Monitoring

Ako monitorujeme zdravie ClubUp v produkcii.

Princípy

• Proactive — alerts nás budia keď je problém, nie užívateľ
• Actionable — každý alert má jasné next steps
• Low noise — radšej menej alertov ale relevantných
• Cost-conscious — pre 0–5000 študentov nepotrebujeme enterprise APM

Stack

Total monitoring cost: ~$60/mes.

Sentry

Setup

// apps/app/instrumentation.ts
import * as Sentry from '@sentry/nextjs';
 
export async function register() {
  Sentry.init({
    dsn: process.env.SENTRY_DSN,
    environment: process.env.VERCEL_ENV ?? 'development',  // production / preview / development
    tracesSampleRate: 0.1,    // 10% trace sampling
    profilesSampleRate: 0.1,
    
    beforeSend(event, hint) {
      // Strip sensitive data
      if (event.request?.cookies) {
        event.request.cookies = '[REDACTED]';
      }
      // Don't send auth errors (expected)
      if (hint.originalException?.code === 'unauthorized') return null;
      return event;
    },
  });
}

Čo trackujeme

• Unhandled exceptions v Server Components, Route Handlers, Server Actions, client components
• Manual capture pri kritických chybách:

import * as Sentry from '@sentry/nextjs';
 
try {
  await issueFinalCertificate(enrollmentId);
} catch (e) {
  Sentry.captureException(e, {
    tags: { feature: 'certificate', priority: 'high' },
    contexts: { enrollment: { enrollmentId } },
  });
  throw e;
}

Alerts

Vercel Analytics + Speed Insights

Auto-enabled v apps/app a apps/admin. Sleduje:

• Core Web Vitals — LCP, FID, CLS, INP
• Page views — top stránky, geo distribution
• Performance per route — server response times

Threshold alerts (Vercel Speed Insights):

• LCP > 2.5s (75th percentile) — varovanie
• CLS > 0.1 (75th percentile) — varovanie

UptimeRobot

External monitoring (mimo nášho infrastruktúry):

Health endpoint vráti 200 ak:

// apps/app/app/api/health/route.ts
export async function GET() {
  try {
    await db.collection('courses').findOne({}, { projection: { _id: 1 } });  // DB ping
    return Response.json({ status: 'ok', timestamp: new Date().toISOString() });
  } catch (e) {
    return Response.json({ status: 'degraded', error: 'database' }, { status: 503 });
  }
}

MongoDB Atlas Charts + Performance Advisor

Monitorujeme:

• Slow queries — Performance Advisor odporučí indexy
• Connection pool usage — pri 80%+ scale-up alebo optimalizácia connection reuse
• Disk usage — alert pri 80% disku
• Replica lag — alert pri > 10s lag medzi primary a secondary

Custom dashboardy v Atlas Charts:

• Active enrollments per Course
• Revenue per týždeň
• Course completion rate
• Test pass rate per Module
• Top failing test questions (kandidáti na revíziu)

Logy

Vercel logs sú default — držia sa 24h v Vercel UI. Pre dlhšie retention:

Drain do Better Stack

# Vercel project settings → Log Drains → Better Stack
endpoint: https://in.logs.betterstack.com
format: ndjson

Better Stack retention: 30 dní (free tier), 180 dní (paid).

Štruktúra logu

Server logy sú JSON s týmito polami:

{
  timestamp: '2026-09-15T14:23:00Z',
  level: 'info' | 'warn' | 'error',
  message: 'Order paid',
  service: 'clubup-app',
  requestId: 'req_abc',
  userId: 'sportup_person_id_X',
  orderId: 'order_id_Y',
  // ...context
}

Logger

// packages/logger/index.ts
export const logger = {
  info: (msg: string, ctx: object) => console.log(JSON.stringify({ level: 'info', message: msg, ...ctx })),
  warn: (msg: string, ctx: object) => console.warn(JSON.stringify({ level: 'warn', message: msg, ...ctx })),
  error: (msg: string, ctx: object) => console.error(JSON.stringify({ level: 'error', message: msg, ...ctx })),
};

Žiadne PII v logoch — emails, mená, IBAN nelogujeme. Iba IDs.

Status page

status.clubup.sk (cez Better Stack subdomain) — verejná status stránka:

• Marketing web
• Študentská app
• Admin app
• Platobná brána (24-pay)
• Video streaming (Mux)

Pri incidents (manuálne updaty + auto detection cez UptimeRobot).

Business metrics

V admin dashboarde zobrazujeme business KPIs (nie technical):

• MRR/ARR (Monthly/Annual Recurring Revenue) — napriek tomu že nemáme subscriptions, useful pre tracking
• Enrollment rate — koľko nákupov / koľko návštev katalógu
• Cart abandonment — koľko vytvorených objednávok zlyhá pri platbe
• Course completion rate — % zapísaných, ktorí dokončia kurz
• Module dropoff — kde študenti prestanú pokračovať
• NPS — Net Promoter Score (po dokončení kurzu)

Implementované cez Mongo aggregations + serverless API endpoint pre admin dashboard.

Incident response

Severity levels

Runbook

1. Detekcia — alert (Sentry, UptimeRobot) alebo user report
2. Acknowledge — odpovedz v Slack, vytvor incident v Better Stack
3. Investigate — Sentry + logs + Vercel deployments
4. Mitigate — rollback, hotfix, scale up
5. Communicate — status page update, ak SEV-1/SEV-2 aj email userom
6. Resolve — overiť, že všetko funguje
7. Postmortem — pre SEV-1/SEV-2 napísať blameless postmortem do internej dokumentácie

Postmortem template

# Postmortem: <názov incidentu>
 
**Date:** YYYY-MM-DD
**Severity:** SEV-?
**Duration:** XX min
**Author:** <meno>
 
## Summary
1–2 vety čo sa stalo a kto bol ovplyvnený.
 
## Timeline
- HH:MM — first alert
- HH:MM — investigation started
- HH:MM — root cause identified
- HH:MM — mitigation applied
- HH:MM — resolved
 
## Root cause
Technický popis.
 
## Impact
Koľko užívateľov, koľko peňazí, aký data loss.
 
## What went well
...
 
## What didn't go well
...
 
## Action items
- [ ] Akcia (owner, deadline)

Cost optimization

Mesačné náklady na monitoring (orientačne):

Pri prekročení 5000 študentov vyhodnotíme prechod na enterprise APM (Datadog, New Relic) — typicky $200–500/mes pre náš objem, ale bohatšie features.