Setting up EBS as a RAID

You are viewing a wiki page. You are welcome to join the group and then edit it. Be bold!

Posted by nullvariable on November 23, 2009 at 4:46pm
Last updated by Greg Coit on Thu, 2010-06-10 17:27

Very similar to setting up an instance with a single EBS drive, you can also fairly easily do a RAID configuration. This leads to better performance usually and higher reliability.

Step 1 - Collect the necessary data

Note the following:

instance ID
region (example: us-east-1). Also be sure to note the letter following it.
access key ID
secret access key.

Step 2 - Create (Provision) the Volumes

You can do this with command line tools or Elastic Fox, I've included a screen shot of the current AWS Management Console. Select the size you'll need the drive to be, choose the availability zone that matches the region your instance is in, including the letter as noted in step 1.

Repeat this step for the total number of EBS volumes you want in your RAID. When writing these instructions I used a 4 volume setup.

Step 3 - Connect the EC2 Instance to the EBS Volumes

Part 1:

attach the EC2 Instance to the EBS. Click the "Attach Volume" button and find the instance previously noted to link this drive to the server. Make sure you note the device name (example: /dev/sda) assigned to the volume, you'll need it in a minute.

Repeat this step for each volume you've created.

Part 2:

apt-get install -y xfsprogs

Part 3:

Now create an XFS file system on the EBS volume:

sudo modprobe xfs
sudo mkfs.xfs [device name]

Device name should look like "/dev/sda" (you can see the device name from the AWS console under "Attachment Information")

Repeat this step for each volume.

Step 4 - Create the RAID Drive

I used mdadm to create my software raid config, there are other tools too. You'll need to install it to your instance:

sudo apt-get install mdadm

I went with a RAID0 configuration here as EBS volumes should already be redundant. Additionally I took the precaution of setting up snapshots on a regular basis so should anything fail I could start back up relatively easily.

sudo mdadm --create /dev/md0 --raid-devices=[num devices] --level=0 [enter device names here like:] /dev/sdf /dev/sdg /dev/sdh /dev/sdi

Now we're done with that and you should have a RAID0 drive at /dev/md0 so let's mount it:

echo "/dev/md0 /vol xfs noatime 0 0" | sudo tee -a /etc/fstab
sudo mkdir -m 000 /vol
sudo mount /vol

replace /dev/md0 with the RAID device you used if it's different.

Congrats! You now have a working RAID0 EBS volume

Step 5 - Move the MySQL Install to the RAID device

Stop the MySQL server.

sudo /etc/init.d/mysql stop

Now move all the database files to the EBS volume.

mkdir /vol/etc /vol/lib /vol/log
mv /mnt/mysql     /vol/etc/
mv /var/lib/mysql /vol/lib/
mv /var/log/mysql /vol/log/

And point everything to the RAID device:

sudo mkdir /mnt/mysql
sudo mkdir /var/lib/mysql
sudo mkdir /var/log/mysql

echo "/vol/etc/mysql /mnt/mysql     none bind" | tee -a /etc/fstab
mount /mnt/mysql

echo "/vol/lib/mysql /var/lib/mysql none bind" | tee -a /etc/fstab
mount /var/lib/mysql

echo "/vol/log/mysql /var/log/mysql none bind" | tee -a /etc/fstab
mount /var/log/mysql

Start up your MySQL server again.

sudo /etc/init.d/mysql start

Now your MySQL files are secure against failure of the EC2 instance and they're running at a higher level of performance. Now you should consider setting up snapshots for these EBS volumes. These snapshots aren't just great backups, they can also be used to create another EBS volume (perfect for spooling up a test server with real content that's almost fresh).

Next step:

Consistent backups with RAID+ec2-consistent-snapshot+cron

This command might be helpful for restoring drives:

mdadm --assemble /dev/md0 /dev/sdh1 /dev/sdi1 /dev/sdj1 /dev/sdk1

Useful links: