Yes, we received that notice too. Amazon is rebooting many of it’s EC2 instances over the next few days. This is to apply some patches to the underlying host systems. There are some articles written with additional details it.
Unfortunately, the time windows in which the updates are going to happen are inconvenient to say the least.
Below are some tips that we have for minimizing the impact of this event.
Run multiple instances across multiple AZs which perform the same task. This is a general best-practice and not just a tip for this event. If you do this, then while one instance is being rebooted, the others are able to continue working. Ideally, no one would notice each instance’s reboot.
Not all instance types are affected. T1, T2, M2, R3, and HS1 instance types are not affected. Prior to your scheduled reboot, you can change the instance type of your instance to one of the unaffected instance types. Doing this during a time of your choosing will allow you to do it under controlled circumstances.
If you have an RDS instance which will be rebooted, then set it up as MultiAZ. This will minimize any downtime to a minute. Similar to tip #2 above, you could then change the instance class of your RDS instance so that any downtime is under controlled conditions.
If your VPC has a single NAT and it is scheduled for reboot, then you can create a new NAT using an unaffected instance type and redirect your VPC route tables to it. Then you can allow the scheduled reboot of the original instance to happen. Next week, redirect your VPC route tables back.
We drank our own Kool-aid and we’re happy to say that the Skeddly infrastructure will not be affected by the reboots.