Checkpoint Lab — Controllers: Simple, Loop, Transaction, If, While, Switch, and Throughput

1. Assumptions and hard ceilings

2. Start clean and archive the tree

mkdir -p results/checkpoint results/parent results/broken results/repaired
python fixtures/controller_fixture.py --log results/checkpoint/server-events.jsonl
curl --fail --silent http://127.0.0.1:8000/health
curl --fail --silent http://127.0.0.1:8000/reset

Save/export the exact controller tree and resolved variables before running. The evidence packet should show every loop count, branch condition, Switch value, Throughput mode/count, and Transaction option.

3. Valid checkpoint tree

Thread Group — 1 thread × 2 iterations
└── Transaction Controller — Journey Transaction
    Parent: OFF
    Include timers/pre-post: OFF
    ├── Simple — Browse Phase
    │   └── Browse
    ├── Loop — SearchLoop ×2
    │   └── Search
    ├── Poll Reset → extractor POLL_MORE=true
    ├── While — WaitForReady (${POLL_MORE})
    │   └── Poll → extractor updates POLL_MORE (true then false)
    ├── If — Do Checkout (${DO_CHECKOUT}=true)
    │   └── Switch (${CHECKOUT_MODE}=standard)
    │       ├── standard → Checkout Standard
    │       ├── express → Checkout Express
    │       └── default → Checkout Default
    └── Throughput Controller — Upsell
        Total executions=1; Per User=OFF
        └── Upsell

4. Required predictions before execution

Prediction A — target requests:

Per outer iteration:
Browse 1
Search 2
Poll Reset 1
Poll 2
Checkout 1
= 7

Two outer iterations = 14
Upsell total-executions branch = 1
Expected target requests = 15

Prediction B — JTL rows, Transaction parent OFF:

15 child protocol rows + 2 transaction rows = 17 CSV rows

Prediction C — Transaction parent ON: target still receives 15 requests, but CSV JTL should contain two parent transaction rows rather than 15 separate child rows.

5. Run the valid checkpoint

jmeter -n   -t plans/checkpoint-controllers.jmx   -l results/checkpoint/results.jtl   -j results/checkpoint/jmeter.log   -Jjmeter.save.saveservice.print_field_names=true   -Jjmeter.save.saveservice.thread_counts=true

python tools/analyze_jtl_labels.py results/checkpoint/results.jtl
python tools/analyze_server_events.py results/checkpoint/server-events.jsonl
curl --fail --silent http://127.0.0.1:8000/stats

6. Verify valid-run counts

Expected target endpoint counts:

Expected additional transaction label count: Journey Transaction = 2. No transaction label should appear in the server event log.

7. Verify parent-sample reporting separately

Reset fixture and rerun the same journey with Generate Parent Sample=ON into results/parent/. Verify:

• target endpoint counts remain 15;
• CSV JTL contains two Journey Transaction parent rows;
• child HTTP results are sub-samples, not separate CSV rows;
• transaction success becomes false if any nested child fails.

8. Deliberate defect — Poll extractor scoped to Poll Reset only

Create a broken copy where the JSON extractor that updates POLL_MORE after Poll is mistakenly moved under Poll Reset. Now Poll responses no longer set false, so While would continue.

Contain the reproduction:

• 1 thread;
• Thread Group duration 4 seconds;
• 200 ms Constant Timer under Poll;
• abort if target Poll count exceeds 20.

Run into results/broken/. Expected: Poll count materially exceeds the valid count of two per journey while other child counts may not be reached because the thread remains inside While.

9. Diagnose the defect from evidence

Required interpretation:

1. JTL shows repeated Poll label.
2. Server event log independently confirms repeated /poll target requests.
3. POLL_MORE never receives the Poll response's false value because extractor scope is wrong.
4. The server is responding normally; the generator's controller state is causing excess traffic.
5. Thread Group duration prevents an unbounded safety incident but does not repair the business logic.

10. Repair the smallest layer

Restore the JSON extractor as a Post-Processor of Poll with default false. Keep the same Thread Group/While/fixture settings and rerun into results/repaired/.

Expected: Poll returns to exactly two executions per journey, the later Checkout/Upsell path executes, and request/JTL counts return to the valid predictions.

11. Optional second defect — wrong Loop Controller count

Change SearchLoop from 2 to 3 for one bounded run. Predict the delta before running:

2 outer iterations × +1 extra Search = +2 target requests and +2 Search JTL rows

This defect is useful for proving multiplicative loop math without involving server errors.

12. Generator/validity observation

Record Task Manager/top CPU/memory during valid and broken runs. The valid lab is tiny and should leave large injector headroom. If the generator is unexpectedly saturated, do not use the result to infer target capacity.

13. Required evidence packet

14. Verification checklist

• Target is exactly loopback:8000.
• Valid plan sends 15 target requests and creates 17 CSV rows in additional transaction mode.
• Parent mode changes CSV schema, not target traffic.
• Search loop count and While poll count match predictions.
• Only standard checkout branch executes in the valid plan.
• Upsell executes once total; it is not interpreted as an RPS setting.
• Broken While reproduction is hard-bounded and preserved before repair.
• Timers/assertions/processors are reviewed for inherited controller scope.

15. Validity statement

16. Cleanup

1. Stop the local fixture.
2. Preserve valid, parent, broken, and repaired evidence until review is complete.
3. Remove debug listeners and synthetic variable dumps from later load plans.
4. Restore all failure-reproduction hard bounds/configuration to the valid baseline.
5. No production target, credential, certificate, RMI engine, paid service, container, database, or system-wide JVM/OS setting was changed.

17. What Chapter 12 adds to the operating model

The performance-testing operating model now has a control-flow contract: outer iteration count, nested loop multiplication, branch conditions and populations, Switch mapping, While termination state/safety bound, Transaction reporting mode, branch-execution controller semantics, scope inheritance, predicted target requests, and expected JTL labels are reviewed before execution.

Chapter 13 moves naturally to Recording Web Traffic, HTTP(S) Test Script Recorder, and Certificate Setup. Recording can create a raw request sequence, but the controller skills from this chapter are what turn that recording into a maintainable business journey instead of replaying every captured request flatly.

Knowledge check

Official references and version notes

• Component Reference — Logic Controllers — Simple, Loop, Throughput, If, While, Switch, and Transaction Controller semantics.
• Elements of a Test Plan — scope and execution-order rules for samplers, controllers, timers, processors, and assertions.
• Functions and Variables — current JEXL3/Groovy/function/variable behavior used in conditions.
• Best Practices — GUI authoring versus CLI load execution and generator-validity guidance.
• Apache JMeter downloads — current production release and Java requirement.