Running Kubernetes on bare metal

Why bare metal Kubernetes?

Our research found that 1 in 5 Kubernetes users are already running on bare metal today in their private clouds and data center environments. Why?

Directly deploying and running Kubernetes on top of physical resources instead of virtual servers means one technology layer less to manage: the hypervisor.

Eliminating the hypervisor with bare metal servers can offer:

Increased performance, particularly for specialized workloads like big data and AI.
Better resource utilization, by as much as 20%.
Savings, both in software licensing and operational efficiencies.
Reduced risk through fewer technical and commercial dependencies.
Better visibility into the health of your hardware.

Learn more about the benefits of bare metal

Super League Gaming is all-in on bare metal Kubernetes

Palette user Super League Gaming hosts up to 15,000 concurrent Minecraft players on a single massive bare metal Kubernetes cluster. They trust the infrastructure to perform when every millisecond counts.

The OS challenge

But deploying Kubernetes on dedicated servers is a challenge, especially at scale. And it’s all to do with the OS.

When you’re running Kubernetes on cloud services in the public cloud, or in a virtualized data center, the infrastructure is ready and waiting for your Kubernetes installation.

With bare metal, you have to install an OS on every server to get it ready for the Kubernetes stack on top. And most Kubernetes management platforms have no visibility of the OS they sit on, so you have to manage the OS separately.

This can expose you to configuration drift that can impact performance and availability, besides increasing the operational hassle.

Discover 5 best practices for running bare metal K8s

See Palette in action
on bare metal

Treat your bare metal clusters just like any other clusters

With Palette, your bare metal servers’ operating system is just another layer being managed by our platform. You can take advantage of all our comprehensive lifecycle management features, from monitoring to patching.

This is what true “full-stack” means: everything automated and managed, from the OS, to the K8s infrastructure and add-on application services.

Powered by cutting-edge open source projects

To make this happen, Palette extends the CNCF’s Cluster API (CAPI). Using CAPI, Palette manages all your clusters in a declarative way: you define what your clusters should look like and Palette ensures they always stay that way, consistent against the declarative profile via continuous reconciliation.

We built an open-source integration between Cluster API and Canonical’s MAAS product, which automates bare metal server provisioning from start to finish, including everything from storage and network discovery to hardware testing and OS imaging.

Together, Palette, CAPI and MAAS mean you can treat your bare metal data center like a cloud!

Read our docs to learn about our MAAS integration

Meet MAAS

Canonical MAAS, or Metal As A Service, manages the lifecycle of bare metal machines and provides a clean and flexible API to do so.

It allows you to easily install, upgrade, and rebuild servers, providing them with an operating system such as Ubuntu, from scratch.

This means you can programmatically manage your bare-metal servers, enabling you to provision, decommission, track inventory, test, and maintain your bare metal cloud.

Find out more about MAAS at: https://maas.io/

What about your VMs?

If you’re thinking about cutting out the hypervisor layer from your Kubernetes stack, you may be wondering where else you can consolidate.

Did you know that you can actually manage virtual machine workloads directly from bare metal Kubernetes, meaning you may not need a separate virtualized stack at all?

Learn more about VMs on Kubernetes here