oha

oha is an HTTP load testing tool inspired by hey and wrk, written in Rust on top of tokio. It bombards a target URL with requests and displays a live, animated TUI showing real-time throughput, latency percentiles, a histogram, and a status code breakdown — all while the test is running.

Features

• Live TUI — a real-time terminal dashboard updates as requests are sent, showing RPS, latency distribution, and a bar chart of response times
• Async I/O — built on Tokio; handles high concurrency with minimal OS thread overhead
• HTTP/1.1 and HTTP/2 — supports both protocols; force HTTP/2 with --http2
• Custom headers, bodies, and methods — full control over the request shape
• Latency percentiles — reports p50, p75, p90, p95, p99, and p99.9 in the summary
• Duration or count mode — run for a fixed duration (-z 30s) or a fixed number of requests (-n 1000)
• Redirect following — configurable redirect behaviour
• Unix socket support — send requests over a Unix domain socket
• JSON output — machine-readable results for use in scripts and CI benchmarks

Installation

cargo install oha

Or via your system package manager:

# Debian / Fedora
# Pre-built Linux binaries are available on the
# [releases page](https://github.com/hatoo/oha/releases).

# macOS
brew install oha

# Arch Linux
pacman -S oha

# Nix
nix-env -iA nixpkgs.oha

Usage

# Basic load test — 200 requests, 50 concurrent
oha -n 200 -c 50 https://example.com

# Run for 30 seconds instead of a fixed request count
oha -z 30s -c 100 https://example.com

# POST request with a JSON body
oha -n 500 -c 20 \
  -m POST \
  -H "Content-Type: application/json" \
  -d '{"key":"value"}' \
  https://api.example.com/endpoint

# Force HTTP/2
oha --http2 -n 1000 -c 50 https://example.com

# Disable TLS certificate verification
oha --insecure -n 100 https://localhost:8443

# Output results as JSON (useful in CI)
oha -n 1000 -c 50 --json https://example.com

# Test with a specific timeout per request
oha --timeout 5 -n 500 https://example.com

# Disable the live TUI (plain output only)
oha --no-tui -n 200 https://example.com

Example Output

After a run completes, oha prints a summary like:

Summary:
  Success rate: 100.00%
  Total:        5.0123 secs
  Slowest:      0.2341 secs
  Fastest:      0.0041 secs
  Average:      0.0501 secs
  Requests/sec: 1995.09

  Total data:   4.77 MiB
  Size/request: 2.44 KiB

Response time histogram:
  0.004 [1]    |
  0.027 [6185] |■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
  0.050 [2898] |■■■■■■■■■■■■■■■
  0.073 [619]  |■■■
  0.096 [196]  |■
  0.119 [72]   |
  0.142 [20]   |
  0.165 [7]    |
  0.188 [1]    |
  0.211 [0]    |
  0.234 [1]    |

Latency distribution:
  10%  in 0.0181 secs
  25%  in 0.0224 secs
  50%  in 0.0312 secs
  75%  in 0.0621 secs
  90%  in 0.0871 secs
  95%  in 0.1021 secs
  99%  in 0.1389 secs

Status code distribution:
  [200] 10000 responses

Tips

• Start with a modest concurrency level (-c 10) and ramp up to find where your service starts degrading
• Use --json in CI to capture benchmark baselines and fail the build if p99 latency regresses
• Pair with hyperfine for command-line benchmarks and oha for HTTP service benchmarks — they complement each other well
• The live TUI is particularly useful for spotting latency spikes in real time that would be invisible in a post-run summary