Where is the official documentation for Amazon EC2 rebooting instances for hardware changes?

I understand that the Amazon EC2 SLA says that EC2 guarantees a 99.95% uptime.

I've read in many places that systems built using EC2 should be designed to cope with individual instances being restarted e.g. ec2 rebooted my instance.

Where is the official Amazon document开发者_如何转开发ation to say that instances may be restarted?

I do not believe Amazon publishes any documentation on rebooting EC2 instances for hardware changes. Instead, they will send customers a notice if there is going to be scheduled maintenance performed on the system. However, I think the issue here is more a matter of servers crashing unexpectedly. That, of course, they cannot announce beforehand. Also, don't forget that they calculate their uptime based upon 5 minute increments so you may have downtime that isn't counted towards their SLA because it was less than the five minutes and didn't get noticed.

Here is a link to the official Amazon EC2 SLA (I'm sure you've seen it). They don't give any indication that maintenance ever affects systems running in production:

http://aws.amazon.com/ec2-sla/

You can contrast this with Amazon RDS, which specifically states what maintenance is and when it occurs:

http://aws.amazon.com/rds/faqs/#12

I would imagine that they expect to never have downtime because of hardware upgrades. Since everything is virtual, they can move live instances to new hardware without taking them down.