Field Guides / Kubernetes Networking Deep Dive
Field Guide · Intermediate to Advanced · 10-12 hours with labsKubernetes Networking Deep Dive
Trace every packet in a Kubernetes cluster, from pod IP to Service, DNS, Gateway and policy, and fix it when it breaks.




70
pages
pages
11
hands-on labs
hands-on labs
26
interview Q&As
interview Q&As
13
chapters
chapters
This is for you if
- You deploy to Kubernetes but guess when pods can't reach each other
- You're on call and want a real playbook for DNS, MTU and conntrack issues
- You're interviewing for platform, SRE or DevOps roles where networking questions go deep
- You need to choose or defend a CNI, kube-proxy mode or ingress strategy
What's inside
- The pod-to-Service journey walked step by step, with a mental model you can reuse in any incident
- Calico vs Cilium vs Flannel compared, and both Calico and Cilium installed on kind so you can see the difference on a node
- Every Service type, EndpointSlice conditions and the deploy-time termination race, with the fix
- kube-proxy iptables, nftables and ipvs status (ipvs deprecated, nftables GA and the planned default) plus Cilium's eBPF replacement
- CoreDNS and ndots:5 captured live on the wire, with tuning and NodeLocal DNSCache guidance
- Ingress vs Gateway API after the ingress-nginx retirement, with a canary HTTPRoute lab on Cilium
- NetworkPolicy default deny done right: DNS egress, the AND/OR selector trap and pod-port matching
- MTU and conntrack failures explained, a symptom-to-fix troubleshooting table and four worked incidents
- 26 senior-level interview Q&As, dense cheat sheets and a 30-day practice plan
After this guide you can
- Explain exactly what happens when a pod calls a Service, layer by layer
- Debug pods, nodes and flows with netshoot, tcpdump, conntrack and Hubble
- Read kube-proxy rules in iptables, nftables and eBPF and plan a move off ipvs
- Lock down namespaces with NetworkPolicy without breaking DNS or monitoring
- Route traffic with Gateway API and choose between overlay and native routing
- Diagnose MTU, conntrack and DNS latency problems that don't show up in kubectl
Contents
- The mental modelp. 4
- Pod networking and choosing a CNIp. 8
- Services and EndpointSlicesp. 17
- kube-proxy, nftables and eBPFp. 25
- DNS: CoreDNS and ndotsp. 30
- Getting traffic in: Ingress and Gateway APIp. 35
- NetworkPolicyp. 40
- MTU, conntrack and the quiet killersp. 45
- Troubleshooting playbookp. 49
- Production checklist and design decisionsp. 54
- Interview questionsp. 57
- Cheat sheetp. 64
- 30-day practice planp. 68
Written by Randy, a Senior Platform Engineer & SRE who runs production platforms on AWS, GCP and Azure. Companion to the Techworld with Randy episode Kubernetes Networking Deep Dive.