Skip to content

LOADTEST.md

Latest commit

 

History

History
167 lines (124 loc) · 4.37 KB

LOADTEST.md

File metadata and controls

167 lines (124 loc) · 4.37 KB
Raw

SyncDoc — Load Testing & Metrics Guide

Your stack runs at http://localhost:8080.

Fix & restart (run these now)

# 1. Stop whatever is currently running
docker-compose -f docker-compose.dev-lb.yml down

# 2. Start fresh with monitoring included
docker-compose -f docker-compose.dev-lb.yml up --build -d

# 3. Wait ~20 seconds, then check all 7 containers are up
docker-compose -f docker-compose.dev-lb.yml ps

Expected — all 7 should show "Up" or "healthy":

NAME                  STATUS
syncdoc_nginx         Up
syncdoc_app1          Up
syncdoc_app2          Up
syncdoc_postgres      Up (healthy)
syncdoc_redis         Up (healthy)
syncdoc_prometheus    Up
syncdoc_grafana       Up

Verify everything is working

# 1. App responds
curl http://localhost:8080/api/health
# → {"status":"ok","db":"postgres","instance":"app1",...}
# Run again — should sometimes say "app2" (load balancing)

# 2. Metrics endpoint works
curl http://localhost:8080/metrics | head -5
# → # HELP ... (Prometheus text format)

# 3. Prometheus is scraping
# Open http://localhost:9090/targets
# → Both syncdoc-app1 and syncdoc-app2 should show State: UP (green)

# 4. Grafana dashboard loaded
# Open http://localhost:3000
# Login: admin / admin
# You should see the SyncDoc dashboard with 10 panels
# (they'll be mostly empty — needs traffic first)

Install k6

# macOS
brew install k6

# Verify
k6 version

Run the load tests

Critical: use port 8080 in BASE_URL

cd loadtest

# Test 1: REST API (~2 min)
k6 run -e BASE_URL=http://localhost:8080 http-api.js

# Test 2: WebSocket sync (~2 min)  
k6 run -e BASE_URL=http://localhost:8080 websocket-sync.js

# Test 3: Stress test — ramps to 250 VUs (~4 min)
k6 run -e BASE_URL=http://localhost:8080 stress.js

Watch Grafana at http://localhost:3000 while tests run — you'll see the graphs fill in.

What good results look like

Test 1 output (http-api.js)

Doc create p50:  8ms       ← excellent
Doc create p95:  45ms      ← good (under 500ms threshold)
Throughput:      38 req/s
Error rate:      0.00%     ← no errors = pass

Test 2 output (websocket-sync.js)

WS connect p95:    180ms   ← good (under 2000ms threshold)
Update RTT p50:    25ms    ← excellent for local
Update RTT p95:    85ms    ← good (under 300ms threshold)
WS error rate:     0.00%   ← pass

What to look for in Grafana during tests

Panel What to watch
WS connections per instance Should split ~50/50 between app1 and app2
HTTP request rate Both instances should show traffic
Redis messages relayed Goes up during websocket test — proves cross-server sync
HTTP latency p50/p95/p99 p99 staying under 500ms is healthy
Memory heap Should stabilise, not grow continuously
Event loop lag Should stay under 10ms locally

Verify Redis cross-server sync

This proves your two containers are actually syncing via Redis:

# Terminal 1: watch Redis pub/sub traffic
docker exec syncdoc_redis redis-cli monitor | grep PUBLISH

# Terminal 2: open the app
open http://localhost:8080
# Create a doc, open same doc in a second browser window
# Type something — you should see PUBLISH ydoc:XXXX lines appear in Terminal 1
# Check which app each request hits
docker-compose -f docker-compose.dev-lb.yml logs app1 app2 | grep "Connected\|client"
# If you see connections on both app1 and app2, Redis sync is being exercised

Common issues

k6 shows 100% errors, 0ms response times → Wrong port. Make sure you use http://localhost:8080 not http://localhost

Prometheus targets show DOWN → Apps not running or metrics not exposed. Check:

docker-compose -f docker-compose.dev-lb.yml logs app1 | tail -20
curl http://localhost:8080/metrics | head -3

Grafana dashboard empty after tests → Go to http://localhost:9090/targets first. If targets are DOWN, Prometheus can't scrape — check app logs. If targets are UP but Grafana is empty, try changing the time range in Grafana to "Last 5 minutes".

Grafana shows "No data" → The SyncDoc dashboard may not have auto-loaded. Go to Dashboards → Browse → SyncDoc folder → SyncDoc dashboard. Or just go to Explore, select Prometheus datasource, and query: syncdoc_ws_connections_active to confirm data is flowing.