website/content/en/docs/tasks/administer-cluster/kubeadm/kubeadm-upgrade-ha.md

reviewers, title, content_template

reviewers	title	content_template
jamiehannaford luxas timothysc jbeda	Upgrading kubeadm HA clusters from v1.11 to v1.12	templates/task

{{% capture overview %}}

This page explains how to upgrade a highly available (HA) Kubernetes cluster created with kubeadm from version 1.11.x to version 1.12.x. In addition to upgrading, you must also follow the instructions in Creating HA clusters with kubeadm.

{{% /capture %}}

{{% capture prerequisites %}}

Before proceeding:

• You need to have a kubeadm HA cluster running version 1.11 or higher.
• Make sure you read the release notes carefully.
• Make sure to back up any important components, such as app-level state stored in a database. kubeadm upgrade does not touch your workloads, only components internal to Kubernetes, but backups are always a best practice.
• Check the prerequisites for Upgrading/downgrading kubeadm clusters between v1.11 to v1.12.

{{< note >}} Note: All commands on any control plane or etcd node should be run as root. {{< /note >}}

{{% /capture %}}

{{% capture steps %}}

Prepare for both methods

Upgrade kubeadm to the version that matches the version of Kubernetes that you are upgrading to:

apt-mark unhold kubeadm && \
apt-get update && apt-get install -y kubeadm && \
apt-mark hold kubeadm

Check prerequisites and determine the upgrade versions:

kubeadm upgrade plan

You should see something like the following:

Upgrade to the latest stable version:

COMPONENT            CURRENT   AVAILABLE
API Server           v1.11.3   v1.12.0
Controller Manager   v1.11.3   v1.12.0
Scheduler            v1.11.3   v1.12.0
Kube Proxy           v1.11.3   v1.12.0
CoreDNS              1.1.3     1.2.2
Etcd                 3.2.18    3.2.24

Stacked control plane nodes

Upgrade the first control plane node

Modify configmap/kubeadm-config for this control plane node:

kubectl get configmap -n kube-system kubeadm-config -o yaml > kubeadm-config-cm.yaml

Open the file in an editor and replace the following values:

• api.advertiseAddress

  This should be set to the local node's IP address.

• etcd.local.extraArgs.advertise-client-urls

  This should be updated to the local node's IP address.

• etcd.local.extraArgs.initial-advertise-peer-urls

  This should be updated to the local node's IP address.

• etcd.local.extraArgs.listen-client-urls

  This should be updated to the local node's IP address.

• etcd.local.extraArgs.listen-peer-urls

  This should be updated to the local node's IP address.

• etcd.local.extraArgs.initial-cluster

  This should be updated to include the hostname and IP address pairs for each control plane node in the cluster. For example:

    "ip-172-31-92-42=https://172.31.92.42:2380,ip-172-31-89-186=https://172.31.89.186:2380,ip-172-31-90-42=https://172.31.90.42:2380"
  

You must also pass an additional argument (initial-cluster-state: existing) to etcd.local.extraArgs.

kubectl apply -f kubeadm-config-cm.yaml --force

Start the upgrade:

kubeadm upgrade apply v<YOUR-CHOSEN-VERSION-HERE>

You should see something like the following:

[upgrade/successful] SUCCESS! Your cluster was upgraded to "v1.12.0". Enjoy!

The kubeadm-config ConfigMap is now updated from v1alpha2 version to v1alpha3.

Upgrading additional control plane nodes

Each additional control plane node requires modifications that are different from the first control plane node. Run:

kubectl get configmap -n kube-system kubeadm-config -o yaml > kubeadm-config-cm.yaml

Open the file in an editor and replace the following values for ClusterConfiguration:

• etcd.local.extraArgs.advertise-client-urls

  This should be updated to the local node's IP address.

• etcd.local.extraArgs.initial-advertise-peer-urls

  This should be updated to the local node's IP address.

• etcd.local.extraArgs.listen-client-urls

  This should be updated to the local node's IP address.

• etcd.local.extraArgs.listen-peer-urls

  This should be updated to the local node's IP address.

You must also modify the ClusterStatus to add a mapping for the current host under apiEndpoints.

Add an annotation for the cri-socket to the current node, for example to use docker:

kubectl annotate node <nodename> kubeadm.alpha.kubernetes.io/cri-socket=/var/run/dockershim.sock

Start the upgrade:

kubeadm upgrade apply v<YOUR-CHOSEN-VERSION-HERE>

You should see something like the following:

[upgrade/successful] SUCCESS! Your cluster was upgraded to "v1.12.0". Enjoy!

External etcd

Upgrade each control plane

Get a copy of the kubeadm config used to create this cluster. The config should be the same for every node. The config must exist on every control plane node before the upgrade begins.

# on each control plane node
kubectl get configmap -n kube-system kubeadm-config -o jsonpath={.data.MasterConfiguration} > kubeadm-config.yaml

Now run the upgrade on each control plane node one at a time.

kubeadm upgrade apply v1.12.0 --config kubeadm-config.yaml

Upgrade etcd

Kubernetes v1.11 to v1.12 only changed the patch version of etcd from v3.2.18 to v3.2.24. This is a rolling upgrade with no downtime, because you can run both versions in the same cluster.

On the first host, modify the etcd manifest:

sed -i 's/3.2.18/3.2.24/' /etc/kubernetes/manifests/etcd.yaml

Wait for the etcd process to reconnect. There will be error warnings in the other etcd node logs. This is expected.

Repeat this step on the other etcd hosts.

Next steps

Manually upgrade your CNI provider

Your Container Network Interface (CNI) provider might have its own upgrade instructions to follow. Check the addons page to find your CNI provider and see whether you need to take additional upgrade steps.

Update kubelet and kubectl packages

Upgrade the kubelet and kubectl by running the following on each node:

# use your distro's package manager, e.g. 'apt-get' on Debian-based systems
# for the versions stick to kubeadm's output (see above)
apt-mark unhold kubelet kubectl && \
apt-get update && \
apt-get install kubelet=<NEW-K8S-VERSION> kubectl=<NEW-K8S-VERSION> && \
apt-mark hold kubelet kubectl && \
systemctl restart kubelet

In this example a deb-based system is assumed and apt-get is used for installing the upgraded software. On rpm-based systems the command is yum install <PACKAGE>=<NEW-K8S-VERSION> for all packages.

Verify that the new version of the kubelet is running:

systemctl status kubelet

Verify that the upgraded node is available again by running the following command from wherever you run kubectl:

kubectl get nodes

If the STATUS column shows Ready for the upgraded host, you can continue. You might need to repeat the command until the node shows Ready.

If something goes wrong

If the upgrade fails, see whether one of the following scenarios applies:

• If kubeadm upgrade apply failed to upgrade the cluster, it will try to perform a rollback. If this is the case on the first master, the cluster is probably still intact.

  You can run kubeadm upgrade apply again, because it is idempotent and should eventually make sure the actual state is the desired state you are declaring. You can run kubeadm upgrade apply to change a running cluster with x.x.x --> x.x.x with --force to recover from a bad state.

• If kubeadm upgrade apply on one of the secondary masters failed, the cluster is upgraded and working, but the secondary masters are in an undefined state. You need to investigate further and join the secondaries manually.

{{% /capture %}}