Load Testing and Benchmarking

Performance testing is important for high-volume or intensive workflows. Two tools commonly used with Ghost are loadtest for a simple endpoint and artillery for multi-step or variable traffic.

loadtest

Run loadtest with pnpm and specify the total requests and concurrency:

pnpm dlx loadtest -n 500 -c 5 http://localhost:2368/

It can also run at a fixed request rate for a duration:

pnpm dlx loadtest -t 30 --rps 50 http://localhost:2368/

Artillery

Artillery uses a YAML test definition and supports phased rates, several requests in one flow, and processor functions for variable input:

config:
  target: "http://localhost:2368"
  phases:
    - duration: 15
      arrivalRate: 50

scenarios:
  - name: "Home page"
    flow:
      - get:
          url: "/"

Run it with:

pnpm dlx artillery run load-test.yml

Getting useful results

Request pooling, keep-alive, timeouts, concurrency, cookies, and headers can all change the result. Match them to the behavior being investigated. Ghost caches public content, so decide whether the test is intended to measure cached or uncached requests.

Local testing starts at http://localhost:2368. Local hardware and production hosting differ, but local results can still show whether a change improves or degrades a focused workflow.

Only send load to a system you own or have explicit permission to test. The public repository does not document production or hosted-service load-testing procedures.

See the loadtest README and Artillery documentation for the current command and scenario options.