- 
                Notifications
    
You must be signed in to change notification settings  - Fork 9
 
Description
Hi,
I find the upgrade instructions a bit vague. It says:
In order to perform a minor or major upgrade without downtime you need to terminate all EC2 instances one-by-one.
So the suggestion is to do a "rolling update", if I understand correctly. The senza "version" would remain the same, as this is used as a service identifier / DNS entry in this case that should remain stable.
Could you elaborate how such a rolling update would be performed? The quote above is confusing, as terminating EC2 instances by hand would just result in new ones being created, because they belong to the same autoscaling group and it would attempt to maintain the number of nodes.
So you'd have to manually add an EC2 instance with a newer etcd version first, and then remove one with the old version?
Could you maybe elaborate a bit on, or even provide an example for, the update of the etcd cluster?