The expected behavior is that the connection to the old master will close and that Vitess will try to reconnect to the new master. 


AWS/Aurora

To ensure that the expected behavior occurs when using AWS/Aurora you will need to set the vttablet flag -pool_hostname_resolve_interval to something other than the default. This is because the default is 0. When this flag is set to the default, Vitess will never re-resolve the AWS/Aurora DNS name.