RAID-0 on EBS

raid0.sh

# Create volumes

for x in {1..6}; do \
  ec2-create-volume --size 100 -z us-east-1c --type io1 --iops 1000; \
done > volumes.txt

# Tag volumes for easy reference

VOLUME_LIST=$(ec2-describe-volumes \
| grep "2013-03-20" |  awk '{print $2}')

for volume in $(echo $VOLUME_LIST); do
  echo "Setting tags on the EBS volume ($volume)..."
	ec2-create-tags $volume -t "raid0"
done

# Attach the volumes to new instance
available_devs=("f" "g" "h" "i" "j" "k")
for vol in $(awk '{print $2}' volumes.txt); do \
  echo $vol -i i-2b74c149 -d /dev/sd${available_devs[i]}; \
  ec2-attach-volume $vol -i i-2b74c149 -d /dev/sd${available_devs[i]}; \
  i=$(( i + 1 )); \
done

# Install relevant dependencies
sudo yum -y install mdadm
sudo yum -y install xfsprogs

# Setup RAID array
sudo mdadm --create /dev/md0 --level 0 --chunk=256 --raid-devices=6 /dev/sdf /dev/sdg /dev/sdh /dev/sdi /dev/sdj /dev/sdk
mdadm: Defaulting to version 1.2 metadata
mdadm: array /dev/md0 started.

# Make it persistent
sudo mdadm --detail --scan --verbose > /etc/mdadm.conf

# See http://aws.amazon.com/amazon-linux-ami/faqs/#raid why raid uses /dev/md127 on reboot and rather use labels

# Create filesystem
sudo mkfs.xfs -L RAID0 /dev/md0 

# Next lets add the /dev/md0 drive to the file systems table for auto-mounting on reboot.
echo "LABEL=RAID0 /raiddrive xfs noatime 0 0" | sudo tee -a /etc/fstab

# Mount the drive
sudo mkdir /raiddrive
sudo mount /raiddrive

#Lets set the block device read ahead have to 64k which seems to really smooth out and improve the EBS drives.
sudo blockdev --setra 65536 /dev/md0

#Next lets verify the drive is operation
df -h /raiddrive

#Prewarm our RAID to avoid 5-50% IOPS penalty on 1st read (Think this is bogus so skipping in future setups)
dd if=/dev/md0 of=/dev/null