oha — HTTP Load Generator with Real-Time TUI Dashboard

Introduction

oha is an HTTP load generator written in Rust that provides a real-time terminal user interface showing live latency histograms, status code distributions, and throughput metrics as the test runs. Inspired by rakyll/hey, it adds visual feedback so you can observe server behavior under load without waiting for a final summary. The Rust implementation ensures low resource consumption on the client side even at high concurrency levels.

What oha Does

• Sends configurable HTTP load with fixed request count or duration-based runs
• Displays a real-time TUI with live latency histogram, percentiles, and error rates
• Supports HTTP/1.1 and HTTP/2 with custom headers, methods, and request bodies
• Outputs final results as text, JSON, or compatible formats for CI integration
• Handles TLS connections with configurable certificate options

Architecture Overview

oha uses Tokio as its async runtime to manage thousands of concurrent connections efficiently. Each connection runs as an independent async task sending requests and recording response times. A dedicated TUI thread powered by ratatui reads timing data from a shared buffer and renders live charts at a fixed refresh rate. After the run completes, oha aggregates all samples and prints a statistical summary including percentiles, mean, and standard deviation.

Self-Hosting & Configuration

• Install via cargo install oha or brew install oha on macOS
• Pre-built binaries available on GitHub Releases for Linux, macOS, and Windows
• Use -n for total request count or -z for duration-based runs (e.g., -z 60s)
• Set -c for concurrent connections and -q for per-second rate limiting
• Disable the TUI with --no-tui for headless CI environments and pipe JSON output with -j

Key Features

• Real-time terminal dashboard with live latency histogram and throughput graph
• Written in Rust for minimal memory footprint and maximum client-side efficiency
• Supports both HTTP/1.1 and HTTP/2 protocols out of the box
• JSON output mode for programmatic consumption in CI/CD pipelines
• Cross-platform: runs natively on Linux, macOS, and Windows

Comparison with Similar Tools

• hey — simpler text-only output with no live dashboard; oha adds real-time visualization
• wrk — higher raw throughput via C/epoll but no TUI and Unix-only; oha is cross-platform
• bombardier — Go-based with a progress bar; oha offers a richer TUI with histograms
• k6 — scriptable in JavaScript with cloud reporting; oha is a lightweight single-purpose tool
• Vegeta — constant-rate attack model with plot output; oha provides live interactive feedback

FAQ

Q: Can I disable the TUI for CI usage? A: Yes. Pass --no-tui to get a simple text summary, or -j for JSON output suitable for automated parsing.

Q: Does oha support HTTP/2? A: Yes. Use --http2 to force HTTP/2 connections, or let oha negotiate via ALPN automatically.

Q: How does oha compare to wrk in throughput? A: wrk typically achieves higher raw RPS due to its C/epoll design. oha trades some peak throughput for a richer user experience and cross-platform support.

Q: Can I send POST requests with a body? A: Yes. Use -m POST -d '{"key":"val"}' or -D @body.json to load the body from a file.

Sources

• https://github.com/hatoo/oha
• https://github.com/hatoo/oha/blob/master/README.md