What is a target utilization percentage?

HPA scales pods to keep average utilization near the target. A target of 70% means HPA adds replicas when average CPU across pods exceeds 70% of the requested CPU. Lower targets give more headroom for spikes; higher targets pack pods tighter but leave less buffer.

Why should I set a minimum replica count above 1?

Running at least 2 replicas ensures availability during rolling updates and node failures. A single replica means zero availability the moment that pod is evicted or the node drains.

How do I get my observed peak load numbers?

Query Prometheus with `max_over_time(container_cpu_usage_seconds_total[7d])` or check Datadog/Grafana dashboards for peak CPU and memory over the busiest recent period.

What safety buffer should I use?

A 20–30% buffer is common. It absorbs request bursts while HPA catches up (the default HPA sync period is 15 seconds). Use a higher buffer for latency-sensitive workloads.

Does this calculator replace HPA?

No. This tool gives you a starting replica count and recommended HPA thresholds. HPA then adjusts replicas dynamically at runtime based on live metrics.

Pod Replica Calculator

Name: Pod Replica Calculator
Availability: InStock
Author: Nham Vu

Enter your pod resource requests and expected load to get the recommended replica count and HPA configuration.

Pod & Workload Parameters

CPU Request per Pod (millicores)

e.g. 250 = 0.25 cores. Use value from your pod spec resources.requests.cpu.

Expected Peak CPU Load (millicores)

Total CPU your workload needs at peak (sum across all replicas).

Memory Request per Pod (MiB)

e.g. 512 = 512 MiB. From resources.requests.memory.

Expected Peak Memory Load (MiB)

Total memory your workload needs at peak across all pods.

HPA Target Utilization: 70%

20%95%

Safety Buffer: 20%

0%50%

—

Replicas (CPU-driven)

—

Replicas (Memory-driven)

—

Recommended Replicas

—

Suggested Min (HA)

HPA Configuration Snippet

Calculation Breakdown

Metric	CPU	Memory

Fill in the parameters and click Calculate Replicas.

Copied!

Summary

Enter your pod resource requests and expected load to get the recommended replica count and HPA configuration.

How it works

Enter the CPU and memory request for a single pod.
Input the expected peak CPU and memory load your application produces.
Set the target utilization percentage (typically 50–80% for HPA).
Optionally add a safety buffer to handle traffic spikes.
The calculator outputs the recommended replica count, minimum replicas, and HPA configuration.

Use cases

Plan the initial replica count before enabling HPA.
Verify your HPA min/max replica bounds are correctly sized.
Estimate cluster node capacity needed for a given workload.
Right-size deployments to avoid under- or over-provisioning.
Translate Prometheus metrics into concrete replica counts.
Quickly iterate over what-if scenarios for load projections.

Frequently Asked Questions

Last updated: 2026-07-22 · Reviewed by Nham Vu