Why deployment, not daemonset?

[holmesb]

Hi,

Any reason you decided to use a Deployment rather than a DaemonSet? DS would be more reliable IMO. For example, in a two-node cluster, if node A goes down, all pods in Deployment will be scheduled on node B. Then, when A returns, if B goes down, drain-node-on-crash is slower to function since it must first wait for a pod to be scheduled on A. Using a DaemonSet means there is already an available pod and drain-node-on-crash doesn't suffer this delay.

Thanks and nice work BTW.