Diagnosing connectivity issues between agent and proxy

The Buoyant Cloud metrics agent scrapes the /metrics admin endpoint served on port 4191 of all linkerd-proxy containers running in your cluster.

There are a few reasons why those scrapes would fail:

The Buoyant Cloud metrics agent needs permission to access the admin port (4191) on all your proxies. It’s possible that Linkerd policy has been configured to disallow that.

Network policy is often used to firewall off traffic between pods or namespaces. Ensure your network policy allows for traffic on port 4191 from the the linkerd-buoyant namespace to all namespaces where proxies are running.

  • Use kubectl to inspect NetworkPolicy objects or contact your platform administrators.

If a Linkerd proxy is unhealthy or is heavily resource constrained it may not be able to surface metrics.

  • Check the resource usage and limits for the pod in question, and if need be consider restarting the pod.