senke/veza

History

senke df8ce52a1e feat(v0.12.4): k6 load test for 1000 concurrent users Three scenarios: smoke (10 VUs), load (500 VUs), stress (1000 VUs). Tests tracks listing, search, track detail, and user profiles. Thresholds: p95 < 100ms, p99 < 200ms, error rate < 1%. Custom metrics for cache hit ratio tracking. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>		2026-03-11 09:58:06 +01:00
..
backend	feat(v0.12.4): k6 load test for 1000 concurrent users	2026-03-11 09:58:06 +01:00
chat	fix(release): v1.0.2 — Conformité complète V1_SIGNOFF (21 critères)	2026-03-03 21:18:53 +01:00
stream	chore: consolidate CI, E2E, backend and frontend updates	2026-02-17 16:43:21 +01:00
config.js	fix(release): v1.0.2 — Conformité complète V1_SIGNOFF (21 critères)	2026-03-03 21:18:53 +01:00
README.md	chore(release): v0.951 — Loadtest (500 req/s, 1000 WS, 50 uploads, perf indexes)	2026-03-02 19:22:38 +01:00
smoke.js	feat(loadtests): audit 3.2 — tests de charge k6 complets	2026-02-15 15:22:48 +01:00

README.md

Load Tests (k6)

Tests de charge pour la plateforme Veza. Audit 3.2.

Prérequis

k6 installé (brew install k6 ou sudo apt install k6)

Variables d'environnement

Variable	Défaut	Description
`BASE_URL`	`http://localhost:8080`	URL de l'API backend
`API_ORIGIN`	idem	Alias pour BASE_URL
`STREAM_ORIGIN`	`http://localhost:8082`	URL du stream server
`CHAT_ORIGIN`	`ws://localhost:8081`	URL WebSocket du chat server
`AUTH_TOKEN`	(vide)	JWT pour tests upload (optionnel)

Exécution locale

Smoke test (validation rapide, ~30s)

k6 run loadtests/smoke.js

Backend API

# Health, readyz, status
k6 run loadtests/backend/health.js

# Auth (register, login, /me, refresh)
k6 run loadtests/backend/auth.js

# Tracks (liste, search, détail)
k6 run loadtests/backend/tracks.js

# Uploads (AUTH_TOKEN optionnel — setup crée des users si absent)
AUTH_TOKEN=xxx k6 run loadtests/backend/uploads.js
# v0.951: 50 uploads concurrents pendant 2 min
k6 run loadtests/backend/uploads.js

# Playlists
k6 run loadtests/backend/playlists.js

# Marketplace
k6 run loadtests/backend/marketplace.js

# Scénario combiné (health, auth, tracks, playlists, marketplace)
k6 run loadtests/backend/full.js

# Stress test 500 req/s (v0.951) — login, tracks, search, products
# Seuils: P99 < 500ms, 500 VUs ramp 0→100→250→500 sur 3 min, maintien 2 min
k6 run loadtests/backend/stress_500rps.js
# Avec rapport JSON:
k6 run --out json=report.json loadtests/backend/stress_500rps.js

Stream server

# Health, healthz, readyz (stream server doit tourner)
k6 run loadtests/stream/http.js

Chat WebSocket

# Backend API + Chat server requis
k6 run loadtests/chat/websocket.js

# Stress test 1000 WebSocket concurrent (v0.951)
# Prérequis: Backend + Chat server, mémoire serveur < 2GB
k6 run loadtests/chat/stress_1000ws.js

Docker

docker run --rm -i --network=host \
  -v $(pwd)/loadtests:/scripts \
  -e BASE_URL=http://localhost:8080 \
  grafana/k6 run /scripts/smoke.js

CI

Le workflow .github/workflows/load-test-nightly.yml s'exécute :

Schedule : 2h UTC chaque nuit
Manual : workflow_dispatch

Il lance smoke + backend full, puis uploade les résultats en artifact.

Structure

loadtests/
├── README.md
├── config.js
├── smoke.js
├── backend/
│   ├── health.js
│   ├── auth.js
│   ├── tracks.js
│   ├── uploads.js
│   ├── playlists.js
│   ├── marketplace.js
│   ├── full.js
│   └── stress_500rps.js
├── stream/
│   └── http.js
└── chat/
    ├── websocket.js
    └── stress_1000ws.js