While developing K3s to run at the edge we had to change our assumptions about how to manage these clusters at scale. A key assumption in a data center is that you have stable network connectivity, but this may not be true at the edge. You may have unreliable cellular service or limited time during the day in which you can connect. In these environments, operations such as upgrading Kubernetes or patching an operating system require a different paradigm.
To address this, Rancher Labs has been working on a solution we're calling the System Upgrade Controller. It enables an operator to initiate an upgrade remotely but manage it locally.In this meetup we will:
Take a technical deep dive of System Upgrade Controller.
See how it can update a standalone K3s cluster.
Import a K3s cluster and upgrade it from Rancher.
Explore how we want to use this to automate node patching.
