Kubernetes Subnet Size Calculator (AKS, ARO, EKS & GKE)

Use this calculator to estimate node subnet and pod CIDR sizes before deploying Kubernetes clusters in Azure, AWS, or Google Cloud.

It supports:

  • AKS (Azure CNI, Azure CNI Overlay, Kubenet)
  • ARO (Azure Red Hat OpenShift)
  • EKS (AWS VPC CNI typical model)
  • GKE (VPC-native clusters with secondary ranges)

The planner models:

  • Cloud provider reserved IP behavior
  • Shared-subnet vs split-range networking
  • Growth headroom
  • Validation of existing CIDR prefixes

Not sure which provider model fits your architecture?

→ Compare Kubernetes Networking Models

What This Planner Calculates

The planner estimates required IP capacity using:

  1. Total planned nodes
    current nodes + future growth

  2. Node IP requirement
    planned nodes × growth buffer

  3. Pod IP requirement
    planned nodes × max pods per node × growth buffer

Then, based on provider networking behavior, it selects the smallest CIDR prefix where: usable IPs ≥ required IPs + cloud-reserved IPs

Inputs

  • Cloud provider (AKS / ARO / EKS / GKE)
  • AKS network mode (Azure CNI / Overlay / Kubenet)
  • ARO control plane node count
  • Number of worker nodes
  • Max pods per node
  • Growth buffer (%)
  • Optional future node growth
  • Optional current CIDR prefixes (to detect undersized ranges)

Outputs

Depending on provider, the planner returns:

  • Required node subnet size
  • Required pod CIDR size
  • Reserved IP impact per subnet/range
  • Suggested CIDR prefix recommendations
  • Validation warnings if an existing range is too small

Provider Networking Models Explained

Understanding how each platform allocates IPs is critical.

AKS Subnet Sizing

→ AKS Networking Deep Dive

Azure CNI (Flat / Shared Subnet)

  • Pods and nodes consume IPs from the same Azure subnet
  • Azure reserves 5 IPs per subnet
  • Planner calculates:
    • One subnet large enough for nodes + pods combined

This is the most common source of unexpected subnet exhaustion.

→ AKS Subnet Calculator

Azure CNI Overlay

  • Nodes consume Azure subnet IPs
  • Pods use an overlay CIDR
  • Azure reserves 5 IPs for the node subnet
  • Pod CIDR does not consume Azure subnet IPs

Planner calculates:

  • Node subnet size
  • Separate pod CIDR size

Kubenet

  • Nodes consume Azure subnet IPs
  • Pods use a separate pod CIDR
  • Azure reserves 5 IPs per subnet

Planning logic matches overlay mode from a subnet capacity perspective.

ARO (Azure Red Hat OpenShift)

→ ARO Networking Deep Dive

ARO typically uses:

  • A master subnet (control plane)
  • A worker subnet
  • A separate OpenShift cluster network CIDR for pods

Key behaviors:

  • Azure reserves 5 IPs per subnet
  • Pods use overlay cluster network CIDR (do not consume subnet IPs)
  • Master and worker subnets must be sized independently

Planner calculates:

  • Master subnet size
  • Worker subnet size
  • Pod cluster network CIDR

This reflects real-world ARO deployments more accurately than treating it like AKS.

→ ARO Subnet Calculator

EKS (Amazon EKS)

→ EKS Networking Deep Dive

  • Pods and nodes typically consume IPs from the same VPC subnet(s)
  • AWS reserves 5 IPs per subnet
  • Subnet must accommodate total (nodes + pods)

⚠ Important:

Even if CIDR math is sufficient, ENI/IP-per-instance limits may cap effective pod density depending on EC2 instance type.

The planner surfaces this limitation as guidance.

→ EKS Subnet Calculator

GKE (VPC-Native)

→ GKE Networking Deep Dive

  • Nodes use the primary subnet
  • Pods use a secondary IP range
  • Google Cloud reserves 4 IPs per subnet/range (first 2 + last 2)

Planner calculates:

  • Primary node subnet size
  • Secondary pod range size

GKE allocates pod IP blocks per node from the secondary CIDR.

→ GKE Subnet Calculator

Example Scenarios

Example 1: AKS Azure CNI (Flat)

Inputs:

  • 20 nodes
  • 30 max pods per node
  • 25% buffer
  • Future node growth: 10

Result:

  • One subnet recommendation covering nodes + pods
  • Azure 5-IP reserved impact included
  • Validation warnings if current subnet too small

Example 2: AKS Azure CNI Overlay

Inputs:

  • 30 nodes
  • 50 max pods per node
  • 20% buffer

Result:

  • Node subnet recommendation (Azure reserved included)
  • Separate pod overlay CIDR recommendation

Example 3: ARO Deployment

Inputs:

  • 3 control plane nodes
  • 25 worker nodes
  • 40 max pods per node
  • 25% buffer

Result:

  • Master subnet recommendation
  • Worker subnet recommendation
  • Pod cluster network CIDR suggestion

Example 4: GKE VPC-Native

Inputs:

  • 40 nodes
  • 32 max pods per node
  • 20% buffer

Result:

  • Primary subnet suggestion
  • Secondary pod range suggestion
  • 4 reserved IP impact per range included

Example 5: EKS with Existing CIDR Validation

Inputs:

  • 25 nodes
  • 40 max pods per node
  • 30% buffer
  • Current subnet: /25

Result:

  • Warning if subnet too small for total planned IPs
  • Reminder that ENI limits may restrict effective pod scaling

Why This Planner Matters

Common real-world failures:

  • AKS Azure CNI subnets running out of IPs unexpectedly
  • EKS scaling blocked despite “enough CIDR”
  • GKE secondary ranges sized too small
  • ARO worker subnet exhaustion during scale events

Under-sizing subnets in Kubernetes environments is difficult and disruptive to fix later.

This planner helps you size correctly before deployment.