Preventing 500 Errors During Sales: Effective Load Testing Strategies

500 errors during sales are a typical pain for e-commerce. If a peak load of 10,000 RPS crashes the server, you lose up to $10,000 per hour. To avoid downtime, we conduct performance testing with Artillery. This tool simulates thousands of simultaneous users, showing how the system behaves under pressure. Before launching promotions or new functionality, we model different test flows: catalog browsing, search, and checkout. As a result, you get not just a report, but specific optimization steps—from caching configuration to database indexes. Our clients save up to $5,000 per incident by timely identifying bottlenecks. Project evaluation takes 1 day, after which we provide a detailed report with recommendations. Contact us to get a consultation.

Load Testing with Artillery Helps Identify Bottlenecks

Artillery is a Node.js tool with YAML configuration, supporting HTTP and WebSocket. It generates load in phases: warm-up, ramp-up, peak. At each stage, we monitor response times, failure rates, and throughput. If p99 exceeds 1 second or 5xx errors >1%—it's a signal to optimize. For example, in one Laravel project, high latency occurred due to an N+1 query to the database when browsing the catalog. A test with 50 RPS revealed p95 growth to 3 s. After adding eager loading, p95 dropped to 200 ms. Learn more about performance testing on Wikipedia.

Why Choose Artillery for Load Testing

Artillery is simpler to configure than K6: no JS code needed for simple scenarios, just YAML. Compared to JMeter, Artillery integrates more easily into CI/CD: one command artillery run and a JSON report. Load can be distributed across multiple workers, achieving up to 100,000 RPS from a single machine (with proper configuration). In our tests, Artillery generates reports 2x faster than K6 under similar load.

# tests/load/basic.yml config: target: "http://localhost:3000" phases: - duration: 60 arrivalRate: 5 name: Warm up - duration: 120 arrivalRate: 20 name: Ramp up load - duration: 300 arrivalRate: 50 name: Sustained load - duration: 60 arrivalRate: 100 name: Stress test defaults: headers: Accept: "application/json" Content-Type: "application/json" ensure: p99: 1000 p95: 500 maxErrorRate: 1 scenarios: - name: Browse catalog weight: 60 flow: - get: url: "/api/products" expect: - statusCode: 200 - hasProperty: "data" - think: 2 - get: url: "/api/products/{{ $randomNumber(1, 100) }}" - name: Search weight: 30 flow: - get: url: "/api/search?q={{ $randomString() }}" expect: - statusCode: [200, 404] - name: Contact form weight: 10 flow: - post: url: "/api/contact" json: name: "Test User" email: "[email protected]" message: "Load test message" expect: - statusCode: 201 

How to Write a Load Test Scenario in 5 Steps

1. Identify critical API endpoints: catalog, search, cart, checkout.
2. Define SLA: p99 < 1 s, error rate < 1%.
3. Write YAML definition with load phases: warm-up, ramp-up, peak.
4. Add authorization if required (token capture).
5. Run test locally, verify correctness, then integrate into pipeline.

Main Artillery Scenarios

Authenticated Scenario

# tests/load/authenticated.yml config: target: "http://localhost:3000" phases: - duration: 300 arrivalRate: 20 variables: users: - email: "[email protected]" password: "pass123" - email: "[email protected]" password: "pass456" scenarios: - name: Authenticated user flow flow: - post: url: "/api/auth/login" json: email: "{{ users[0].email }}" password: "{{ users[0].password }}" capture: - json: "$.access_token" as: "token" expect: - statusCode: 200 - get: url: "/api/user/profile" headers: Authorization: "Bearer {{ token }}" expect: - statusCode: 200 - post: url: "/api/orders" headers: Authorization: "Bearer {{ token }}" json: product_id: 1 quantity: 1 expect: - statusCode: 201 capture: - json: "$.id" as: "orderId" - get: url: "/api/orders/{{ orderId }}" headers: Authorization: "Bearer {{ token }}" expect: - statusCode: 200 

Custom JS Processor

# tests/load/custom.yml config: processor: "./processor.js" scenarios: - name: Dynamic flow flow: - function: "generateDynamicPayload" - post: url: "/api/data" json: "{{ payload }}" 

// processor.js module.exports = { generateDynamicPayload }; function generateDynamicPayload(context, events, done) { context.vars.payload = { id: Math.floor(Math.random() * 10000), timestamp: new Date().toISOString(), data: Array.from({ length: 10 }, (_, i) => ({ key: `item_${i}`, value: Math.random() })), }; return done(); } 

GitHub Actions Integration

- name: Load Test run: | artillery run --output results.json tests/load/basic.yml artillery report --output load-report.html results.json - name: Check SLA run: | ERRORS=$(cat results.json | jq '.aggregate.counters["http.codes.5xx"] // 0') P99=$(cat results.json | jq '.aggregate.latency.p99') if [ "$ERRORS" -gt "10" ] || [ "$(echo "$P99 > 2000" | bc)" = "1" ]; then echo "Load test failed: too many errors or high latency" exit 1 fi 

Comparison of Load Testing Tools

Common Mistakes in Load Testing

• Testing only one API endpoint, ignoring business processes.
• Incorrect emulation of user behavior (e.g., missing think time).
• Running load from a single IP (network-level blocking).
• Ignoring application-level caching.
• Lack of server-side monitoring (CPU, RAM, DB connections).

Load Test Development Stages

What's Included

• Documentation: technical specification, scenario description, run instructions.
• Scenarios: 3–5 YAML files with support for authorization, WebSocket, custom processors.
• Report: HTML dashboard with metrics (response time, RPS, errors), JSON data for CI.
• Recommendations: list of bottlenecks and specific improvement steps (indexes, caching, replication).
• Support: 1 month of consultation after delivery.

Our experience: 50+ projects in e-commerce and SaaS. After completion, you'll get a clear understanding of your site's capacity and be able to prevent crashes during peaks. Order load test development and be confident in your site's stability. Get a consultation: we'll test your site under load and give recommendations.

Timeline Estimate

Development of 3–5 turnkey scenarios takes from 2 to 5 days depending on complexity. Cost is calculated individually after analyzing your project.