Amazon just announced two related features for getting super-fast, consistent performance with EBS volumes: (1) Provisioned IOPS EBS volumes, and (2) EBS-Optimized Instances.
Starting new instances and EBS volumes with these features is fine, but what if you already have some running instances you’d like to upgrade for faster and more consistent disk performance?
Given the two AWS features, there are two separate powers that need to be engaged to take full advantage:
Convert the EBS volume(s) from standard EBS volumes into new Provisioned IOPS EBS volume(s).
Convert the standard EC2 instance into an EBS-Optimized instance.
This article demonstrates how to take an existing EBS boot instance that is already running and convert it to use both of these two EBS performance features. Note that there will be some increased costs; please study Amazon’s published pricing before attempting.
For this demo, we start a temporary EBS boot instance (Ubuntu 12.04 LTS). Save the instance id and EBS volume id:
zone=us-east-1d ec2-run-instances --availability-zone $zone --key $USER --instance-type m1.small ami-013f9768 instance_id=... ec2-describe-instances $instance_id volume_id=...
Here are the steps to take a running EBS boot instance and convert it into an EBS-Optimized Instance with a Provisioned IOPS EBS volume.
Stop the EC2 instance (and wait for it to stop):
Detach the original (non-Provisioned IOPS) EBS volume(s):
Snapshot the original EBS volume(s) and save the snapshot ids:
ec2-create-snapshot $volume_id snapshot_id=...
Create Provisioned IOPS EBS volume from the snapshot(s) in the same availability zone as the instance, specifying the new size in GB, and specifying the IOPS level that you require:
ec2-create-volume --availability-zone $zone --size 10 --type io1 --iops 100 --snapshot $snapshot_id new_volume_id=...
Attach new Provisioned IOPS EBS volume(s) to the instance:
ec2-attach-volume --instance $instance_id --device /dev/sda1 $new_volume_id
If the instance type is not already one of the ones that supports EBS-Optimized instances, then you’ll need to change it to one that is. For this example, we’ll use
ec2-modify-instance-attribute --instance-type m1.large $instance_id
Convert the EC2 instance to EBS-Optimized:
ec2-modify-instance-attribute --ebs-optimized True $instance_id
Start the EBS-Optimized EC2 instance with its new, attached Provisioned IOPS EBS volume(s):
If you had an Elastic IP address associated with the instance before you stopped it, now’s the time to re-associate it.
When you’re comfortable with the new provisioned IOPS EBS volume, delete the original EBS volume and its snapshot:
ec2-delete-volume $volume_id ec2-delete-snapshot $snapshot_id
Terminate any test instance you started to experiment with in this demo:
Since you manually created the new EBS volume that was attached to the test instance, it will not be automatically deleted when the instance is terminated, so you must delete it manually:
In order for these commands to work with these features, you must be running the latest version of the EC2 API command line tools (or at least v126.96.36.199).
The new Provisioned IOPS EBS volume must be at least 10 GB.
The size of the new Provisioned IOPS EBS volume in GB must be at least 1/10th the value of the IOPS you are requesting. For example, 1000 IOPS requires an EBS volume of at least 100 GB.
If you’re running an official Ubuntu AMI, then your root file system will automatically be extended to the new size of the EBS volume. Other distros might need a little
xfs_growfsto get the benefit.