Observability — tizimda nima bo‘layotganini ko‘rish

Monolitda debug oson: bitta jarayon, bitta log, kirib grep qilasan. Lekin biz qurgan taqsimlangan tizimda (mikroservis — 1-mavzu) bitta so‘rov 5–6 xizmatdan o‘tadi.

Observability uch "ustun"ga tayanadi — har biri boshqa savolga javob beradi:

• Logs (loglar): alohida hodisalar — "aynan shu lahzada nima bo‘ldi" (xato, muhim voqea). Diskret, batafsil.
• Metrics (metrikalar): vaqt bo‘yicha yig‘ilgan raqamlar — "qancha so‘rov, qancha xato, qanday kechikish". Arzon, dashboard va alert uchun.
• Traces (izlar): bitta so‘rovning boshidan oxirigacha yo‘li — "vaqt qayerda ketdi, qaysi xizmat muammoli". Taqsimlangan tizim uchun hal qiluvchi.

Har ustunning o‘z o‘rni bor — ularni to‘g‘ri ishlatish muhim:

Strukturalangan log — eng muhim odat

NestJS’da uch ustunni amalga oshiramiz: strukturalangan log, Prometheus metrika va (keyingi bo‘limda) tracing.

1 · Correlation ID — hammani bog‘lovchi ip

// Correlation ID: har so'rovga noyob ID beriladi va servislar bo'ylab uzatiladi
@Injectable()
export class CorrelationMiddleware implements NestMiddleware {
  use(req: Request, res: Response, next: NextFunction) {
    const id = (req.headers["x-correlation-id"] as string) ?? randomUUID();
    req["correlationId"] = id;
    res.setHeader("x-correlation-id", id);      // javobda ham qaytar
    next();
  }
}

2 · Strukturalangan log (nestjs-pino)

// Strukturalangan (JSON) log — nestjs-pino. correlationId bilan bog'lanadi
this.logger.info(
  {
    correlationId: req.correlationId,
    userId: user.id,
    action: "order.created",
    orderId: order.id,
    durationMs: 42,
  },
  "Order created",
);
// -> Loki/ELK'da correlationId bo'yicha butun so'rovning barcha loglarini topasan

3 · Prometheus metrika

// Metrika: har so'rov davomiyligini o'lchaymiz (prom-client)
const httpDuration = new Histogram({
  name: "http_request_duration_seconds",
  help: "HTTP request duration",
  labelNames: ["method", "route", "status"],   // kam kardinallik!
});

@Injectable()
export class MetricsInterceptor implements NestInterceptor {
  intercept(ctx: ExecutionContext, next: CallHandler) {
    const end = httpDuration.startTimer();
    const req = ctx.switchToHttp().getRequest();
    return next.handle().pipe(
      tap(() => end({ method: req.method, route: req.route?.path, status: 200 })),
    );
  }
}
// GET /metrics -> Prometheus scrape qiladi -> Grafana dashboard

Qanday ishlaydi

Gateway (8-mavzu) so‘rovga trace ID beradi. Bu ID har xizmatga header orqali (W3C traceparent) uzatiladi. Har xizmat o‘z ishini span sifatida qo‘shadi. Natijada — so‘rovning to‘liq "sharshara" (waterfall) ko‘rinishi, har bosqichning vaqti bilan:

// tracing.ts — ilova boshida ishga tushadi (OpenTelemetry)
const sdk = new NodeSDK({
  traceExporter: new OTLPTraceExporter({ url: "http://jaeger:4318/v1/traces" }),
  instrumentations: [getNodeAutoInstrumentations()],   // HTTP, Prisma, Redis avtomatik
});
sdk.start();
// endi har so'rov trace ID oladi va span'lar servislar bo'ylab avtomatik tarqaladi

RED va USE usullari

Four golden signals (Google SRE): latency (kechikish), traffic (yuk), errors (xatolar), saturation (to‘lganlik). Bu to‘rttasini kuzatsang — tizim sog‘lig‘ining 90%ini ko‘rasan.

SLI, SLO, SLA

• SLI (indicator): o‘lchanadigan ko‘rsatkich — "so‘rovlarning 99%i 200ms’dan tez".
• SLO (objective): ichki maqsad — "99.9% so‘rov muvaffaqiyatli bo‘lsin".
• SLA (agreement): mijoz bilan rasmiy kelishuv (buzilsa — jarima/kompensatsiya).

Observability — incidentdan keyin emas, birinchi kundan qo‘yiladi (muammo chiqganda tayyor bo‘lishi kerak).

• Observability = tizim chiqishlaridan uning ichki holatini tushunish. Taqsimlangan tizimda hayotiy (so‘rov ko‘p xizmatdan o‘tadi). Monitoring (known) ≠ observability (unknown savollar).
• Uch ustun: Logs (nima bo‘ldi — hodisalar), Metrics (qancha — yig‘ilgan sonlar), Traces (vaqt qayerda ketdi — so‘rov yo‘li).
• Log strukturalangan (JSON) bo‘lsin (qidiriladi); metrika kam kardinalli (kardinallikka ehtiyot); trace bitta so‘rovni servislar bo‘ylab ko‘rsatadi.
• Distributed tracing — trace ID gateway’da beriladi, header (W3C traceparent) orqali tarqaladi. OpenTelemetry avtomatik qiladi. Correlation ID uch ustunni bog‘laydi.
• Metrikani tizimli yig‘: RED (xizmat), USE (resurs), golden signals. SLI/SLO belgila; alert’ni alomatga (sababga emas) qo‘y.