Starter Script Template

Rather than starting a new project from an empty Test Plan, start from this one. It has the whole checklist from Script Requirements already wired in.

📥 Download loadlitmus-starter.jmx — source: static/templates/loadlitmus-starter.jmx in this repository.

What is in it

Warning: Test Editor is in beta and has known bugs. Use it to look at a script, not as the place you build one. Keep a copy of the .jmx before saving from the editor, and check changes in JMeter's own GUI.

Test Plan  ← comment: what this test answers, and what it does NOT
├── UDV "Environment"        host · port · protocol · password · smoke · thinkMin · thinkMax
├── HTTP Request Defaults · Global Headers · Cookie Manager · Cache Manager
└── Thread Group "01 - <PERSONA A>"     users/ramp/loops = ${__P(a.users)} etc.
    ├── CSV - users                     ${__P(user.dir)}/${__P(a.csv,test_data/users.csv)}
    ├── TC01_Login
    │   ├── 01_GET_Login     Init all vars = NOT_FOUND · extract csrf · assert 200
    │   └── 02_POST_Login    think time · extract landing_url · assert not bounced · save failed response
    ├── If: logged in        ${__jexl3("${landing_url}" != "NOT_FOUND")}
    │   └── TC02_<READ_ONLY_JOURNEY>    03_GET_<PAGE> · assert 200 · assert content · save failed response
    └── If: smoke == false
        └── TC03_<SUBMIT>    04_POST_<SUBMIT> · assert saved · save failed response

Feature	Why it is there
${__P(...)} directly in the Thread Group	Keeps the Load card editable
UDV control panel with a description per row	One place to change settings; explains itself to the next person
password with no default	Typed into the parameter form, never stored in the file
${__P(user.dir)}/... CSV path	Same path works on the master and on every slave
NOT_FOUND init + If guards	A failed login stops that user instead of flooding the error rate
smoke flag around the write steps	Smoke runs leave the system untouched
Result Savers, failures only	Fleet can collect the actual failing pages
Content assertions, not just HTTP 200	A 200 carrying an error page is still a 200
No listeners, no Backend Listener	LoadLitmus writes the JTL and injects monitoring itself

Fill in

Placeholder	Where
<CLIENT> <SYSTEM> - <the one question...>	Test Plan name and its comment
host, protocol, port defaults	UDV "Environment"
/login, /home, /submit	Sampler paths
name="_token" value="([^"]+)"	The csrf extractor's regex
<SOMETHING ONLY THIS PAGE CONTAINS>	Content assertions
USERNAME	CSV column names
01 - <PERSONA A>, TC02_<READ_ONLY_JOURNEY>	Thread group and transaction names

Adding a second persona

Copy the whole Thread Group, rename it 02 - <PERSONA B>, and rename its parameters a.users → b.users, a.ramp → b.ramp, a.loops → b.loops, a.csv → b.csv. The Runner then shows one load row per persona.

Keep every group a classic Thread Group. A setUp or tearDown group switches the Load card off, and you go back to setting load from the parameter form.

Running it

1. Copy the file into data/projects/<slug>/test_plan/ and rename it.
2. Put the CSV in test_data/ with a USERNAME header.
3. On the Runner page, select it and check that the Load card shows one row per persona and Test Data lists the CSV.
4. Smoke run: 1 user, 1 iteration, smoke=true.
5. Real run: raise the load, set smoke=false when the write steps should fire.

What was verified

Run against a local stub server with JMeter 5.6.3 on 2026-09-23, and through LoadLitmus 1.4.5:

Check	Result
2 users, smoke=false	All 4 samplers and 3 transactions pass
smoke=true (default)	The submit step is skipped
Broken login	The guard stops the user after 02_POST_Login; the failing page lands in failed_response/02_login/alice.unknown
LoadLitmus Load card	Editable; a 50 VU / 60 s ramp / 5 iteration override was applied to the run
Secret scan	No tokens, hosts or absolute paths in the file

Habits worth keeping

The template is the distillation of several scripts from finished projects. Habits worth naming:

• A Test Plan comment stating the one question the test answers, a description on every variable, and a setUp "preflight" check that catches a wrong environment before any load starts.
• One OAuth token fetched once and shared through props.put(), so thousands of users do not each log in.
• Retry loops around an unreliable SSO, and a Synchronizing Timer to release users together.
• NOT_FOUND guards, and failed responses saved per user and per step.
• ${persona} in every sampler name, so the report splits by role.