Add k6 load test scripts for Voiceprint verification endpoints (FRE-4930)

- k6 script with P99 latency thresholds (enrollment <500ms, verification <250ms, model retrieval <100ms)
- Configurable 500 req/s sustained throughput for 5 minutes
- Mixed workload scenario + individual endpoint scenarios
- GitHub Actions workflow for automated load testing
- Runner script with environment configuration
- JSON result export for CI artifact collection
- .gitignore entry for load test results
This commit is contained in:
Senior Engineer
2026-05-09 07:50:29 -04:00
committed by Michael Freno
parent bce4787802
commit cb5851ec8c
5 changed files with 402 additions and 0 deletions

54
.github/workflows/load-test.yml vendored Normal file
View File

@@ -0,0 +1,54 @@
name: Load Test
on:
push:
branches: [main]
workflow_dispatch:
inputs:
target_rps:
description: 'Target requests per second'
required: false
default: '500'
duration:
description: 'Test duration'
required: false
default: '300s'
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
jobs:
load-test-voiceprint:
name: Voiceprint Load Test
runs-on: ubuntu-latest
timeout-minutes: 15
steps:
- uses: actions/checkout@v4
- name: Install k6
run: |
curl -s https://github.com/grafana/k6/releases/download/v0.50.0/k6-linux-amd64.tar.gz -L | tar xz
sudo mv k6 /usr/local/bin/
k6 version
- name: Run Voiceprint load tests
run: |
cd load-tests/voiceprint
./run.sh mixed
env:
VOICEPRINT_BASE_URL: ${{ secrets.LOAD_TEST_BASE_URL || 'http://localhost:3000' }}
API_TOKEN: ${{ secrets.LOAD_TEST_API_TOKEN || 'test-token' }}
TARGET_RPS: ${{ github.event.inputs.target_rps || '500' }}
DURATION: ${{ github.event.inputs.duration || '300s' }}
ENROLLMENT_P99_MS: '500'
VERIFICATION_P99_MS: '250'
MODEL_RETRIEVAL_P99_MS: '100'
- name: Upload results
if: always()
uses: actions/upload-artifact@v4
with:
name: load-test-results-${{ github.sha }}
path: load-tests/voiceprint/results/
retention-days: 30

1
.gitignore vendored
View File

@@ -3,3 +3,4 @@ dist
.env .env
*.log *.log
.DS_Store .DS_Store
load-tests/voiceprint/results/

View File

@@ -0,0 +1,19 @@
# Voiceprint Load Test Configuration
# Copy to .env and adjust values
# Base URL of the Voiceprint API
VOICEPRINT_BASE_URL=http://localhost:3000
# API authentication token
API_TOKEN=test-token
# Test duration (default: 300s = 5 minutes)
DURATION=300s
# Target requests per second (default: 500)
TARGET_RPS=500
# P99 latency thresholds in milliseconds
ENROLLMENT_P99_MS=500
VERIFICATION_P99_MS=250
MODEL_RETRIEVAL_P99_MS=100

69
load-tests/voiceprint/run.sh Executable file
View File

@@ -0,0 +1,69 @@
#!/usr/bin/env bash
# Run k6 load tests for Voiceprint endpoints
# Usage: ./run.sh [scenario]
# scenario: mixed (default), enrollment, verification, model-retrieval
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
cd "$SCRIPT_DIR"
# Load environment variables from .env if present
if [[ -f .env ]]; then
set -a
source .env
set +a
fi
SCENARIO="${1:-mixed}"
OUTPUT_DIR="${SCRIPT_DIR}/results"
TIMESTAMP="$(date +%Y%m%d-%H%M%S)"
mkdir -p "$OUTPUT_DIR"
echo "=== Voiceprint Load Test ==="
echo "Scenario: $SCENARIO"
echo "Target RPS: ${TARGET_RPS:-500}"
echo "Duration: ${DURATION:-300s}"
echo "Base URL: ${VOICEPRINT_BASE_URL:-http://localhost:3000}"
echo ""
case "$SCENARIO" in
mixed)
k6 run voiceprint.js \
--out json="$OUTPUT_DIR/results-${TIMESTAMP}.json" \
<<EOF
EOF
;;
enrollment)
k6 run --scenario enrollment_only voiceprint.js \
--out json="$OUTPUT_DIR/results-${TIMESTAMP}.json"
;;
verification)
k6 run --scenario verification_only voiceprint.js \
--out json="$OUTPUT_DIR/results-${TIMESTAMP}.json"
;;
model-retrieval)
k6 run --scenario model_retrieval_only voiceprint.js \
--out json="$OUTPUT_DIR/results-${TIMESTAMP}.json"
;;
*)
echo "Unknown scenario: $SCENARIO"
echo "Available: mixed, enrollment, verification, model-retrieval"
exit 1
;;
esac
EXIT_CODE=$?
if [[ $EXIT_CODE -eq 0 ]]; then
echo ""
echo "✅ All thresholds passed!"
echo "Results saved to: $OUTPUT_DIR/results-${TIMESTAMP}.json"
else
echo ""
echo "❌ Thresholds failed. Check output above."
echo "Results saved to: $OUTPUT_DIR/results-${TIMESTAMP}.json"
fi
exit $EXIT_CODE

