Opscraft
Networking fundamentals · Lesson 7 of 9 · 3 min read

Kubernetes networking and service meshes (bridge to other topics)

  • Pod networking comes from the CNI plugin: on EKS the VPC CNI gives pods VPC IPs; others use overlays (Calico VXLAN, Flannel) or eBPF (Cilium).
  • Services are virtual IPs implemented by kube-proxy (iptables/IPVS) or eBPF; DNS by CoreDNS.
  • NetworkPolicy requires an enforcing CNI; a policy "with no effect" means the CNI doesn't enforce it.
  • Ingress controllers / Gateway API provide L7 entry; on EKS the AWS LB Controller creates ALBs/NLBs.
  • Service mesh (Istio, Linkerd, Cilium service mesh) adds mTLS between services, retries/timeouts/circuit breaking, traffic splitting for canaries, and per-request telemetry — via sidecar proxies (Envoy) or sidecar-less "ambient" modes. It costs complexity and some latency; adopt it when you need zero-trust service identity or fine traffic control across many services.

Pod-to-RDS connectivity checklist: resolve the RDS endpoint from a pod (DNS) → RDS security group allows the node SG or pod SG (security groups for pods) on 5432 → subnet route tables and NACLs between the pod subnet and DB subnet → credentials/TLS settings.

Enlarge diagramDNSthe RDSendpointresolves frominside the podSecurity groupRDS SG allowsthe node SG orpod SG on 5432Routes andNACLsbetween the podsubnet and theDB subnetCredentialsand TLSsecret mounted,CA bundlepresentThe apppool settings,timeouts, retries12345
Figure 1. Pod cannot reach RDS? Check each hop in order.
Swipe sideways to see all of it, or tap to open it full screen.