Caution
Before proceeding with the upgrade procedure, verify that you have updated DriveTrain including Aptly, Gerrit, Jenkins, Reclass, Salt formulas, and their subcomponents to the current MCP release version. Otherwise, the current MCP product documentation is not applicable to your MCP deployment.
This section describes how to automatically upgrade Kubernetes
to a major version or update to a minor version.
During this procedure, the downtime occurs only
while the kube-apiserver VIP of a node to be updated is moving to another
active node. The MCP cluster workload is not affected.
Caution
During the execution of the Kubernetes upgrade pipeline job:
The Kubernetes Docker container back end is replaced by containerd and all Kubernetes workloads are moved to containerd. However, the Docker service is not stopped and removed in the event some third-party Docker workloads that are not related to the MCP Kubernetes cluster can be running in Docker. If this is not the case and you do not need Docker anymore, you can disable it after the upgrade.
Starting from the MCP 2019.2.3 maintenance update, due to the
conflict between the docker-engine and the containerd
runc packages versions, Docker is removed
during the upgrade to prevent the conflict. Therefore, Mirantis
recommends migrating the third-party Docker workloads
running on the MCP Kubernetes cluster, if any,
before the upgrade.
If any DaemonSet was deployed on your cluster, all Kubernetes nodes will be rebooted to reflect the containerd changes to DaemonSets.
No third-party Docker workloads will be removed from Docker but any Docker workload will be stopped during a node reboot.
Starting from the MCP 2019.2.3 maintenance update, any third-party Docker workload running on the MCP Kubernetes cluster is stopped and removed along with Docker during a node reboot.
See also