Hardware requirements

Cozystack utilizes Talos Linux, a minimalistic Linux distribution designed solely to run Kubernetes. Usually, this means you cannot share a server with any services other than those run by Cozystack. The good news is that whichever service you need, Cozystack will run it perfectly: securely, efficiently, and in a fully containerized or virtualized environment.

Hardware requirements depend on your usage scenario. Below are several common deployment options; review them to determine which setup fits your needs best.

Small lab

Here are the baseline requirements for running a tiny installation. The minimum recommended configuration for each node is as follows:

Compute:

• Three physical or virtual servers with amd64/x86_64 architecture, with at least 16 GB RAM and 4 CPU cores each.
• Virtualized servers need nested virtualization enabled and the CPU model set to host (without emulation).
• PXE installation requires an extra management instance connected to the same network, with any Linux system able to run a Docker container. It should also have x86-64-v2 architecture, which most probably may be achieved by setting CPU model to host in case of a VM.

Storage:

Storage in a Cozystack cluster serves two primary roles: one for the system and one for your workloads. Understanding each role ensures the stability and scalability of your environment.

• Primary Disk: This disk contains the Talos Linux operating system, essential system kernel modules and Cozystack system base pods, logs, and base container images.

  Minimum: 32 GB; approximately 26 GB is used in a standard Cozystack setup. The Talos installation expects /dev/sda as the system disk (virtio drives usually appear as /dev/vda).

• Secondary Disk: Dedicated to workload data and can be increased based on workload requirements. Used for provisioning volumes via PersistentVolumeClaims (PVCs).

  Suggested: 100 GB. Disk path (usually /dev/sdb) will be defined in the storage configuration. It does not affect the Talos installation.

  Learn more about configuring Linstor StorageClass from the Deploy Cozystack tutorial

Networking:

• Machines must be allowed to use additional IPs, or an external load balancer must be available. Using additional IPs is disabled by default and must be enabled explicitly in most public clouds.
• Additional public IPs for ingress and virtual machines may be needed. Check if your public cloud provider supports floating IPs.

Production Cluster

For a production environment, consider the following:

Compute:

• Having at least three worker nodes is mandatory for running highly available applications. If one of the three nodes becomes unavailable due to hardware failure or maintenance, you’ll be operating in a degraded state. While database clusters and replicated storage will continue functioning, starting new database instances or creating replicated volumes won’t be possible.
• Having separate servers for Kubernetes master nodes is highly recommended, although not required. It’s much easier to take a pure worker node offline for maintenance or upgrades, than if it also serves as a management node.

Networking:

• In a setup with multiple data centers, it’s ideal to have direct, dedicated optical links between them.
• Servers must support out-of-band management (IPMI, iLO, iDRAC, etc.) to allow remote monitoring, recovery, and management.

Distributed Cluster

You can build a distributed cluster with Cozystack.

Networking:

• Distributed cluster requires both a fast and reliable network, and it must have low RTT (Round Trip Time), as Kubernetes is not designed to operate efficiently over high-latency networks.

  Data centers in the same city typically have less than 1 ms latency, which is ideal. The maximum acceptable RTT is 10 ms. Running Kubernetes or replicated storage over a network with RTT above 20 ms is strongly discouraged. To measure actual RTT, you can use the ping command.

• It’s also recommended to have at least 2–3 nodes per data center in a distributed cluster. This ensures that the cluster would be able to survive one data center loss without major disruption.

• If it’s hard to keep a single address space between data centers, instead of using some external VPN, you can enable KubeSpan, a Talos Linux feature that creates a WireGuard-backed full-mesh VPN between nodes.

Highly Available Applications

Achieving high availability adds to the basic production environment requirements.

Networking:

• It is recommended to have multiple 10 Gbps (or faster) network cards. You can separate storage and application traffic by assigning them to different network interfaces.

• Expect a significant amount of horizontal, inter-node traffic inside clusters. It is usually caused by multiple replicas of services and databases deployed across different nodes exchanging data. Also, virtual machines with live migration require replicated volumes, which further increases the amount of traffic.