I am working with Kubernetes 1.4.3 and my nodes look like the following:
ip-10-0-0-105.eu-central-1.compute.internal Ready 1d
ip-10-0-0-50.eu-central-1.compute.internal Ready,SchedulingDisabled 1d
ip-10-0-1-126.eu-central-1.compute.internal Ready 1dEven though the master node is set to SchedulingDisabled, Daemonsets are still being scheduled on it.
Firstly, why? this did not happen on prior to K8S 1.4, and if its a new, how do i cancel that option or maybe use pod affinity to exclude the master node from running daemonset pods.
Thanks.
This was answered in https://github.com/kubernetes/kubernetes/issues/29108#issuecomment-233432397
Basically, this is working as intended. DaemonSet pods will get scheduled on unschedulable nodes. In the future (not v1.4), this behavior will be selectable at the pod level (e.g., see https://github.com/kubernetes/kubernetes/issues/29178). For now, you can choose not to register your master node to avoid this problem.