Add k6 smoke test triggered on every deployment #126

New Issue

marcel · 2026-03-27T18:36:47+01:00

marcel commented

2026-03-27 18:36:47 +01:00

Problem

There are no load or performance tests. After every deployment, there is no automated check that the application responds within acceptable latency or that the error rate is acceptable. A performance regression — a slow query introduced by a new index, a MinIO connection pool misconfiguration — will only be noticed when a user complains.

Why This Matters

This application handles file uploads to MinIO and full-text search queries against PostgreSQL. Both are I/O-bound operations with real latency characteristics. A smoke test with thresholds catches obvious breakage immediately after deploy without requiring a full load test run.

What Needs To Be Done

Create load-tests/smoke.js:

import http from 'k6/http';
import { check } from 'k6';

export const options = {
  vus: 5,
  duration: '1m',
  thresholds: {
    http_req_duration: ['p(95)<500'],  // 95th percentile under 500ms
    http_req_failed: ['rate<0.01'],    // less than 1% error rate
  },
};

export default function () {
  const res = http.get(`${__ENV.BASE_URL}/api/documents`);
  check(res, { 'status is 200': (r) => r.status === 200 });
}

Add a smoke-test CI job that runs after the E2E job on merges to main:

smoke-test:
  needs: e2e-tests
  # Runs k6 smoke test against the deployed instance
  # Fails if p95 > 500ms or error rate > 1%

Define thresholds as non-negotiable CI gates:
- p(95) < 500ms for all API endpoints
- Error rate < 1%
Separate nightly load test (ramp to expected traffic, 10 minutes) can be added later — the smoke test is the minimum viable gate.

Acceptance Criteria

load-tests/smoke.js committed to repo
Smoke test runs in CI on merge to main
CI fails if p95 > 500ms or error rate > 1%
k6 summary report uploaded as CI artifact

## Problem There are no load or performance tests. After every deployment, there is no automated check that the application responds within acceptable latency or that the error rate is acceptable. A performance regression — a slow query introduced by a new index, a MinIO connection pool misconfiguration — will only be noticed when a user complains. ## Why This Matters This application handles file uploads to MinIO and full-text search queries against PostgreSQL. Both are I/O-bound operations with real latency characteristics. A smoke test with thresholds catches obvious breakage immediately after deploy without requiring a full load test run. ## What Needs To Be Done 1. Create `load-tests/smoke.js`: ```javascript import http from 'k6/http'; import { check } from 'k6'; export const options = { vus: 5, duration: '1m', thresholds: { http_req_duration: ['p(95)<500'], // 95th percentile under 500ms http_req_failed: ['rate<0.01'], // less than 1% error rate }, }; export default function () { const res = http.get(`${__ENV.BASE_URL}/api/documents`); check(res, { 'status is 200': (r) => r.status === 200 }); } ``` 2. Add a `smoke-test` CI job that runs after the E2E job on merges to `main`: ```yaml smoke-test: needs: e2e-tests # Runs k6 smoke test against the deployed instance # Fails if p95 > 500ms or error rate > 1% ``` 3. Define thresholds as non-negotiable CI gates: - `p(95) < 500ms` for all API endpoints - Error rate `< 1%` 4. Separate nightly load test (ramp to expected traffic, 10 minutes) can be added later — the smoke test is the minimum viable gate. ## Acceptance Criteria - [ ] `load-tests/smoke.js` committed to repo - [ ] Smoke test runs in CI on merge to `main` - [ ] CI fails if p95 > 500ms or error rate > 1% - [ ] k6 summary report uploaded as CI artifact

marcel added the test descoped labels 2026-03-28 08:06:02 +01:00

marcel commented

2026-03-28 08:06:21 +01:00

Currently there is no deployment, so will keep that for later

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: marcel/familienarchiv#126