Skip to content

Execution-Guide.md

Latest commit

 

History

History
116 lines (96 loc) · 4.37 KB

Execution-Guide.md

File metadata and controls

116 lines (96 loc) · 4.37 KB
Raw

Execution Guide

Three execution tiers: Local (your machine), RedLine13 (cloud, modest), and Production-Scale (distributed, multi-region).

A. Local Execution (start here — minimal load)

Prerequisites

  • Apache JMeter installed. Either add <JMETER>/bin to PATH, or set JMETER_HOME (e.g. C:\apache-jmeter-5.6.3).
  • Java 17+ (java -version).

Quickest smoke (GUI sanity check)

  1. Open JMeter GUI → File → Open → jmx/SmokeTest.jmx.
  2. Enable View Results Tree (currently disabled) temporarily.
  3. Click Start. Confirm green results and passing assertions.

    Never run Load/Stress/Spike/Soak in the GUI — GUI is for editing/debug only.

Non-GUI run (the real way) — Windows

# From PerformanceFramework/
.\scripts\run-test.ps1 -Plan SmokeTest -Env qa
.\scripts\run-test.ps1 -Plan EndToEndFlow -Env qa
# Small local load with overrides:
.\scripts\run-test.ps1 -Plan LoadTest -Env qa -Props @{users=20; rampup=30; duration=120}

Non-GUI run — Linux/macOS/CI

chmod +x scripts/run-test.sh
./scripts/run-test.sh SmokeTest qa
./scripts/run-test.sh LoadTest qa -Jusers=20 -Jrampup=30 -Jduration=120

Raw JMeter command (what the scripts wrap)

jmeter -n -t jmx/SmokeTest.jmx \
  -q config/env/qa.properties \
  -p config/user.properties \
  -l results/smoke.jtl -j results/jmeter.log \
  -e -o reports/smoke \
  -Jtestdata_file=data/testdata.csv

Outputs

  • results/<run-id>/results.jtl — raw samples (CSV).
  • reports/<run-id>/index.html — JMeter HTML dashboard (open in browser).
  • results/<run-id>/jmeter.log — engine log.

Do not attempt 5k/50k threads locally. One laptop ≈ a few hundred light threads before the generator itself becomes the bottleneck. Scale on RedLine13.

B. RedLine13 Execution (cloud, no edits required)

  1. Validate locally first (Section A) — only upload green plans.
  2. RedLine13 → New Test → JMeter Test.
  3. Attach files:
    • Plan: jmx/LoadTest.jmx (or any plan).
    • Additional file: data/testdata.csv.
  4. JMeter Properties box — paste your environment + load profile, e.g.: 
    base_url=app-perf.internal.company.com
protocol=https
port=443
users=500
rampup=180
duration=900
think_time_min=1000
think_time_range=3000
testdata_file=testdata.csv
  5. Generators: instance type + count + region (start: 1 × c5.large).
  6. Start and watch live RPS / response time / error% / active threads.
  7. Download the JTL when done; regenerate the HTML dashboard locally: 
    jmeter -g results/redline13-results.jtl -o reports/redline13-run

Because the JMX uses filename-only data references and ${__P()} everywhere, no field needs editing after upload — change behaviour via the Properties box.

C. Production-Scale Execution (distributed, multi-region)

For 5k → 50k VUs. Treat as a coordinated event, not an ad-hoc run.

Pre-flight checklist

  • AWS configured per docs/AWS-Setup.md (IAM key, EC2 quota, SGs).
  • SUT allows generator IP ranges (network/firewall change approved).
  • Change window / sign-off obtained (especially anything touching PROD).
  • Test data volume sufficient (unique users/bookings) — expand data/testdata.csv so threads don't collide unrealistically.
  • Monitoring ready on the SUT side (APM, infra metrics, DB, logs).
  • Baseline captured (a Smoke/Load run for comparison).

Run

  1. Configure multiple generator groups, one per region, with users split by the traffic-share table (RedLine13-Guide.md §4).
  2. Match instance type/count to the scaling table (RedLine13-Guide.md §5).
  3. Stagger ramp: set rampup to ≥ 300–600s for 10k+ so you observe degradation curves, not a thundering herd.
  4. Start; monitor both RedLine13 (client-side) and SUT APM (server-side). Correlate response-time inflection with server CPU/DB/queue saturation.
  5. For Stress, let the staged thread groups climb until error% spikes — that inflection is the breaking point. For Soak, hold steady for hours and watch for slow memory growth / GC pressure / connection-pool exhaustion.

After

  • Confirm all generators terminated in EC2 (cost hygiene).
  • Archive JTL + HTML dashboard with the run metadata (date, build, profile).
  • Compare P95/P99, throughput, and error% against the baseline and SLAs.