Skip to main content
With Managed Kubernetes, worker nodes in a Kubernetes cluster deploy directly on Bare Metal servers. This setup lets workloads use dedicated hardware resources while keeping Kubernetes orchestration.

​
Architecture

Applications often run on Virtual Machines (VMs) with Kubernetes and containers layered on top. This setup involves multiple layers: VMs, hypervisors, and operating systems. Kubernetes on Bare Metal removes the virtualization layer. Containers run directly on the host server operating system (OS), so they have direct access to the underlying hardware.

​
Advantages

  • Improved performance. Bare Metal support uses the full hardware capacity because it eliminates the virtualization layer.
  • Enhanced security and isolation. Bare Metal runs applications in a completely isolated environment without shared resources or noisy neighbors.
  • Higher network performance. Applications on Bare Metal servers can handle higher traffic because nodes have dedicated access to the network card and can use the full network card bandwidth.
  • Faster disk response time. Bare Metal gives applications direct access to underlying hardware, including disks, which reduces latency because no additional layers sit between the workload and storage.
  • Direct internet access. A public interface can be assigned to Bare Metal servers, enabling direct internet access without intermediary layers and reducing latency to end users.

​
Available configurations

Gcore offers Bare Metal flavor categories for Kubernetes worker node pools:
  • High frequency. Single-socket servers with AMD 7950x and 9950x processors, suitable for applications that require high processor frequency.
  • Infrastructure. Multi-core, multi-socket configurations designed for applications that require a high number of cores and are optimized for multithreading.
  • Storage. Configurations optimized for storage-intensive workloads.

​
Create a Bare Metal Kubernetes cluster

In the Gcore Customer Portal, navigate to Cloud > Managed Kubernetes and click Create Cluster. On the cluster form, open the Pools section and select Bare metal instances under Type. The portal also offers GPU Bare metal instances for GPU-accelerated workloads.
Bare metal worker nodes may be deleted and recreated when Kubernetes reschedules pods, because Managed Kubernetes treats nodes as stateless infrastructure. To protect critical data, use a Gcore storage service such as Managed PostgreSQL, NFS, S3, or File Shares.