VPC CIDR Planning Procedure

VPC CIDR Planning Procedure A workflow diagram generated by Archify. 01 / CIDR Planning EX / Overlap Check Cluster scale · Estimate node count · CIDR Planning Cluster scale Estimate node count IP demand · nodes × max pods · CIDR Planning IP demand nodes × max pods Reserve headroom · 20-30% spare · CIDR Planning Reserve headroom 20-30% spare Existing networks · On-premises · peering · CIDR Planning Existing networks On-premises · peering Finalize VPC CIDR · RFC 1918 private range · CIDR Planning Finalize VPC CIDR RFC 1918 private range Range overlap · Needs redesign · Overlap Check Range overlap Needs redesign no overlap overlap found pick a new range Legend Workload Security gate AWS service External

Recommended CIDR by scale

  • • Small (1-10 nodes): /24 · 256 IPs · dev and test
  • • Medium (10-100 nodes): /20 · 4,096 IPs · multi-AZ recommended
  • • Large (100+ nodes): /16 · 65,536 IPs · multi-AZ required

Worked IP calculation

  • • m5.large: 3 ENIs × 10 IPs per ENI → up to 29 pods
  • • Required IPs = node count × max pods per node + system reservations
  • • Split subnets with small public ranges (/24) and larger private ranges (/20-/22)

Common mistakes

  • • Sizing the CIDR too small and running out of pod IPs
  • • Overlapping on-premises or peered ranges, which blocks connectivity later
  • • Unbalanced subnet sizes per AZ, so scheduling fails in one AZ only