Load tests for the backend reminder engine and analytics endpoints

Overview

• Two lightweight shell scripts are provided that use npx autocannon to drive load against the backend HTTP API.
• Scripts do not add dependencies to the repo; they rely on npx autocannon being available (installed transiently by npx).

Files

• run-reminder-loadtest.sh — public status check + admin-trigger (POST /api/reminders/process). Requires ADMIN_API_KEY env var.
• run-analytics-loadtest.sh — authenticated analytics endpoints (/api/analytics/summary and /api/analytics/spending). Requires X_API_KEY env var.

Representative scenarios

1. Light (smoke): DURATION=10 CONCURRENCY=10 ./run-analytics-loadtest.sh
2. Medium (normal load): DURATION=30 CONCURRENCY=50 ./run-analytics-loadtest.sh
3. Heavy (stress): DURATION=60 CONCURRENCY=200 ./run-analytics-loadtest.sh

Example commands Export required keys and run a medium analytics load test:

export X_API_KEY="your-test-api-key"
export TARGET="http://localhost:3001"
DURATION=30 CONCURRENCY=50 PIPES=10 ./scripts/load-tests/run-analytics-loadtest.sh

Run reminder engine tests (requires admin key):

export ADMIN_API_KEY="your-admin-key"
export TARGET="http://localhost:3001"
DURATION=30 CONCURRENCY=50 PIPES=10 ./scripts/load-tests/run-reminder-loadtest.sh

Measuring bottlenecks

• Fetch ops/monitoring metrics before and after a run (requires ADMIN_API_KEY):

curl -s -H "x-admin-api-key: $ADMIN_API_KEY" "$TARGET/api/admin/metrics/throughput?w=1" | jq .
curl -s -H "x-admin-api-key: $ADMIN_API_KEY" "$TARGET/api/admin/metrics/latency?w=1" | jq .
curl -s -H "x-admin-api-key: $ADMIN_API_KEY" "$TARGET/api/admin/metrics/retries?w=1" | jq .
curl -s -H "x-admin-api-key: $ADMIN_API_KEY" "$TARGET/api/admin/health" | jq .

• Use database query plans for slow queries: backend has scripts/capture-query-plans.js and scripts/benchmark-performance-indexes.js to capture plans and run benchmarks.
• Enable Sentry profiling or sampling during a controlled run to capture CPU/stack traces (Sentry is configured in backend/src/index.ts).

Interpreting results and next steps

• Look for increased 5xx responses, long tail latency, DB connection pool exhaustion, and retry queue growth.
• If DB is the bottleneck, enable or tune the indexes found in backend/migrations/20260527000000_add_performance_indexes.sql and re-run benchmarks.
• If CPU/GC is the bottleneck, consider scaling worker processes, optimizing heavy queries, or offloading analytics to materialized views.