1import pino from 'pino';
2import { AsyncLocalStorage } from 'async_hooks';
3 
4// Request context — automatically included in all logs
5const requestContext = new AsyncLocalStorage<{
6  requestId: string;
7  userId?: string;
8  traceId?: string;
9}>();
10 
11const logger = pino({
12  level: process.env.LOG_LEVEL ?? 'info',
13  formatters: {
14    log: (obj) => ({
15      ...obj,
AsyncLocalStorage injects request context without threading it through every function
16      ...requestContext.getStore(), // Auto-inject requestId, traceId
17    }),
18  },
19});
20 
21// Middleware: set context for all logs in this request
22app.use((req, res, next) => {
Propagate X-Request-ID — it links logs across services
23  const requestId = req.headers['x-request-id'] ?? crypto.randomUUID();
24  const traceId = req.headers['x-trace-id'] ?? requestId;
25  requestContext.run({ requestId, traceId }, () => {
26    res.setHeader('x-request-id', requestId);
27    next();
28  });
29});
30 
31// Every log automatically includes requestId, traceId
32logger.info({ operation: 'order.create', orderId: 42, duration: 45 }, 'Order created');
33// ✅ Structured — machine-queryable in Grafana Loki
34// ❌ Avoid: logger.info(`Order ${orderId} created`) — unqueryable free text

1# SLO alerting: burn rate approach
2 
3# Error rate alert (page the on-call)
4- alert: SLOErrorBudgetBurningFast
5  expr: |
6    (
7      sum(rate(http_requests_total{status=~"5.."}[1h])) /
8      sum(rate(http_requests_total[1h]))
9    ) > 0.01  # > 1% error rate for 1 hour
>1% for 5 minutes → page. Not every spike — sustained burn.
10  for: 5m
11  labels:
12    severity: page
13  annotations:
14    summary: "Error budget burning — SLO at risk"
15 
16# Burn rate: how fast is the monthly budget being consumed?
17# Burn rate of 1 = on track (sustainable)
18# Burn rate of 14.4 = budget exhausted in 50 hours (urgent)
Burn rate > 14.4 = monthly budget exhausted in 50 hours
19- record: slo:error_budget_burn_rate:1h
20  expr: |
21    (sum(rate(http_requests_total{status=~"5.."}[1h])) /
22     sum(rate(http_requests_total[1h])))
23    / 0.001  # divide by (1 - SLO target) = (1 - 0.999)

1import { NodeSDK } from '@opentelemetry/sdk-node';
2import { OTLPTraceExporter } from '@opentelemetry/exporter-trace-otlp-grpc';
3import { getNodeAutoInstrumentations } from '@opentelemetry/auto-instrumentations-node';
4 
5// Auto-instruments HTTP, Express, pg, Redis, etc.
Auto-instrumentation captures HTTP, DB, Redis spans with zero manual code
6const sdk = new NodeSDK({
7  resource: new Resource({
8    'service.name': 'orders-service',
9    'service.version': process.env.APP_VERSION,
10  }),
11  traceExporter: new OTLPTraceExporter({
12    url: process.env.OTEL_EXPORTER_OTLP_ENDPOINT,
13  }),
14  instrumentations: [getNodeAutoInstrumentations()],
15});
16sdk.start();
17 
18// Manual span for critical business operations
19const tracer = trace.getTracer('orders-service');
20 
21async function processPayment(orderId: string, amount: number) {
22  return tracer.startActiveSpan('payment.process', async (span) => {
23    span.setAttributes({ 'order.id': orderId, 'payment.amount': amount });
24    try {
25      const result = await paymentGateway.charge(amount);
26      span.setStatus({ code: SpanStatusCode.OK });
27      return result;
28    } catch (error) {
29      span.recordException(error as Error);
30      span.setStatus({ code: SpanStatusCode.ERROR });
31      throw error;
Always end spans in finally — unclosed spans cause memory leaks
32    } finally {
33      span.end(); // ALWAYS end spans — unclosed spans leak memory
34    }
35  });
36}