The distributed nature of cloud native DNS solution creates a variety of reliability concerns. Container crashes, network disruptions or traffic spikes, could disrupt a cloud DNS solution build for facilitating workload specific record types (NATPR for telco purposes), or service discovery across multi-cluster or hybrid cloud environments. Chaos Engineering can be used to evaluate the fault tolerance, performance, and correctness of these systems, helping ensure robustness under such conditions.

In this talk, we will show how to set up a DNS service with cloud-native components like CoreDNS, ExternalDNS and PowerDNS, then test its resilience with a steady‑state hypothesis and realistic fault injection. We will highlight the key metrics for observing and measuring the impact of injected faults and the overall health of the service. Finally, we will demonstrate how this chaos‑testing approach can drive architectural changes towards a more reliable service.