EKS Subnet Calculator (Amazon EKS CIDR Calculator)

Correct EKS subnet sizing is one of the most common “gotchas” in Amazon EKS.

With the default AWS VPC CNI, pods and nodes typically consume IPs from the same VPC subnets. AWS also reserves 5 IP addresses per subnet, which reduces usable capacity.

This EKS subnet calculator helps you choose the right subnet CIDR prefix before deployment.

Want to understand ENI limits and why EKS can stop scaling?

Even if your subnet looks large enough, ENI / IP-per-instance limits can cap pods per node.

→ Read the EKS Networking Deep Dive

If you’re unsure what to enter, these are practical real-world assumptions:

Max pods per node (EKS)

  • Conservative planning value: 50
  • Many clusters run 30–50 pods per node depending on instance type and workloads

If you’re uncertain, use 50 to avoid subnet exhaustion.

Growth buffer

  • Recommended: 20–30%
  • Common production value: 25%
  • Fast-growing workloads: 40%+

Always size for future growth, not just day-one capacity.

How EKS Networking Affects Subnet Size (AWS VPC CNI)

1️⃣ Shared-subnet model (most common)

With AWS VPC CNI:

  • Nodes use VPC subnet IPs
  • Pods also use VPC subnet IPs (via ENIs on the worker node)
  • Your subnets must accommodate:

(planned nodes) + (planned nodes × max pods per node)

…and still leave headroom for buffer and AWS reserved IPs.

⚠ The “CIDR math is fine but EKS still won’t scale” problem

Even if your subnet CIDR looks large enough, your effective pod scaling can be limited by ENI/IP-per-instance limits on the EC2 instance type.

So the planner helps you size subnets correctly, but also treats instance type limits as an important design constraint.

(If you want, we can add a short “instance type ENI limits” hint section later, but the core calculator stays simple.)

How the EKS Subnet Calculator Works

The planner calculates:

Planned nodes

plannedNodes = nodes + futureGrowth
bufferFactor = 1 + (buffer% / 100)

Total IP requirement (shared subnet model)

requiredIPs = ceil(plannedNodes × bufferFactor) + ceil(plannedNodes × maxPodsPerNode × bufferFactor)

It then finds the smallest subnet prefix where:

usableIPs ≥ requiredIPs + AWS reserved IPs (5)

Most Common EKS Subnet Sizing Mistakes

  • Using a small subnet (like /24) with high pod density and running out of IPs
  • Forgetting AWS reserves 5 IPs per subnet
  • Ignoring growth buffer
  • Not accounting for future node groups / scaling events
  • Assuming “CIDR math” is the only constraint (ENI limits matter too)

FAQ – EKS Subnet Sizing

How many IP addresses does AWS reserve per subnet?

AWS reserves 5 IP addresses per subnet (reducing usable capacity).

Do pods consume subnet IPs in EKS?

Yes—typically with the default AWS VPC CNI, pods receive VPC IPs and consume subnet capacity.

What happens if my EKS subnet runs out of IPs?

  • Nodes may fail to join the cluster
  • Pods may fail to schedule
  • Scaling events can stall
  • Fixing it often requires subnet redesign or migration