For those who have been using kops for a while should know the upgrade from 1.11 to 1.12 poses a greater risk, as it will upgrade etcd2 to etcd3.
Since this upgrade is disruptive to the control plane (master nodes), although brief, it's still something we take very seriously because nearly all the Buffer production services are running on this single cluster. We felt a more thorough backup process than the currently implemented Heptio Velero was needed.
To my surprises, my Google searches didn't yield any useful result on how to carry out the backup steps. To be fair, there are a few articles that are specifically for backing up master nodes created by kubeedm but nothing too concrete for `kop