JMeter cheatsheet
A one-page reference for JMeter. For distributed testing and metrics deep-dives, see the complete guide.
📖 Full guide: JMeter →Test plan structure
Test Plan
└─ Thread Group
├─ HTTP Request Sampler
├─ Assertions (Response, Duration)
├─ Listeners (View Results Tree, Summary Report)
└─ Config Elements (CSV Data Set, HTTP Header Manager)
Thread group config
| Field | Meaning |
|---|---|
| Threads | concurrent virtual users |
| Ramp-up | seconds to start all threads |
| Loop count | iterations per thread |
Types of load testing
| Type | Goal |
|---|---|
| Load | expected traffic, verify SLAs |
| Stress | push past capacity, find breaking point |
| Spike | sudden traffic burst |
| Soak | sustained load over hours — finds leaks |
Correlation & parameterization
CSV Data Set Config → ${username}, ${password}
Regular Expression Extractor → capture sessionId from response
${sessionId} → reused in next request
Correlation replays dynamic values (tokens, IDs) that change per session.
Distributed (remote) testing
# on each worker
jmeter-server
# on the controller
jmeter -n -t plan.jmx -R worker1,worker2 -l results.jtl
Key metrics
- Throughput — requests/sec the system sustains.
- Latency vs response time — latency excludes time to read the full response.
- Error % — failed requests under load.
- 90th/95th/99th percentile — tail latency, more meaningful than average.
Non-GUI mode for CI
jmeter -n -t plan.jmx -l results.jtl -e -o report/
Always run load tests non-GUI — the GUI itself consumes resources and skews results.
Common pitfalls
- Running from a resource-starved laptop instead of a dedicated load generator.
- Ignoring think-time — unrealistic hammering vs real user pacing.
- Not correlating dynamic tokens — replays fail silently after session 1.