Transition to containers¶

Transitioning from virtual machines to containers is a lengthy and complex process that in some environments may take years. If you want to leverage Kubernetes features while you continue using the existing applications that run in virtual machines, Mirantis provides an interim solution that enables running virtual machines orchestrated by Kubernetes.

To enable Kubernetes to run virtual machines, you need to deploy and configure a virtual machine runtime for Kubernetes called Virtlet. Virtlet is a Kubernetes Container Runtime Interface (CRI) implementation that is packaged as a Docker image and contains such components as libvirt daemon, QEMU/KVM wrapper, and so on.

Virtlet enables you to run unmodified QEMU/KVM virtual machines that do not include an additional containerd layer as in similar solutions in Kubernetes. Virtlet supports all standard Kubernetes objects, such as ReplicaSets, deployments, DaemonSets, and so on, as well as their operations. For information on operations with Kubernetes objects, see: Kubernetes documentation.

Unmodified QEMU/KVM virtual machines enable you to run:

• Unikernels
• Applications that are hard to containerize
• NFV workloads
• Legacy applications

Compared to regular Kubernetes pods, Virtlet pods have the following limitations:

• Only one virtual machine per pod is allowed.
• Virtual machine volumes (pod volumes) must be specified using the FlexVolume driver. Standard Kubernetes directory-based volumes are not supported except for the use case of Kubernetes Secrets and ConfigMaps. If a Secret or a ConfigMap is mounted to a VM pod, its content is copied into an appropriate location inside the VM using the cloud-init mechanism.
• No support for kubectl exec.

For details on Virtlet operations, see: Virtlet documentation.

This section describes how to create and configure Virtlet pods as well as provides examples of pods for different services.

For an instruction on how to update Virtlet, see Update Virtlet.