If you have ever managed an etcd cluster on AWS and tried to keep it running reliably for several months, you certainly know that it might be a bit of a rocky road. For instance, the way etcd is designed does not natively allow it to be ran inside an auto-scaling group, so when your EC2 instance / NAT Gateway / AZ dies, you have no choice but to cleanup the membership and sync a new instance manually. It is Saturday night and you just lost your evening with the family. This is if you were smart enough to run a replica in each availability zone. Otherwise your cluster has likely already lost quorum.
Senior Engineer / Team Lead at CoreOS and in direct interaction with the etcd / etcd-operator authors, I decided to write a little project on my spare time to save you that kind of pain.
