I have a cluster of i2.2xlarge instances that I use to run performance tests. These are quite costly to run and I'd like to stop them when they are not in use. Unfortunately, the ephemeral storage on these instances does not survive the stop/start cycle. I am looking for a way to "snapshot" these instances, including all data on all volumes and then be able to bring them back to life when I need them.
Naively, I tried creating AMIs for each instance, but that does not seem to capture the storage volumes, so I can't even launch an instance from the AMI.
I would appreciate any pointers in the right direction.