diff --git a/source/clear-linux/tutorials/kubernetes/kubernetes-bp.rst b/source/clear-linux/tutorials/kubernetes/kubernetes-bp.rst index 3c241cb2..2afcd3ee 100644 --- a/source/clear-linux/tutorials/kubernetes/kubernetes-bp.rst +++ b/source/clear-linux/tutorials/kubernetes/kubernetes-bp.rst @@ -3,6 +3,9 @@ Kubernetes Best Practices on |CL| ################################# +Update Kubernetes clusters +************************** + This tutorial shows you how to manage your Kubernetes cluster while using :command:`swupd` to update |CL-ATTR|. @@ -13,12 +16,19 @@ RPM/DEB). In contrast, |CL| manages packages in bundles. This document describes best practices to manage cluster upgrades with kubeadm on a |CL|-based cluster. Prerequisites -************** +************* Let's review some basic assumptions: * |CL| includes Kubernetes support with the `cloud-native-basic` bundle. +* Ensure `autoupdate` is enabled. See :ref:`swupd-guide`. + To enable, run: + + .. code-block:: bash + + sudo swupd autoupdate + * Execute :command:`sudo swupd update` **on each node** to do the update. * Repeat the above command for both master and worker nodes. @@ -29,7 +39,7 @@ binaries simultaneously, which are part of `cloud-native-basic` bundle .. note:: - Oher Linux\* distros shown in Kubernetes upgrade documentation reflect + Other Linux\* distros shown in Kubernetes upgrade documentation reflect `apt-get update`, `apt-mark hold kubeadm`, and similar commands; however, such commands **aren not valid** on |CL|. @@ -49,20 +59,18 @@ Update the cluster cat /run/motd -#. Next, ....[Explain resultant actions/choices from above] +.. TODO: As of 01/29/19 the motd is only available if the k8s minor version changes. Under consideration that the motd will ALWAYS show when k8s changes. + +#. If the 'motd' indicates a change...., follow instructions below to update. .. note:: - Swupd post install script adds a message-of-the-day to warn/notify - users if the OS update installs a new kubernetes version that is not - compatible (according to kubernetes API guarantee) with the previous - version. If the `motd` appears, a kubelet restart on master and nodes - must be postponed until the control plane is properly updated. - :command:`swupd update` does not restart services automatically unless - explicitly configured to do so (be warned about making kubelet - automatically restartable). + Do not restart the kubelet before the control plane is updated. - .. note:: Do not restart the kubelet before the control plane is updated. + If the `motd` appears, a kubelet restart on master and nodes + **must be postponed** until the control plane is properly updated. + :command:`swupd update` does not restart services automatically unless + explicitly configured to do so. #. Now follow these instructions in kubernetes documentation. @@ -87,7 +95,7 @@ Update worker nodes * `Restart Kubelet and undrain node`_ -.. _Restart Keubelet and undrain node: https://kubernetes.io/docs/tasks/administer-cluster/kubeadm/kubeadm-upgrade-1-13/#restart-the-kubelet-for-all-nodes +.. _Restart Kubelet and undrain node: https://kubernetes.io/docs/tasks/administer-cluster/kubeadm/kubeadm-upgrade-1-13/#restart-the-kubelet-for-all-nodes .. _Update kubelet configuration: https://kubernetes.io/docs/tasks/administer-cluster/kubeadm/kubeadm-upgrade-1-13/#upgrade-the-kubelet-config-on-worker-nodes