observability（可观测性）

Observability

Three Pillars

1. Logs

Discrete events with context.

{
  "timestamp": "2024-01-01T12:00:00Z",
  "level": "error",
  "message": "Failed to process order",
  "orderId": "123",
  "error": "Payment declined",
  "traceId": "abc123"
}

2. Metrics

Numeric measurements over time.

http_requests_total{method="GET", status="200"} 1234
http_request_duration_seconds{quantile="0.95"} 0.23

3. Traces

Request flow through services.

Trace: abc123
├── API Gateway (50ms)
│   ├── Auth Service (10ms)
│   └── Order Service (35ms)
│       └── Database (20ms)

OpenTelemetry Setup

import { NodeSDK } from '@opentelemetry/sdk-node';
import { OTLPTraceExporter } from '@opentelemetry/exporter-trace-otlp-http';

const sdk = new NodeSDK({
  traceExporter: new OTLPTraceExporter({
    url: 'http://collector:4318/v1/traces',
  }),
  serviceName: 'my-service',
});

sdk.start();

Key Metrics

RED Method (Request-focused)

• Rate: Requests per second
• Errors: Failed requests per second
• Duration: Request latency

USE Method (Resource-focused)

• Utilization: % time busy
• Saturation: Queue depth
• Errors: Error count

Alerting

Good Alerts

• Actionable: Something can be done
• Urgent: Needs immediate attention
• Specific: Clear what's wrong

Alert Template

alert: HighErrorRate
expr: rate(http_requests_total{status=~"5.."}[5m]) > 0.01
for: 5m
labels:
  severity: critical
annotations:
  summary: "High error rate on {{ $labels.service }}"
  description: "Error rate is {{ $value | humanizePercentage }}"

Dashboards

Essential panels:

1. Request rate
2. Error rate
3. Latency (P50, P95, P99)
4. Saturation (CPU, memory)
5. Active alerts