可观测性：日志、指标与追踪

概述

可观测性是理解系统内部状态的能力。本教程介绍日志、指标和分布式追踪。

1. 日志

import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.slf4j.MDC;

@Component
public class LoggingService {
    private static final Logger logger = LoggerFactory.getLogger(LoggingService.class);
    
    public void processRequest(String requestId) {
        MDC.put("requestId", requestId);
        
        try {
            logger.info("开始处理请求");
            
            // 业务逻辑
            
            logger.info("请求处理完成");
        } catch (Exception e) {
            logger.error("请求处理失败", e);
        } finally {
            MDC.clear();
        }
    }
}

2. 指标

import io.micrometer.core.instrument.*;
import org.springframework.stereotype.Component;

@Component
public class MetricsService {
    private final Counter requestCounter;
    private final Timer requestTimer;
    private final Gauge activeUsers;
    
    public MetricsService(MeterRegistry registry) {
        this.requestCounter = Counter.builder("http.requests.total")
            .description("Total HTTP requests")
            .register(registry);
        
        this.requestTimer = Timer.builder("http.request.duration")
            .description("HTTP request duration")
            .register(registry);
        
        this.activeUsers = Gauge.builder("app.active.users", this, MetricsService::getActiveUserCount)
            .description("Active users count")
            .register(registry);
    }
    
    public void recordRequest() {
        requestCounter.increment();
    }
    
    public void recordDuration(long durationMs) {
        requestTimer.record(Duration.ofMillis(durationMs));
    }
    
    private double getActiveUserCount() {
        return activeUserService.getActiveCount();
    }
}

3. 分布式追踪

import brave.Tracing;
import brave.sampler.Sampler;
import zipkin2.reporter.AsyncReporter;
import zipkin2.reporter.okhttp3.OkHttpSender;

@Configuration
public class TracingConfig {
    @Bean
    public Tracing tracing() {
        OkHttpSender sender = OkHttpSender.create("http://localhost:9411/api/v2/spans");
        AsyncReporter reporter = AsyncReporter.builder(sender).build();
        
        return Tracing.newBuilder()
            .localServiceName("my-service")
            .spanReporter(reporter)
            .sampler(Sampler.ALWAYS_SAMPLE)
            .build();
    }
}

@Component
public class TracingService {
    private final Tracer tracer;
    
    public void processRequest() {
        Span span = tracer.newTrace().name("processRequest").start();
        
        try (Tracer.SpanInScope ws = tracer.withSpanInScope(span)) {
            // 业务逻辑
            span.annotate("processing");
        } catch (Exception e) {
            span.error(e);
        } finally {
            span.finish();
        }
    }
}

4. 实际应用示例

健康检查

@Component
public class HealthIndicator implements org.springframework.boot.actuate.health.HealthIndicator {
    @Override
    public Health health() {
        if (isDatabaseConnected() && isRedisConnected()) {
            return Health.up().withDetail("database", "connected").build();
        }
        return Health.down().withDetail("database", "disconnected").build();
    }
}

Prometheus指标

management:
  endpoints:
    web:
      exposure:
        include: health,metrics,prometheus
  metrics:
    export:
      prometheus:
        enabled: true

5. 最佳实践

1. 结构化日志：使用JSON格式记录日志
2. 统一指标命名：遵循Prometheus命名规范
3. 采样策略：合理设置追踪采样率
4. 上下文传播：在服务间传播追踪上下文
5. 监控告警：设置合理的告警阈值

总结

可观测性是理解系统内部状态的能力。掌握日志、指标和分布式追踪，可以构建可维护、可诊断的系统。