, with different partitions for /boot, /home, and perhaps others, you will need to rsync those file systems separately. For example:<br />
[root@arch ~]# rsync -avxHAXS --delete --progress /boot /mnt/new-raid/boot<br />
[root@arch ~]# rsync -avxHAXS --delete --progress /home /mnt/new-raid/home<br />
<br />
Alternatively, you can use tar instead of the above rsync command if you prefer. rsync will, however, be quicker if you are only copying over changes. The tar command is: <code>tar -C / -clspf - . | tar -xlspvf -</code><br />
<br />
=== Update GRUB legacy ===<br />
Use your preferred text editor to open {{ic|/mnt/new-raid/boot/grub/menu.lst}}.<br />
<br />
--- SNIP ---<br />
default 0<br />
color light-blue/black light-cyan/blue<br />
<br />
## fallback<br />
fallback 1<br />
<br />
# (0) Arch Linux<br />
title Arch Linux - Original Disc<br />
root (hd0,0)<br />
kernel /vmlinuz26 root=/dev/sda1<br />
<br />
# (1) Arch Linux<br />
title Arch Linux - New RAID<br />
root (hd1,0)<br />
#kernel /vmlinuz26 root=/dev/sda1 ro<br />
kernel /vmlinuz26 root=/dev/md0 md=0,/dev/sda2,/dev/sdb2<br />
--- SNIP ---<br />
Notice we added the {{Ic|fallback}} line and duplicated the Arch Linux entry with a different {{Ic|root}} directive on the kernel line.<br />
<br />
Also update the "kopt" and "groot" sections, as shown below, if they are in your {{ic|/mnt/new-raid/boot/grub/menu.lst}} file, because it will make applying distribution kernel updates easier:<br />
- # kopt=root=UUID=fbafab1a-18f5-4bb9-9e66-a71c1b00977e ro<br />
+ # kopt=root=/dev/md0 ro md=0,/dev/sda2,/dev/sdb2<br />
<br />
## default grub root device<br />
## e.g. groot=(hd0,0)<br />
- # groot=(hd0,0)<br />
+ # groot=(hd0,1)<br />
<br />
=== Update Grub2 ===<br />
Please refer to https://wiki.archlinux.org/index.php/GRUB2#Other_Options<br />
<br />
=== Alter fstab ===<br />
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.<br />
<br />
To find the UUID to use:<br />
[root@arch ~]# blkid<br />
/dev/sda1: TYPE="swap" UUID="34656682b-34ad-8ed5-9233-dfab42272212" <br />
/dev/sdb1: UUID="9ff5682b-d5a1-4ed5-8d63-d1df911e0142" TYPE="swap" LABEL="NEW-SWAP" <br />
/dev/md0: UUID="6f2ea3d3-d7be-4c9d-adfa-dbeeedaf128e" SEC_TYPE="ext2" TYPE="ext3" LABEL="RAID-ONE" <br />
/dev/sda2: UUID="13dd2227-6592-403b-931a-7f3e14a23e1f" TYPE="ext2" <br />
/dev/sdb2: UUID="b28813e7-15fc-d4aa-dc8a-e2c1de641df1" TYPE="mdraid" <br />
<br />
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.<br />
<br />
[root@arch ~]# cat /mnt/new-raid/etc/fstab<br />
/dev/md0 / ext3 defaults 0 1<br />
UUID=9ff5682b-d5a1-4ed5-8d63-d1df911e0142 none swap sw 0 0<br />
<!-- /dev/sdb1 swap swap defaults 0 0 --><br />
<br />
=== Rebuild initcpio or initramfs ===<br />
<br />
==== First chroot into the RAID system ====<br />
<br />
[root@arch ~]# mount --bind /sys /mnt/new-raid/sys<br />
[root@arch ~]# mount --bind /proc /mnt/new-raid/proc<br />
[root@arch ~]# mount --bind /dev /mnt/new-raid/dev<br />
[root@arch ~]# chroot /mnt/new-raid/<br />
[root /]# <br />
<br />
If the chroot command gives you an error like <code>chroot: failed to run command `/bin/zsh': No such file or directory</code>, then use <code>chroot /mnt/new-raid/ /bin/bash</code> instead.<br />
<br />
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, regardless of the Linux variant).<br />
<br />
==== Record mdadm's config ====<br />
<br />
For Arch Linux, use "/etc/mdadm.conf", for Ubuntu or Debian, use "/etc/mdadm/mdadm.conf"<br />
nano /etc/mdadm.conf<br />
... and change the "MAILADDR" line to be your email address, if you want emailed alerts of problems with the RAID-1.<br />
<br />
Then save the array configuration with UUIDs to make it easier for the system to find /dev/md0 on boot-up. If you do not do this, you can get an "ALERT! /dev/md0 does not exist" error when booting :<br />
