We have a 4 node Kubernetes cluster, installed using Juju 2.0 and the official canonical Kubernetes charm. The sole purpose of this very local and private installation is development and testing.
One of the biggest and documented issues of this install is complete lack of a graceful server shutdown procedure. As only as etcd nodes go down, the whole config is lost.
We have spent a considerable amount of time looking for an automated backup / restore script / procedures for Kubernetes. The most frequent answer to the problem we have is to deploy the server from a scratch using Juju whenever we boot it. Given 30-35min. deployment time it doesn't seem like an attractive solution?
Do you happen to have any feasible solution to this problem?
what about backing up the etcd config to restore on start up?
https://coreos.com/etcd/docs/latest/v2/admin_guide.html
Doesn't that save you quite some time?