Skip to content

README.md

Latest commit

 

History

History
66 lines (49 loc) · 3.18 KB

README.md

File metadata and controls

66 lines (49 loc) · 3.18 KB
Raw

Trivela load tests

k6 scripts that exercise the Trivela backend API under synthetic traffic. The scripts are designed to run against a locally deployed backend (http://localhost:3001 by default) but can target any environment via the BASE_URL env variable.

Scenarios

File Profile What it covers
scenarios/read-campaigns.js 100 VUs · 30s GET /api/v1/campaigns (read heavy)
scenarios/write-campaigns.js 10 VUs · 30s POST /api/v1/campaigns (writes)
scenarios/mixed-read-write.js 80R + 20W VUs · 60s Combined read/write traffic
scenarios/burst-registration.js 0→200 VUs · 70s User registration spike (ramping burst)
scenarios/claim-storm.js 0→150 VUs · 75s Concurrent reward claims (storm scenario)

Most scenarios apply the project pass/fail thresholds: http_req_duration{expected_response:true} p(95) < 200ms and http_req_failed rate < 0.01. Burst/storm scenarios use more lenient thresholds (500-800ms p95, 3-5% error rate) to account for traffic spikes. Set LATENCY_P95_MS and ERROR_RATE_THRESHOLD to override.

Prerequisites

  • Install k6 (Homebrew: brew install k6).
  • A running backend that listens on BASE_URL (defaults to http://localhost:3001).
  • For write scenarios, an API key with permission to call POST /api/v1/campaigns exposed as API_KEY.

Running locally

# 1. start the backend (writes need an API key)
TRIVELA_API_KEY=sk_dev_local npm run dev:backend

# 2. run the scenarios from the repo root
npm run load-test                                      # default: read-campaigns
LOAD_SCENARIO=write-campaigns API_KEY=sk_dev_local npm run load-test
LOAD_SCENARIO=mixed-read-write API_KEY=sk_dev_local npm run load-test
LOAD_SCENARIO=burst-registration npm run load-test    # registration spike
LOAD_SCENARIO=claim-storm API_KEY=sk_dev_local npm run load-test  # claim surge

npm run load-test is a thin wrapper around k6 run. To run a scenario directly:

BASE_URL=http://localhost:3001 \
API_KEY=sk_dev_local \
k6 run load-tests/scenarios/mixed-read-write.js

Interpreting results

After each run k6 prints a summary block. The two metrics that gate the suite:

  • http_req_duration{expected_response:true}: p(95) < 200ms — 95th percentile of successful responses. A regression here usually means SQLite write contention or rate limiter saturation.
  • http_req_failed: rate < 0.01 — proportion of non-2xx responses. Spikes typically indicate the rate limiter is rejecting traffic too aggressively (defaults are 60 req/min per IP).

When a threshold is exceeded k6 exits non-zero, which is what the CI workflow (.github/workflows/load-test.yml) uses to mark a manual run as red.

CI

The load suite is deliberately not part of PR CI. The workflow is workflow_dispatch-triggered so anyone with write access can run it on-demand from the Actions tab against a staging URL.