cluster.ninja

Upgrade hop

Kubernetes 1.17 → 1.18

1.17End of lifeEOL 13 Jan 2021 1.18End of lifeEOL 18 Jun 2021

1.18 is itself end of life. If you are on 1.17 today, plan the whole path to a supported version (at least 1.35) with the upgrade planner.

1 · Fix removed APIs first

No API versions are removed between 1.17 and 1.18.

2 · Other changes that can bite

Deprecated in 1.18

  • The kubernetes.io/ingress.class annotation is described as deprecated; the new ingressClassName field and IngressClass resource take its place.

Our summary of the 1.18 release announcement (CC BY 4.0). Read the full changelog for your components.

3 · Upgrade in this order

  1. Prepare. Move every node to 1.16 or newer if any are older: a 1.18 API server supports kubelets back to 1.16 only. Upgrade to the newest 1.17 patch, 1.17.17.
  2. kube-apiserver to 1.18 (in HA clusters, one instance at a time; old and new may differ by one minor during the roll). Use the newest patch, 1.18.20.
  3. kube-controller-manager, kube-scheduler, cloud-controller-manager to 1.18, once every API server is on 1.18.
  4. kubelet on each node (drain first). Optional in this hop: kubelets from 1.16 to 1.18 all work with a 1.18 control plane.
  5. kube-proxy: no newer than the API server, and within the same window as the kubelet it runs beside.
  6. kubectl: 1.17 to 1.19 all talk to a 1.18 API server.

From the official version skew policy. Managed services run the control-plane steps for you; node pools are still yours to schedule.