I have the following scenario.

For various reasons I chose to use embedded tomcat for my rest backend that will form a cluster of AWS machines. I will want to deploy fairly frequently updates without service disruption. I have been reading on the web but am not perfectly sure about the methodology used.

As I understand I will deploy using some automation software (anything to recommend here or just plain old unix scripts?) to stop one by one a server of the cluster and redeploy the REST jar but while the server is down another one will be running (because of the cluster) and the user will not have a service interruption because of session sharing (and a load balancer?). Should I use a persisted session storage with jdbc for example.

Or is it that if a user is currently using the server I want to bring down for update will have an interruption no matter what.