Amazon EC2 Documentation confusion

I recently received a notification from Amazon EC2 to the effect that one of my server instances was running on degraded hardware, and that if I didn’t take action, this would result in loss of the instance.

The Cloud Computing service offered by Amazon EC2 is a great service, but it does throw up new challenges that do not exist in more traditional hosting environments. The tendency for server instances to simply ‘disappear’ is one of them, and you need to have your Cloud Computing setup well organised to ensure that disappearing instances don’t adversely impact on your business.

In this case, I was advised that if my instances was EBS-backed, all I had to do was stop and start my instance and that this would migrate my instance to stable hardware.

I try to use EBS-backed instances for all my servers in Amazon EC2, so this seemed like a fairly straightforward solution.

However, when I consulted the advisory link that Amazon provided in the relevant email, I was a little less certain. This states:

If an instance reboots (intentionally or unintentionally), the data on the instance store will survive.

However, under the following circumstances the data in the instance store will be lost:

o Failure of an underlying drive.

o Running an instance on degraded hardware.

o Stopping an Amazon EBS-backed instance.

o Terminating an instance.

To me, that suggested that if I stopped an EBS-backed instance, I may incur some data loss. That didn’t fit with my previous experience of using EBS-backed instances, which I’ve stopped and started frequently, without issue.

Any way, I went ahead and did the stop/start and everything worked out fine, which left me wondering what the warning was about.

The answer is this.

Whenever you deploy an Amazon EC2 instance, whether it is EBS-backed or Instance Store-backed, it comes with ephemeral storage. What differentiates an EBS-backed instance from an Instance Store-backed instance, is the nature of its root volume, not the totality of its storage.

That means that you can have an EBS-backed instance with an EBS root volume, but which might use ephemeral storage for other purposes, or in other words, and EBS-backed instance doesn’t mean that this instances uses only EBS storage.

This is the situation in the document referenced above. If you have an EBS-backed instance that uses its allocated ephemeral storage, and you stop/start that instance, the data on the ephemeral storage will be lost.

Leave a Reply

Your email address will not be published.

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>