Goldpinger

Goldpinger helps reveal whether different parts of a Kubernetes cluster can reach one another. Instances placed around the cluster make requests to peers and report the results as monitoring metrics and a network graph. It is useful for spotting connectivity failures that depend on which machine a workload uses. Its checks cover the configured probe paths, so a green graph does not establish that every application route or access policy works.

Deployment and operating notes

Bloomberg’s current Goldpinger README describes calls between DaemonSet instances, Prometheus metrics and a network graph, with additional configurable probe types. It is useful for observing connectivity between locations in a cluster, but an all-green graph does not prove every application Service, ingress route or NetworkPolicy works.

Decide whether probes should use pod networking or another network path, and ensure scheduling covers the relevant node pools, operating systems and taints. Review API discovery permissions and restrict access to the topology UI. Choose a probe interval and peer scope that remain affordable as node count grows. Test an intentional, isolated connectivity restriction and confirm the expected metric and alert, then restore it. Correlate failures with node placement and the actual application path before attributing them to the CNI. Preserve dashboard and alert label assumptions during chart upgrades. This review did not run a network benchmark or demonstrate that Goldpinger can identify every cause of packet loss.