BYOC and BYOK requirements

BYOC resource requirements

These minimum requirements are enforced when creating a cluster or updating its node pools.

A single-node cluster must therefore use a node with at least 8 vCPUs and 16 GB memory. Clusters using self-hosted log storage require at least 12 vCPUs and 32 GB memory.

BYOK resource requirements

Northflank does not provision nodes on imported clusters (BYOK), so these requirements are not enforced. Treat them as guidance: below them, Northflank's system components may not have enough capacity to schedule alongside your workloads.

Optimization recommendations

Node sizing: Prefer fewer larger nodes over many smaller nodes. Some system components must run on every node, so larger nodes minimize the relative per-node overhead, leaving more capacity for your workloads.

Storage ratio: Maintain at least 5 GB of ephemeral storage per vCPU for optimal performance.

If you're importing an existing Kubernetes cluster (BYOK), your cluster must have these components pre-installed and meet additional requirements. See import an existing cluster (BYOK) for the import process.

Kubernetes version

Your cluster must run a supported Kubernetes version: currently 1.34 and 1.35. Newer versions may work, but are not officially supported.

Required system components

Your cluster must have these components already installed:

The following Cilium flags must be enabled for L7 networking features (for example L7 network policies and port-level security policies):

• enable-l7-proxy: "true"
• enable-envoy-config: "true"

Note on Kube-DNS: If you're currently using Kube-DNS instead of CoreDNS, contact Northflank support before importing your cluster.

L4 load balancer support

Your Kubernetes installation must be able to provision external, public IPs for Kubernetes Service resources of type LoadBalancer.

Requirements:

• Must provision L4 load balancers by default
• If your provider requires specific annotations on Service resources to provision L4 load balancers, contact Northflank support for assistance

Component conflicts

Ensure the following components are NOT pre-installed on your cluster (except CoreDNS), as they will be installed during the import process:

• Istio
• Envoy Gateway
• Prometheus
• Promtail
• RuntimeClass resources (except those required by your provider)

If these components exist before import, the installation process may fail due to conflicting resources.

See import an existing cluster (BYOK) for a walkthrough of the import process, the components Northflank installs on your cluster, and the split of responsibilities between Northflank and you.

Use a new cluster for BYOK import

We strongly recommend using a new, dedicated cluster for BYOK import. Do NOT import clusters that:

• Are running production workloads
• Host important or business-critical applications
• Are shared with other systems or teams

Potential risks:

• The installation might fail and leave the cluster in an unhealthy state
• There is currently no full deinstallation process
• Existing configurations may be overwritten or modified

If something goes wrong: If the import fails or your cluster becomes unhealthy, you may need to redeploy a fresh cluster.

Contact Northflank support if:

• You have questions about the installation process
• Your Kubernetes provider doesn't meet some requirements
• Your provider requires customization (e.g., specific annotations, non-standard configurations)
• You need assistance during or after the import process
• Your cluster uses Kube-DNS instead of CoreDNS