Checkpoint Lab — Pre-Processors, Post-Processors, Extractors, and Correlation

1. Assumptions and hard ceilings

2. Prepare a clean workspace

jmeter-ch09-checkpoint/
├── fixtures/correlation_fixture.py
├── plans/checkpoint-good.jmx
├── plans/checkpoint-broken.jmx
├── tools/analyze_correlation.py
├── evidence/
│   ├── predictions.txt
│   ├── extractor-settings.txt
│   └── validity.txt
└── results/
    ├── good/
    ├── broken/
    └── repaired/

Start the fixture with a fresh event log and preflight /health.

3. Build the correct checkpoint tree

Test Plan
└── Thread Group — 2 users × 2 loops
    ├── HTTP Request Defaults — HttpClient4 / 127.0.0.1:8000
    ├── Start Session — GET /session/start
    │   ├── JSON JMESPath Extractor
    │   │   variable: CORR_TOKEN
    │   │   expression: session.token
    │   │   match: 1
    │   │   default: CORR_MISSING
    │   └── JSR223 Assertion — CORR_TOKEN must not be CORR_MISSING
    └── Use Session — GET /session/use?token=${CORR_TOKEN}&prepared=${PREPARED_TOKEN}&thread=${__threadNum}
        ├── JSR223 PreProcessor — uppercase CORR_TOKEN → PREPARED_TOKEN
        └── JSON JMESPath Assertion — status == accepted

4. Predict before running

1. Four Start Session responses should issue four distinct synthetic tokens/session IDs.
2. Each Use Session should be accepted only when it carries the token produced earlier in its own thread/iteration plus the correct prepared value.
3. Thread echo should show both thread 1 and 2.
4. JTL should contain eight successful samples and zero failures in the good profile.
5. Server logs should contain token fingerprints, not full token values.

5. One-thread authoring proof

Before the two-thread run, temporarily use 1 thread × 1 loop with Debug Sampler between Start and Use. Verify:

• CORR_TOKEN has a synthetic tok-... value;
• PREPARED_TOKEN is not set until the PreProcessor runs before Use;
• Use Session returns status=accepted.

Then disable Debug Sampler/View Results Tree for load execution.

6. Run the good two-thread profile

mkdir -p results/good
jmeter -n \
  -t plans/checkpoint-good.jmx \
  -l results/good/results.jtl \
  -j results/good/jmeter.log \
  -Jjmeter.save.saveservice.print_field_names=true \
  -Jjmeter.save.saveservice.response_code=true \
  -Jjmeter.save.saveservice.response_message=true \
  -Jjmeter.save.saveservice.assertion_results_failure_message=true
python tools/analyze_correlation.py results/good/results.jtl
curl --fail --silent http://127.0.0.1:8000/stats

7. Verify independent multi-user state

Inspect the synthetic server JSONL:

• 4 Start Session entries with distinct session IDs/token fingerprints;
• 4 Use Session entries with prepared_ok=true;
• Use entries include thread echo 1 or 2;
• no Use request has token_fp=missing;
• JTL good profile is 8/8 successful.

Exact request interleaving can differ. Independence is proven by unique producer state and valid downstream use, not by assuming a fixed thread execution order.

8. Intentionally break the extractor expression

Copy the plan to checkpoint-broken.jmx, reduce to 1 thread × 1 loop, and change only:

JMESPath expression:
  session.token
becomes:
  session.missingToken

Keep Default Value = CORR_MISSING. Predict that Start Session still returns HTTP 200 and contains the real token, but the extractor produces the sentinel and the extraction-guard assertion fails.

9. Preserve the broken run

jmeter -n \
  -t plans/checkpoint-broken.jmx \
  -l results/broken/results.jtl \
  -j results/broken/jmeter.log \
  -Jjmeter.save.saveservice.print_field_names=true \
  -Jjmeter.save.saveservice.assertion_results_failure_message=true
python tools/analyze_correlation.py results/broken/results.jtl

Do not delete or overwrite these artifacts after repair.

10. Diagnose the broken chain

11. Apply the least destructive repair

Restore only the extractor expression to session.token. Do not paste a token into the JMX, loosen the server, convert the token into a global property, or add retries. Rerun the original two-thread profile into results/repaired/ and compare with the good baseline.

12. Optional second failure: scope, not expression

For a one-thread diagnostic only, move the correct extractor under /noise or Thread Group. Preserve how the variable becomes missing/overwritten. Restore the child-of-producer placement. This separates “wrong expression” from “wrong sample/scope” as two distinct correlation failure classes.

13. Generator and validity check

Record CPU/memory during the tiny good/repaired runs. Built-in JSON extraction plus one cached Groovy uppercase operation should have negligible cost here. If generator saturation appears, stop; do not infer target capacity from an invalid injector.

14. Required evidence packet

15. Verification checklist

• Target exactly 127.0.0.1:8000.
• Good/repaired profile = 2 threads × 2 loops; broken profile = 1 × 1.
• Extractor is child of Start Session and uses session.token in good/repaired plans.
• Default sentinel is explicit and producer assertion rejects it.
• PreProcessor is child of Use Session and reads/writes vars, not props.
• Server evidence proves distinct token fingerprints/session IDs and prepared_ok=true.
• No real token/credential exists in JMX/JTL/logs.
• Broken evidence remains preserved after repair.

16. Validity statement

17. Cleanup

1. Stop the local fixture.
2. Keep good/broken/repaired artifacts until review is complete.
3. Remove Debug Sampler/View Results Tree from load profiles.
4. Do not carry synthetic token dumps into real test practices.
5. No production target, credential, certificate, remote engine, database, plugin, container, or system-wide JVM/OS setting was changed.

18. What Chapter 09 adds to the operating model

The performance-testing operating model now includes a correlation provenance contract: producer sampler, extraction source/expression/default, processor scope, variable ownership, transformation logic, consumer request, failure policy, and redaction strategy are all reviewable.

The current repository curriculum bridges next to Chapter 10, Regular Expressions, CSS/JQuery, XPath, JSONPath, and Boundary Extraction, where each extractor family is explored in depth. The published site may lag curriculum updates, so the repository curriculum remains the chapter-path/title source for generated content.

Knowledge check

Official references and version notes

• Elements of a Test Plan — Pre-Processor/Post-Processor purpose, scope, and execution order: configuration → pre-processors → timers → sampler → post-processors → assertions → listeners.
• Component Reference — Regular Expression, JSON/JMESPath, Boundary, JSR223 Pre/Post Processor, User Parameters, and Result Status Action Handler semantics.
• Functions and Variables — thread-local JMeter variables versus process-wide JMeter properties.
• Best Practices — CLI load execution and scripting/performance guidance.
• Apache JMeter downloads — current stable release and Java requirement.