View File

@@ -0,0 +1,259 @@
import http from 'k6/http';
import { check, sleep } from 'k6';
import { Rate, Trend } from 'k6/metrics';
// ── Configuration ────────────────────────────────────────────────────────────
const BASE_URL = __ENV.VOICEPRINT_BASE_URL || 'http://localhost:3000';
const API_TOKEN = __ENV.API_TOKEN || 'test-token';
const DURATION = __ENV.DURATION || '300s'; // 5 minutes
const TARGET_RPS = parseInt(__ENV.TARGET_RPS || '500', 10);
// P99 latency thresholds (ms)
const THRESHOLDS = {
enrollment: parseInt(__ENV.ENROLLMENT_P99_MS || '500', 10),
verification: parseInt(__ENV.VERIFICATION_P99_MS || '250', 10),
modelRetrieval: parseInt(__ENV.MODEL_RETRIEVAL_P99_MS || '100', 10),
};
// ── Custom Metrics ───────────────────────────────────────────────────────────
const enrollmentLatency = new Trend('enrollment_p99');
const verificationLatency = new Trend('verification_p99');
const modelRetrievalLatency = new Trend('model_retrieval_p99');
const enrollmentSuccess = new Rate('enrollment_success');
const verificationSuccess = new Rate('verification_success');
const modelRetrievalSuccess = new Rate('model_retrieval_success');
// ── Helpers ──────────────────────────────────────────────────────────────────
function uuidv4() {
return 'xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx'.replace(/[xy]/g, (c) => {
const r = (Math.random() * 16) | 0;
const v = c === 'x' ? r : (r & 0x3) | 0x8;
return v.toString(16);
});
}
// Generate a realistic audio payload (base64-encoded WAV-like buffer)
// ~3 seconds of 16kHz mono 16-bit audio = ~96KB
function generateAudioPayload() {
const size = 96000;
const audio = new Array(size);
for (let i = 0; i < size; i++) {
audio[i] = Math.floor(Math.random() * 256);
}
return btoa(String.fromCharCode(...audio.slice(0, 2048)));
}
const headers = {
'Content-Type': 'application/json',
'Authorization': `Bearer ${API_TOKEN}`,
};
// ── Scenario: Enrollment (POST /voiceprint/enroll) ──────────────────────────
function testEnrollment() {
const payload = JSON.stringify({
name: `voice_profile_${uuidv4()}`,
audio: generateAudioPayload(),
});
const res = http.post(`${BASE_URL}/voiceprint/enroll`, payload, { headers });
const duration = res.timings.duration;
enrollmentLatency.add(duration);
const success = res.status === 201;
enrollmentSuccess.add(success);
check(res, {
'enrollment: status 201': (r) => r.status === 201,
'enrollment: has enrollment.id': (r) => {
try {
const json = JSON.parse(r.body);
return !!json.enrollment && !!json.enrollment.id;
} catch {
return false;
}
},
`enrollment: P99 < ${THRESHOLDS.enrollment}ms`: (r) => duration < THRESHOLDS.enrollment,
});
return res.json()?.enrollment?.id || uuidv4();
}
// ── Scenario: Verification (POST /voiceprint/analyze) ───────────────────────
function testVerification() {
const payload = JSON.stringify({
audio: generateAudioPayload(),
});
const res = http.post(`${BASE_URL}/voiceprint/analyze`, payload, { headers });
const duration = res.timings.duration;
verificationLatency.add(duration);
const success = res.status === 201;
verificationSuccess.add(success);
check(res, {
'verification: status 201': (r) => r.status === 201,
'verification: has analysis.id': (r) => {
try {
const json = JSON.parse(r.body);
return !!json.analysis && !!json.analysis.id;
} catch {
return false;
}
},
`verification: P99 < ${THRESHOLDS.verification}ms`: (r) => duration < THRESHOLDS.verification,
});
return res.json()?.analysis?.id || uuidv4();
}
// ── Scenario: Model Retrieval (GET /voiceprint/results/:id) ─────────────────
function testModelRetrieval(modelId) {
const id = modelId || uuidv4();
const res = http.get(`${BASE_URL}/voiceprint/results/${id}`, { headers });
const duration = res.timings.duration;
modelRetrievalLatency.add(duration);
// 200 = found, 404 = not found (both valid for load testing)
const success = res.status === 200 || res.status === 404;
