Files

Ho Ngoc Hai 33c2e5ac1d feat(load-tests): add K6 coverage for search, admin, and MCP endpoints

Add three new K6 load test scripts to cover previously untested API surfaces:

- search-advanced.js: Combined geo + text + filter queries, paginated deep
  search, and sort variations against /search and /search/geo (300 peak VUs)
- admin.js: Moderation queue CRUD, bulk moderation, dashboard stats, audit
  logs, and user management endpoints (50 peak VUs)
- mcp.js: MCP server discovery, SSE connection, property-search tool calls,
  valuation/batch-valuation, and feature extraction (120 peak VUs)

Also updates README with new suite documentation, per-suite custom thresholds,
and adds the new suites to the CI workflow_dispatch selector.

Co-Authored-By: Paperclip <noreply@paperclip.ing>

2026-04-11 20:14:52 +07:00

2.9 KiB

Raw Blame History

K6 Load Testing — Goodgo Platform

Overview

Performance load tests for critical API paths using K6.

Test Suites

Script	Target	Peak VUs	Duration
`auth.js`	Login/Register	100	2min
`listings.js`	Search + Detail	500	3min
`search.js`	Text + Geo search	200	3min
`search-advanced.js`	Combined geo + text + filters, pagination	300	3min
`admin.js`	Moderation queue, dashboard, audit logs	50	2.5min
`mcp.js`	MCP server discovery, property-search, valuation	120	2.5min
`payments.js`	Create + List	50	2min

SLA Thresholds

Metric	Threshold
p50 latency	< 200ms
p95 latency	< 500ms
p99 latency	< 1000ms
Error rate	< 1%

Per-Suite Custom Thresholds

Suite	Metric	Threshold
search-advanced	advanced_search_duration p95	< 800ms
search-advanced	geo_filter_search_duration p95	< 800ms
admin	moderation_action_duration p95	< 800ms
admin	admin_dashboard_duration p95	< 500ms
mcp	mcp_property_search_duration p95	< 1500ms
mcp	mcp_valuation_duration p95	< 1000ms
mcp	mcp_batch_valuation_duration p95	< 2000ms

Prerequisites

# Install K6
brew install k6        # macOS
# or: https://grafana.com/docs/k6/latest/set-up/install-k6/

# Start the API
pnpm --filter @goodgo/api run dev

Running Tests

# Run individual suite
k6 run load-tests/scripts/auth.js
k6 run load-tests/scripts/listings.js
k6 run load-tests/scripts/search.js
k6 run load-tests/scripts/search-advanced.js
k6 run load-tests/scripts/admin.js
k6 run load-tests/scripts/mcp.js
k6 run load-tests/scripts/payments.js

# Run against a custom API URL
k6 run -e API_BASE_URL=https://staging.goodgo.vn load-tests/scripts/auth.js

# Run all suites sequentially
for f in load-tests/scripts/*.js; do k6 run "$f"; done

# Export results as JSON
k6 run --out json=results/auth.json load-tests/scripts/auth.js

CI Integration

The load-test.yml workflow runs as an optional manual stage in GitHub Actions. Trigger via workflow_dispatch with a suite selector.

Directory Structure

load-tests/
├── lib/
│   └── config.js              # Shared config, helpers, SLA thresholds
├── scripts/
│   ├── auth.js                # Auth flow load tests
│   ├── listings.js            # Listings search + detail
│   ├── search.js              # Full-text + geo search (basic)
│   ├── search-advanced.js     # Combined geo + text + filter search
│   ├── admin.js               # Admin moderation, dashboard, audit
│   ├── mcp.js                 # MCP server endpoints (property-search, valuation)
│   └── payments.js            # Payment creation + listing
├── results/
│   └── BASELINE-REPORT.md     # Baseline performance report
└── README.md

2.9 KiB Raw Blame History