The service catalog apiserver and controller are deployed as daemonsets to the first master only. An HA mode cluster will be crippled if you shutdown the master which has these daemonsets deployed. IMHO, we should do one of two things: 1) make them normal pods so they can be migrated in cases of failure or 2) deploy the daemonset to each master (instead of only master01). [root@master01 ~]# oc get pods --all-namespaces -owide |grep catalog kube-service-catalog apiserver-jcb4w 1/1 Running 1 22h 10.129.0.16 master01.ocp.nicknach.net kube-service-catalog controller-manager-9pwg7 1/1 Running 2 22h 10.129.0.18 master01.ocp.nicknach.net one hour later, after shutting down master01: [root@master02 ~]# oc get pods --all-namespaces -owide |grep catalog kube-service-catalog apiserver-jcb4w 1/1 NodeLost 1 23h 10.129.0.16 master01.ocp.nicknach.net kube-service-catalog controller-manager-9pwg7 1/1 NodeLost 2 23h 10.129.0.18 master01.ocp.nicknach.net -Nick
*** This bug has been marked as a duplicate of bug 1554623 ***