mdadm --detail --scan >> /etc/mdadm.conf<br />
<br />
==== For Arch Linux: Rebuild initcpio ====<br />
<br />
Edit {{ic|/etc/mkinitcpio.conf}} to include {{Ic|mdadm}} in the HOOKS array. Place it after {{Ic|autodetect}}, {{Ic|sata}}, {{Ic|scsi}} and {{Ic|pata}} (whichever is appropriate for your hardware).<br />
[root /]# mkinitcpio -g /boot/kernel26.img<br />
[root /]# exit<br />
<br />
==== For Ubuntu or Debian: Rebuild initramfs ====<br />
<br />
Then rebuild initramfs, incorporating the two above changes:<br />
update-initramfs -k `uname -r` -c -t<br />
<br />
This will rebuild your running version - to rebuild others, this will show a listing (in Ubuntu):<br />
ls /boot/ | perl -lne "/^[A-z\.\-]+/m && print $'" | egrep -e 'openvz$|generic$|server$' | sort -u<br />
Then substitute/script in these others so that all are available for use with the new RAID setup.<br />
<br />
=== Install GRUB on the RAID array ===<br />
Start grub:<br />
<br />
[root@arch ~]# grub --no-floppy<br />
<br />
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:<br />
<br />
grub> find /boot/grub/stage1<br />
(hd0,0)<br />
(hd1,1)<br />
<br />
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.<br />
<br />
grub> device (hd0) /dev/sdb<br />
<br />
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:<br />
<br />
grub> root (hd0,1)<br />
Filesystem type is ext2fs, partition type 0xfd<br />
grub> setup (hd0)<br />
Checking if "/boot/grub/stage1" exists... yes<br />
Checking if "/boot/grub/stage2" exists... yes<br />
Checking if "/boot/grub/e2fs_stage1_5" exists... yes<br />
Running "embed /boot/grub/e2fs_stage1_5 (hd0)"... 16 sectors are embedded. succeeded<br />
Running "install /boot/grub/stage1 (hd0) (hd0)1+16 p (hd0,1)/boot/grub/stage2 /boot/grub/grub.conf"... succeeded<br />
Done<br />
grub> quit<br />
<br />
== Verify success ==<br />
Reboot your computer, making sure it boots from the new RAID disk ({{ic|/dev/sdb}}) and not the original disk ({{ic|/dev/sda}}). You may need to change the boot device priorities in your BIOS to do this.<br />
<br />
Once the GRUB on the '''new''' disk loads, make sure you select to boot the new entry you created in {{ic|menu.lst}} earlier.<br />
<br />
Verify you have booted from the RAID array by looking at the output of mount. Also check mdstat again only to confirm which disk is in the array.<br />
[root@arch ~]# mount<br />
/dev/md0 on / type ext3 (rw)<br />
<br />
[root@arch ~]# cat /proc/mdstat<br />
Personalities : [linear] [raid0] [raid1] [raid5] [multipath] [raid6] [raid10]<br />
md0 : active raid1 sdb2[1]<br />
40064 blocks [2/1] [_U]<br />
<br />
unused devices: <none><br />
<br />
<br />
Also {{Ic|swapon -s}}:<br />
[root@arch ~]# swapon -s<br />
Filename Type Size Used Priority<br />
/dev/sdb1 partition 4000144 16 -1<br />
Note it is the swap partition on {{ic|sdb}} that is in use, nothing from {{ic|sda}}.<br />
<br />
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.<br />
<br />
== Add original disk to array ==<br />
<br />
=== Partition original disk ===<br />
<br />
Copy the partition table from /dev/sdb (newly implemented RAID disk) to /dev/sda (second disk we are adding to the array) so that both disks have exactly the same layout.<br />
[root@arch ~]# sfdisk -d /dev/sdb | sfdisk /dev/sda<br />
<br />
Alternate method - this will output the /dev/sdb partition layout to a file, then it's used as input for partitioning /dev/sda.<br />
[root@arch ~]# sfdisk -d /dev/sdb > raidinfo-partitions.sdb<br />
[root@arch ~]# sfdisk /dev/sda < raidinfo-partitions.sdb<br />
Use the --force if needed.<br />
[root@arch ~]# sfdisk --force /dev/sda < raidinfo-partitions.sdb<br />
<br />
Verify that the partitioning is identical:<br />
