Skip to content

General Notes

Rebooting Nodes

  1. Cordon nodes so that no new pods are schedulable
  2. Drain the nodes prior to restarting. 
    oc adm drain bmgpu-03.ocp.mlopslab.ncs --delete-emptydir-data --ignore-daemonsets --disable-eviction --force
  3. SSH into core@bm hostname or through OpenShift Console > Node > bm hostname > Terminal > chroot /host > sudo su - > sudo reboot now
  4. Once node is back up, mark as schedulable.

MCP

 watch oc logs deploy/machine-config-controller -n openshift-machine-config-operator
Delete any pods that might be stuck

Restarting the nodes

  1. Worker nodes first, then Master nodes.
  2. When turning them back on, cluster dependencies first like Bastion, Firewall etc.