It's always DNS

When DNS Meets eBPF: How a Resolver Choice caused production downtime We hit an outage after shipping network policy changes on a cluster that uses Cilium in kube-proxy replacement mode. Minutes later a nginx server being used by other services start...