!! BREAKING CHANGES !!
vcluster will now deploy coredns itself instead of depending on the k3s coredns deployment. This makes it possible to customize the coredns deployment (e.g. for rootless mode) and also support other kubernetes distributions besides k3s. However, this will break the current coredns deployment created by k3s.
This means that after you have upgraded vcluster from an v0.4.x version, you need to delete the coredns deployment inside vcluster and then restart vcluster by deleting the vcluster pod to let it redeploy the correct coredns deployment:
# Delete coredns deployment inside vcluster
export KUBECONFIG=vcluster-kubeconfig.yaml
kubectl delete deployment coredns -n kube-system
# Restart vcluster by deleting its pod
export KUBECONFIG=host-kubeconfig.yaml
kubectl delete pod vcluster-pod-0 -n vcluster-namespace
Support for k0s and vanilla k8s virtual Kubernetes Clusters
vcluster now supports creating k0s and vanilla k8s virtual Kubernetes clusters instead of k3s. You can now specify the target virtual Kubernetes cluster distribution via the --distro flag in vcluster create. For vanilla k8s there is now also a high availability mode.
Examples:
# Create a virtual k3s Kubernetes cluster
vcluster create my-vcluster -n my-vcluster
# Create a virtual k0s Kubernetes cluster
vcluster create my-vcluster -n my-vcluster --distro k0s
# Create a virtual k8s Kubernetes cluster with kubernetes version v1.20
vcluster create my-vcluster -n my-vcluster --distro k8s --kubernetes-version v1.20
Other Changes
- cli: Deprecation of
vcluster createflags--release-valuesand--k3s-image. Use--extra-valuesinstead - cli: Changed default value of
vcluster createflag--upgradefromtruetofalse. vcluster will not try to upgrade an existing vcluster by default anymore - cli: New flag
--kubernetes-versionforvcluster createto override the kubernetes version used for creating the vcluster. - chart: New helm chart to deploy vcluster with k0s instead of k3s (#167) (see docs for more information)
- chart: New helm chart to deploy vcluster with vanilla k8s instead of k3s (see docs for more information)
- chart: Added support for rootless mode (see docs for more information)
- chart: Added support for openshift v4 for k3s distro
- syncer: New flag
--namethat specifies the vcluster name (replaces--suffix) - syncer: Deprecated flag
--suffix, use--nameinstead - syncer: Flag
--service-namenow defaults to--nameif omitted - syncer: Flag
--set-ownernow defaults totrueand vcluster does not fail if setting the owner is not possible - syncer: Fixed an issue where nodes wouldn't get deleted correctly
- syncer: Fixed an issue where vcluster would return a conflict error instead of already exists during service create
- syncer: Fixed an error where vcluster would not correctly sync nodes and create node services (#214)
- syncer: Removed flag
--service-namespaceas this would break vcluster if it's not the same value as current vcluster namespace - syncer: Fixed an issue where glusterfs volume endpoints would not get correctly translated
- syncer: Fixed an issue where vcluster couldn't find the current pod if the flag
--target-namespacewas provided (#214) - syncer: New flag
--leader-electthat defaults to false to avoid unnecessary leader election - build: Removed file name from the sha256 files (thanks @developer-guy)