Software raid - mdadm¶
2010 Archive
Danger
Create a backup of the data with the fingerprints before proceeding.
Note
It is better to use zpool because it is used on a large scale in production, plus it is easier to configure, stable and reliable.
Setup of RAID 1¶
We assume that the two 500GB hard drives are identical and install, format in ext4. We configure them so that it is in raid 1
Go to root
su -
Create a new logical devices
mknod /dev/md0 b 9 0
Install mdadm
apt install mdadm
Cluster creation. We indicate to mdadm the type of RAID we are doing (level = 1), the number of hard drives making up this cluster (raid-devices = 2), the name of the cluster (/dev/md0) and the partitions making up the cluster (/dev/sda1 and /dev/sdb1)
mdadm --create --verbose /dev/md0 --level=raid1 --raid-devices=2 /dev/sda1 /dev/sdb1
This displays
mdadm --create --verbose /dev/md1 --level=raid1 --raid-devices=2 /dev/sdb2 /dev/sdc2
mdadm: /dev/sdb2 appears to contain an ext2fs file system
size=293065760K mtime=Thu Jan 1 01:00:00 1970
mdadm: /dev/sdc2 appears to contain an ext2fs file system
size=293065760K mtime=Thu Jan 1 01:00:00 1970
mdadm: size set to 293065664K
Continue creating array? y
mdadm: array /dev/md1 started.
View raid status
mdadm --detail -scan
What should display this
ARRAY /dev/md0 level=raid1 num-devices=2 UID=b0be37d4:54202dc1:0e9efb29:244465d1
Now you have to wait for the raid to be built, you can see the progress details like this:
mdadm --detail /dev/md0
Mount the raid volume. It is necessary to Save the characteristics of the cluster.
mdadm --detail --scan
The previous command shows us the following line
ARRAY /dev/md0 level=raid1 num-devices=2 metadata=00.90 UUID=c06c3k41:gk63de5c:4e5m5245:5ead0852
This line is added to the /etc/mdadm/mdadm.conf file,
then you have to reboot the machine and enter everything in order
Delete a RAID volume¶
Turn off the volume
sudo mdadm --stop /dev/md0\nsudo mdadm --remove /dev/md0
Remove superblocks
sudo mdadm --zero-seperblock /dev/sda1
sudo mdadm --zero-seperblock /dev/sdb1
Check if the partitions have returned to “normal”
sudo mdadm --examine /dev/sda3\nmdadm : no md superberblock detected on /dev/sda3
You must then comment out the line concerning the cluster in the etc/mdadm/mdadm.conf file, otherwise there is an error when restarting.
Setting up RAID 5¶
We assume that three identical 500GB hard drives are installed. We will configure them so that it is in Raid 5
Run terminal as root
su -
Installing the mdadm package
apt install mdadm
Creating the Raid 5 volume
mdadm --create --verbose /dev/md0 --level=5 --assume-clean --raid-devices=4 /dev/sd[bcd]
⚠ --assume-clean: prevents synchronization after creation. This synchronization can last tens of hours depending on the size of the raid and as part of a new creation we will in any case format the volume afterwards.
Demonizing the RAID volume, we will make the system load the volume at each startup:
mdadm --daemonise /dev/md0
Format raid volume
mkfs.ext4 -j /dev/md0
For automatic editing add in /etc/fstab
/dev/md0 /media/raid ext4 defaults 0 1
Case study: loss of a disk¶
When the raid group is in “degraded” mode (missing disk). It will be assembled but not started, to force it use “–run” in addition to the other options, or once the raid group is assembled
mdadm --run /dev/md0
Examine the cluster
mdadm --detail /dev/md0
Add the replacement disk
mdadm /dev/md0 --add /dev/sde
Removing the failed hard drive
mdadm /dev/md0 --fail /dev/sda\nmdadm /dev/md0 --remove /dev/sda
Check raid reconstruction. Do not restart the computer until raid is re-built.
mdadm --detail /dev/md0
Case study: migrate data to a new machine¶
Consider that you want to install a pre-existing cluster on a machine that does not have a RAID volume
Install mdadm
apt install mdadm
Create a new FIFO file, identifier 0 may be different if it is already in use
mknod /dev/md0 b 9 0
Boot the old RAID as follows:
mdadm -A /dev/md1 --update=super-minor -m0 /dev/sd... /dev/sd...
To make the raid persistent
mdadm --daemonise /dev/md0