[root@arch ~]# fdisk -l<br />
<br />
==== Note ====<br />
If you get an error when attempting to add the parition to the array: <br />
mdadm: /dev/sda1 not large enough to join array<br />
You might have seen an earlier warning message when partitioning this disk that the kernel<br />
still sees the old disk size - a reboot ought to fix this, then try adding again to the array.<br />
<br />
=== Add disk partition to array ===<br />
[root@arch ~]# mdadm /dev/md0 -a /dev/sda2<br />
mdadm: hot added /dev/sda2<br />
<br />
Verify that the RAID array is being rebuilt.<br />
[root@arch ~]# cat /proc/mdstat<br />
Personalities : [linear] [raid0] [raid1] [raid5] [multipath] [raid6] [raid10]<br />
md0 : active raid1 sda2[2] sdb2[1]<br />
80108032 blocks [2/1] [_U]<br />
[>....................] recovery = 1.2% (1002176/80108032) finish=42.0min speed=31318K/sec<br />
<br />
unused devices: <none><br />
<br />
<br />
Syncing can take a while. If the machine is not needed for other tasks the speed limit can be increased.<br />
<br />
[root@arch ~]# cat /proc/mdstat <br />
Personalities : [raid1] <br />
md0 : active raid1 sda3[2] sdb3[1]<br />
155042219 blocks super 1.2 [2/1] [_U]<br />
[>....................] recovery = 0.0% (77696/155042219) finish=265.8min speed=9712K/sec<br />
<br />
unused devices: <none><br />
<br />
Check the current speed limit.<br />
<br />
[root@arch ~]# cat /proc/sys/dev/raid/speed_limit_min <br />
1000<br />
[root@arch ~]# cat /proc/sys/dev/raid/speed_limit_max<br />
200000<br />
<br />
Increase the limits.<br />
<br />
[root@arch ~]# echo 400000 >/proc/sys/dev/raid/speed_limit_min<br />
[root@arch ~]# echo 400000 >/proc/sys/dev/raid/speed_limit_max<br />
<br />
Then check out the syncing speed and estimated finish time.<br />
<br />
[root@arch ~]# cat /proc/mdstat <br />
Personalities : [raid1] <br />
md0 : active raid1 sda3[2] sdb3[1]<br />
155042219 blocks super 1.2 [2/1] [_U]<br />
[>....................] recovery = 1.3% (2136640/155042219) finish=158.2min speed=16102K/sec<br />
<br />
unused devices: <none><br />
<br />
=== Add second swap partition ===<br />
The partition was created with sfdisk, but it still has to be formatted for swap.<br />
[root@arch ~]# mkswap -L SWAP /dev/sda1<br />
Setting up swapspace version 1, size = 271314 kB<br />
LABEL=SWAP, UUID=1acd55dc-f73f-4639-94bc-3f30c33710c9<br />
<br />
Then add this UUID to the fstab exactly like the other one earlier. When done, it should look similar to this:<br />
[root@arch ~]# cat /mnt/new-raid/etc/fstab<br />
/dev/md0 / ext3 defaults 0 1<br />
UUID=1acd55dc-f73f-4639-94bc-3f30c33710c9 none swap sw 0 0<br />
UUID=9ff5682b-d5a1-4ed5-8d63-d1df911e0142 none swap sw 0 0<br />
<br />
It can be activated immediately:<br />
[root@arch ~]# swapon /dev/sda1<br />
<br />
=== Verify that email alerts are working ===<br />
<br />
If you run this command, then you should get a notification email showing the contents of /proc/mdstat :<br />
[root@arch ~]# mdadm --monitor --test --oneshot /dev/md0<br />
Check that you get the test email notification. This way you can be aware if one of the disks in the array fails (otherwise it may fail silently, putting you at risk of data loss if another drive should also fail). If you do not get an email notification, check the "MAILADDR" line in mdadm.conf, and also check that sending an email to this address from the command line (e.g. using the "mail" command) works.<br />
<br />
== Automatic conversion tool alternative - Raider ==<br />
<br />
You may consider to use '''Raider''', which is a tool that is able to convert a single disk in to a Raid system ('''1, 4, 5, 6''' or '''10''') with a two-pass command.<br />
<br />
*Website: http://raider.sourceforge.net/<br />
*It is also available in AUR: {{AUR|Raider}}</div>Mcarreira