Join us in San Francisco for TailscaleUp!Grab your ticket ->

CoreWeave best practices

Last validated:

This document details best practices for using Tailscale with CoreWeave, particularly for connecting to CoreWeave Kubernetes Service (CKS) clusters using the Tailscale Kubernetes Operator.

Use the CoreWeave fork of the Tailscale Kubernetes Operator

CoreWeave maintains its own build of the Tailscale Kubernetes Operator so that it integrates cleanly with CKS: a purpose-built version of the tailscale-operator Helm chart, paired with a CoreWeave-specific container image. Deploy CoreWeave's fork rather than the upstream Helm chart when running on CKS.

For setup instructions, refer to CoreWeave's tailscale-operator documentation.

For details on the chart's customizations, refer to Deploy the tailscale-operator chart.

Networking and connectivity

CoreWeave nodes do not receive public IP addresses and are behind a hard NAT, which can prevent direct connections between peers and reduce bandwidth and connection stability. To optimize connectivity:

  • Add CoreWeave's DERP servers to your tailnet's derpMap configuration alongside, or instead of, Tailscale's default relays for optimal connectivity into CoreWeave's networks.
  • Deploy and configure peer relays in your network, with a public IP address, and allow incoming UDP traffic to the relay port to support connections through CoreWeave's NAT gateway.