modelRetrievalSuccess.add(success);
check(res, {
'model_retrieval: status 200 or 404': (r) => r.status === 200 || r.status === 404,
`model_retrieval: P99 < ${THRESHOLDS.modelRetrieval}ms`: (r) => duration < THRESHOLDS.modelRetrieval,
});
}
// ── Default Scenario: Weighted mixed workload ────────────────────────────────
export const options = {
scenarios: {
sustained_load: {
executor: 'constant-arrival-rate',
duration: DURATION,
rate: TARGET_RPS,
preAllocatedVUs: 20,
maxVUs: 100,
startTime: '0s',
exec: 'mixedWorkload',
tags: { scenario: 'sustained_load' },
},
},
thresholds: {
`enrollment_p99`: [`p(99)<${THRESHOLDS.enrollment}`],
`verification_p99`: [`p(99)<${THRESHOLDS.verification}`],
`model_retrieval_p99`: [`p(99)<${THRESHOLDS.modelRetrieval}`],
`enrollment_success`: ['rate>0.95'],
`verification_success`: ['rate>0.95'],
`model_retrieval_success`: ['rate>0.95'],
http_req_duration: [`p(95)<400`, `p(99)<500`],
http_req_failed: ['rate<0.05'],
},
};
// Mixed workload: 30% enrollment, 45% verification, 25% model retrieval
export function mixedWorkload() {
const rand = Math.random();
if (rand < 0.3) {
const modelId = testEnrollment();
sleep(0.1);
testModelRetrieval(modelId);
} else if (rand < 0.75) {
const modelId = testVerification();
sleep(0.05);
testModelRetrieval(modelId);
} else {
testModelRetrieval();
}
sleep(0.05);
}
// ── Individual endpoint scenarios for targeted testing ───────────────────────
export const endpointScenarios = {
enrollment_only: {
executor: 'constant-arrival-rate',
duration: DURATION,
rate: TARGET_RPS,
preAllocatedVUs: 20,
maxVUs: 100,
exec: 'enrollmentOnly',
startTime: '0s',
tags: { scenario: 'enrollment_only' },
},
verification_only: {
executor: 'constant-arrival-rate',
duration: DURATION,
rate: TARGET_RPS,
preAllocatedVUs: 20,
maxVUs: 100,
exec: 'verificationOnly',
startTime: '0s',
tags: { scenario: 'verification_only' },
},
model_retrieval_only: {
executor: 'constant-arrival-rate',
duration: DURATION,
rate: TARGET_RPS,
preAllocatedVUs: 20,
maxVUs: 100,
exec: 'modelRetrievalOnly',
startTime: '0s',
tags: { scenario: 'model_retrieval_only' },
},
};
export function enrollmentOnly() {
testEnrollment();
sleep(0.1);
}
export function verificationOnly() {
testVerification();
sleep(0.05);
}
export function modelRetrievalOnly() {
testModelRetrieval();
sleep(0.02);
}
// ── Summary Hook ─────────────────────────────────────────────────────────────
export function handleSummary(data) {
return {
'stdout': `\n=== Voiceprint Load Test Results ===\n`,
'summary.json': JSON.stringify({
timestamp: new Date().toISOString(),
duration: DURATION,
targetRPS: TARGET_RPS,
thresholds: THRESHOLDS,
metrics: {
enrollment: {
p99: data.metrics.enrollment_p99?.values['p(99)']?.toFixed(2) || 'N/A',
p95: data.metrics.enrollment_p99?.values['p(95)']?.toFixed(2) || 'N/A',
avg: data.metrics.enrollment_p99?.values.avg?.toFixed(2) || 'N/A',
count: data.metrics.enrollment_p99?.values.count || 0,
successRate: (data.metrics.enrollment_success?.values.rate || 0) * 100 + '%',
},
verification: {
p99: data.metrics.verification_p99?.values['p(99)']?.toFixed(2) || 'N/A',
p95: data.metrics.verification_p99?.values['p(95)']?.toFixed(2) || 'N/A',
avg: data.metrics.verification_p99?.values.avg?.toFixed(2) || 'N/A',
count: data.metrics.verification_p99?.values.count || 0,
successRate: (data.metrics.verification_success?.values.rate || 0) * 100 + '%',
},
modelRetrieval: {
p99: data.metrics.model_retrieval_p99?.values['p(99)']?.toFixed(2) || 'N/A',
p95: data.metrics.model_retrieval_p99?.values['p(95)']?.toFixed(2) || 'N/A',
avg: data.metrics.model_retrieval_p99?.values.avg?.toFixed(2) || 'N/A',
count: data.metrics.model_retrieval_p99?.values.count || 0,
successRate: (data.metrics.model_retrieval_success?.values.rate || 0) * 100 + '%',
},
},
passed: Object.entries(data.metrics).every(
([_, metric]) => metric?.thresholds?.every?.((t) => t.pass)
),
}, null, 2),
};
}