Issue

• We have pods running on the master node even though scheduling is disabled:

# oc get nodes
NAME                     STATUS                     AGE       VERSION
master.example.com   Ready,SchedulingDisabled   5d        v1.7.6+a08f5eeb62
node1.example.com   Ready                      5d        v1.7.6+a08f5eeb62
node2.example.com   Ready                      5d        v1.7.6+a08f5eeb62

# oc get pods -o wide --all-namespaces | grep master
test                                test-website-2-33j41                                          1/2       Error              966        3d        10.0.0.1      master.example.com
test                                fluentd-pod-1-1i44j                                           1/1       Running            6          3d        10.0.0.2      master.example.com
test                                another-example-pod-3-ffjae                                   1/1       Running            0          3d        10.0.0.3      master.example.com

• I marked a node unschedulable with below command but pod continues to run there:

# oc adm manage-node <nodename> --schedulable=false

• I cordoned a node but continue to run pod deployments there