kata-deploy
provides a Dockerfile, which contains all of the binaries
and artifacts required to run Kata Containers, as well as reference DaemonSets, which can
be utilized to install Kata Containers on a running Kubernetes cluster.
Note: installation through DaemonSets successfully installs
katacontainers.io/kata-runtime
on a node only if it uses either containerd or CRI-O CRI-shims.
For your k3s cluster, run:
$ git clone https://github.com/kata-containers/kata-containers.git
Check and switch to the stable branch of your choice, if wanted, and then run:
$ cd kata-containers/tools/packaging/kata-deploy
$ kubectl apply -f kata-rbac/base/kata-rbac.yaml
$ kubectl apply -k kata-deploy/overlays/k3s
For your RKE2 cluster, run:
$ git clone https://github.com/kata-containers/kata-containers.git
Check and switch to the stable branch of your choice, if wanted, and then run:
$ cd kata-containers/tools/packaging/kata-deploy
$ kubectl apply -f kata-rbac/base/kata-rbac.yaml
$ kubectl apply -k kata-deploy/overlays/rke2
For your k0s cluster, run:
$ git clone https://github.com/kata-containers/kata-containers.git
Check and switch to "main", and then run:
$ cd kata-containers/tools/packaging/kata-deploy
$ kubectl apply -f kata-rbac/base/kata-rbac.yaml
$ kubectl apply -k kata-deploy/overlays/k0s
The supported version of k0s is v1.27.1+k0s and above, since the k0s support leverages a special dynamic containerd configuration mode:
From 1.27.1 onwards k0s enables dynamic configuration on containerd CRI runtimes. This works by k0s creating a special directory in /etc/k0s/containerd.d/ where user can drop-in partial containerd configuration snippets.
k0s will automatically pick up these files and adds these in containerd configuration imports list. If k0s sees the configuration drop-ins are CRI related configurations k0s will automatically collect all these into a single file and adds that as a single import file. This is to overcome some hard limitation on containerd 1.X versions. Read more at containerd#8056
However, this would also require a magic string set in the beginning of the line for /etc/k0s/containerd.toml
:
# k0s_managed=true
The latest image refers to pre-release and release candidate content. For stable releases, please, use the "stable" instructions.
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-rbac/base/kata-rbac.yaml
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-deploy/base/kata-deploy.yaml
The stable image refers to the last stable releases content.
Note: if you use a tagged version of the repo, the stable image does match that version. For instance, if you use the 2.2.1 tagged version of the kata-deploy.yaml file, then the version 2.2.1 of the kata runtime will be deployed.
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-rbac/base/kata-rbac.yaml
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-deploy/base/kata-deploy-stable.yaml
$ kubectl -n kube-system wait --timeout=10m --for=condition=Ready -l name=kata-deploy pod
Workloads specify the runtime they'd like to utilize by setting the appropriate runtimeClass
object within
the Pod
specification. The runtimeClass
examples provided define a node selector to match node label katacontainers.io/kata-runtime:"true"
,
which will ensure the workload is only scheduled on a node that has Kata Containers installed
runtimeClass
is a built-in type in Kubernetes. To apply each Kata Containers runtimeClass
:
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/runtimeclasses/kata-runtimeClasses.yaml
The following YAML snippet shows how to specify a workload should use Kata with Dragonball
:
spec:
template:
spec:
runtimeClassName: kata-dragonball
The following YAML snippet shows how to specify a workload should use Kata with Cloud Hypervisor:
spec:
template:
spec:
runtimeClassName: kata-clh
The following YAML snippet shows how to specify a workload should use Kata with Firecracker:
spec:
template:
spec:
runtimeClassName: kata-fc
The following YAML snippet shows how to specify a workload should use Kata with QEMU:
spec:
template:
spec:
runtimeClassName: kata-qemu
To run an example with kata-dragonball
:
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/examples/test-deploy-kata-dragonball.yaml
To run an example with kata-clh
:
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/examples/test-deploy-kata-clh.yaml
To run an example with kata-fc
:
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/examples/test-deploy-kata-fc.yaml
To run an example with kata-qemu
:
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/examples/test-deploy-kata-qemu.yaml
The following removes the test pods:
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/examples/test-deploy-kata-dragonball.yaml
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/examples/test-deploy-kata-clh.yaml
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/examples/test-deploy-kata-fc.yaml
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/examples/test-deploy-kata-qemu.yaml
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-deploy/base/kata-deploy.yaml
$ kubectl -n kube-system wait --timeout=10m --for=delete -l name=kata-deploy pod
After ensuring kata-deploy has been deleted, cleanup the cluster:
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-cleanup/base/kata-cleanup.yaml
The cleanup daemon-set will run a single time, cleaning up the node-label, which makes it difficult to check in an automated fashion. This process should take, at most, 5 minutes.
After that, let's delete the cleanup daemon-set, the added RBAC and runtime classes:
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-cleanup/base/kata-cleanup.yaml
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-rbac/base/kata-rbac.yaml
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/runtimeclasses/kata-runtimeClasses.yaml
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-deploy/base/kata-deploy-stable.yaml
$ kubectl -n kube-system wait --timeout=10m --for=delete -l name=kata-deploy pod
After ensuring kata-deploy has been deleted, cleanup the cluster:
$ kubectl apply -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-cleanup/base/kata-cleanup-stable.yaml
The cleanup daemon-set will run a single time, cleaning up the node-label, which makes it difficult to check in an automated fashion. This process should take, at most, 5 minutes.
After that, let's delete the cleanup daemon-set, the added RBAC and runtime classes:
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-cleanup/base/kata-cleanup-stable.yaml
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/kata-rbac/base/kata-rbac.yaml
$ kubectl delete -f https://raw.githubusercontent.com/kata-containers/kata-containers/main/tools/packaging/kata-deploy/runtimeclasses/kata-runtimeClasses.yaml
The Dockerfile used to create the container image deployed in the DaemonSet is provided here. This image contains all the necessary artifacts for running Kata Containers, all of which are pulled from the Kata Containers release page.
Host artifacts:
cloud-hypervisor
,firecracker
,qemu
, and supporting binariescontainerd-shim-kata-v2
(go runtime and rust runtime)kata-collect-data.sh
kata-runtime
Virtual Machine artifacts:
kata-containers.img
andkata-containers-initrd.img
: pulled from Kata GitHub releases pagevmlinuz.container
andvmlinuz-virtiofs.container
: pulled from Kata GitHub releases page
Two DaemonSets are introduced for kata-deploy
, as well as an RBAC to facilitate
applying labels to the nodes.
This DaemonSet installs the necessary Kata binaries, configuration files, and virtual machine artifacts on
the node. Once installed, the DaemonSet adds a node label katacontainers.io/kata-runtime=true
and reconfigures
either CRI-O or containerd to register three runtimeClasses
: kata-clh
(for Cloud Hypervisor isolation), kata-qemu
(for QEMU isolation),
and kata-fc
(for Firecracker isolation). As a final step the DaemonSet restarts either CRI-O or containerd. Upon deletion,
the DaemonSet removes the Kata binaries and VM artifacts and updates the node label to katacontainers.io/kata-runtime=cleanup
.
This DaemonSet runs of the node has the label katacontainers.io/kata-runtime=cleanup
. These DaemonSets removes
the katacontainers.io/kata-runtime
label as well as restarts either CRI-O or containerd
systemctl
daemon. You cannot execute these resets during the preStopHook
of the Kata installer DaemonSet,
which necessitated this final cleanup DaemonSet.