Skip to main content

Upgrading Kubernetes with kubeadm

In this guide we will use as an example upgrading from kubernetes v1.28.1 to v1.28.3

The the control-plane nodes must be upgraded first, followed by the worker nodes.

Upgrade Control-Plane Nodes​

Pick a control-plane node to be the first to upgrade, followed by:

1: Upgrading kubeadm and kubectl to the latest patch version of the desired major version:

apt-get update
apt-mark unhold kubeadm
apt-get install -y kubeadm='1.28.3-*'
apt-mark hold kubeadm

2: Verify the upgrade plan:

kubeadm upgrade plan v1.28.3

3: Drain the node:

kubectl drain <node-name> --ignore-daemonsets

4: Perform the upgrade:

kubeadm upgrade apply v1.28.3

5: Upgrade the node's CRI-O or other container runtime to an appropriate version if need be. For CRI-O that would be changing the minor version in the repositories added to /etc/apt/sources.list.d and then running:

apt-get update
apt-get install cri-o cri-o-runc
systemctl daemon-reload
systemctl restart crio

6: Upgrade kubelet and kubectl

apt-get update
apt-mark unhold kubelet
apt-mark unhold kubectl
apt-get install -y kubelet='1.28.3-*'
apt-get install -y kubectl='1.28.3-*'
apt-mark hold kubelet
apt-mark hold kubectl

7: Restart kubelet

systemctl daemon-reload
systemctl restart kubelet

8: Uncordon the node

kubectl uncordon <node-name>

9: Possibly, upgrade CNI. Particularly if it's a minor version upgrade there may be a need to update the CNI to a new version as well according to the vendor's release notes for the upgrade process

Upgrade remaining Control-Plane Nodes​

For the remaining control-plane nodes, execute steps 1 to 8, one at a time but:

  • step 2 is skipped, no need to plan the upgrade anymore
  • step 4 is replaced by:
kubeadm upgrade node

Upgrade Worker Nodes​

After all the control-plane nodes are upgraded, it's time to upgrade your worker nodes by following the previous steps from 1 to 8 but:

  • step 2 is skipped
  • step 4 is replaced by:
kubeadm upgrade node

Note: You can upgrade as many worker nodes in parallel as you see fit and/or find adequate to your availability requirements, as the nodes being upgraded will be drained from workloads.