In this HowTo we build on the Cluster Provisioning with CAPI guide to demonstrate two
useful operations in the lifecycle of the CAPI management cluster:
- the upgrade of providers
- the migration of the management workloads to another cluster
We assume we already have a management cluster and the infrastructure provider configured as described in the Cluster Provisioning with CAPI guide. The selected infrastructure provider is AWS. We have not yet called
clusterctl init to initialise the cluster.
To demonstrate the steps of upgrading the management cluster, we will begin by initialising an old version of the MicroK8s CAPI providers.
To set the version of the providers to be installed we use the following notation:
clusterctl init --bootstrap microk8s:v0.4.0 --control-plane microk8s:v0.4.0 --infrastructure aws
With clusterctl we can check if there are any new versions on the running providers:
clusterctl upgrade plan
The output shows the existing version of each provider as well as the version that we can upgrade into:
NAME NAMESPACE TYPE CURRENT VERSION NEXT VERSION bootstrap-microk8s capi-microk8s-bootstrap-system BootstrapProvider v0.4.0 v0.5.0 control-plane-microk8s capi-microk8s-control-plane-system ControlPlaneProvider v0.4.0 v0.5.0 cluster-api capi-system CoreProvider v1.3.5 Already up to date infrastructure-aws capa-system InfrastructureProvider v2.0.2 Already up to date
To apply the upgrade plan recommended by
clusterctl upgrade plan simply:
clusterctl upgrade apply --contract v1beta1
To upgrade each provider one by one, issue:
clusterctl upgrade apply --bootstrap capi-microk8s-bootstrap-system/microk8s:v0.5.0 clusterctl upgrade apply --control-plane capi-microk8s-control-plane-system/microk8s:v0.5.0
Management cluster migration is a really powerful operation in the cluster’s lifecycle as it allows admins
to move the management cluster in a more reliable substrate or perform maintenance tasks without disruptions.
In the Cluster Provisioning with CAPI guide we showed how to provision a workloads cluster. Here, we start from the point where the workloads cluster is available and we will migrate the management cluster to the one cluster we just provisioned.
Before migrating a cluster, we must make sure that both the target and source management clusters run the same version of providers (infrastructure, bootstrap, control plane). To do so
clusterctl init should be called against he target cluster:
clusterctl get kubeconfig <provisioned-cluster> > targetconfig clusterctl --kubeconfig=$PWD/targetconfig --bootstrap microk8s --control-plane microk8s --infrastructure aws
clusterctl move --to-kubeconfig=$PWD/targetconfig