OS Upgrades

Summary (Formula);

• drain = terminate + schedule + cordoned
• cordon = cordoned

Introduction of OS Upgrades

Assume you have a cluster with 2 nodes. Now, you want to upgrade one of the nodes to a new version. If the node is down for more than 5 minutes, all the pods running on that node will be dead, else the pods will come back online when the kubectl process starts. If the pods are part of a replicaset, those pods will be provisioned on other nodes.

In this case, we need to upgrade the node in a safer way.

kubectl drain <node-name>
# For pods, you have to force it, but the pod will lose forever
kubectl drain <node-name> --force
kubectl drain <node-name> --ignore-daemonsets

What we are doing her is, we are draining the node. When we drain the node, the pods are gracefully terminated from the node and are scheduled on other nodes (it does not move from one node to another). Besides, the node is also marked as unschedulable or cordoned, that means no new pods will be scheduled on that node until you uncordon it.

kubectl uncordon <node-name>

When the node is come back online, you have to uncordon it, so that the node is not marked as unschedulable. Once you uncordon it, the pods can be scheduled on that node again.

One thing we have to keep in mind it, those pods that are scheduled on other nodes won't automatically come back to the original node.

kubectl cordon <node-name>

If you want to make the node unschedulable or cordoned, you can use this command. This will make sure that no new pods are scheduled on that node.