Networking
Module infographic
32 / 69

Revision poster: VPC CNI, IP planning and load balancing.

Infographic: Pods with VPC IPs on node ENIs, IP exhaustion fixes, the AWS Load Balancer Controller with ALB and NLB, and VPC Lattice
Click the poster to enlarge it

Left: how Pods consume VPC IPs and the three tools for IP pressure. Right: how traffic enters the cluster. Bottom: cross-cluster services.

  1. 1Pods use real VPC IPs from node ENIs; capacity per node is limited by the instance type.
  2. 2Prefix delegation raises Pods per node; custom networking adds a secondary CIDR; IPv6 removes the limit.
  3. 3Security groups for pods control Pod to AWS resource access.
  4. 4The Load Balancer Controller builds ALBs for HTTP and gRPC and NLBs for TCP, UDP and TLS.
  5. 5IP target mode sends traffic straight to Pods.
  6. 6VPC Lattice connects services across clusters and accounts.

Self-check: cover the poster and answer

  • Which fix helps when the whole VPC CIDR is nearly full?
  • Why is IP target mode preferred over instance mode?

Plan pod IP capacity first, then pick the load balancer.

BackNext