Table of Contents
Table of Contents
Further Reading
What Does Artillery Mean?
Artillery is an open-source load testing framework for modern APIs and backend services. Developers and QA teams use it to simulate real-world traffic patterns.
It targets HTTP, WebSocket, and Socket.io endpoints, and plugins extend it to other protocols. Test scenarios are written in YAML, with JavaScript for custom logic.
How Artillery Works
- Define phases: set how many virtual users arrive and for how long.
- Write scenarios: describe the requests each virtual user makes.
- Run the test: Artillery generates traffic against the target.
- Read the report: review latency percentiles, request rates, and errors.
Artillery vs. k6 vs. JMeter
Why Artillery Matters
Applications must handle large and unpredictable request volumes. Without proactive load testing, bottlenecks and outages appear only in production. Artillery helps teams:
- Find performance limits and latency issues
- Validate behavior under stress and traffic spikes
- Confirm scalability before customers feel failures
Common Use Cases
- API load testing: REST or GraphQL APIs under realistic traffic
- Real-time services: WebSocket and Socket.io validation
- Release stress tests: checks before new features ship
- Peak event prep: simulating seasonal or promotional traffic
How Artillery Works with Testkube
Testkube runs Artillery scripts inside your Kubernetes cluster as part of Test Workflows. That gives teams:
- Accurate results: load reflects real cluster scaling and networking.
- Horizontal scale: runs can spread across pods for larger distributed load testing.
- Central reporting: results and artifacts are stored for every execution.
- CI/CD automation: tests trigger on every deployment from Jenkins, GitHub Actions, GitLab, or ArgoCD.
Real-World Example
A B2B SaaS company expects a spike when a large customer onboards thousands of users at once. The team writes an Artillery scenario for login and dashboard APIs.
Running it in Kubernetes with Testkube mirrors production infrastructure. They find a database connection limit and fix it before launch day.
Best Practices for Artillery Load Tests
- Ramp gradually: use warm-up phases before peak load.
- Model real behavior: base scenarios on actual traffic patterns.
- Watch percentiles: p95 and p99 latency reveal problems averages hide.
- Test close to production: run load from inside the cluster when possible.
- Automate baselines: run tests on every release and compare trends.