OS Upgrades

• Use kubectl drain <node-name> this will terminate the pod on this node and create them on other node, making this node available for maintenance.

• After necessary maintainance, we have to do kubectl uncordon <node-name> to make it available for pod scheduling again.

• kubectl cordon <node-name> makes the node non-schedulable. does not terminate pods and stuff like drain.

• When the node is cordoned, the pods are moved (Not technically moved but terminated on that node & started on other) from that node to another.

• When the node is uncordoned the pods are are not moved back and only newly created pods will be placed on that node.

• If the node is forcefully drained, then the pod that is not part of repliaset will be lost permanently (So insted of draining it, we cordon it)

• When we drain the node, it is first cordoned and the pods are terminated.

• drain - Unschedulable & terminate pods

• cordon - Unschedulable

• uncordon - Make is schedulable again