第七章：可观测性

本章介绍 API 网关的可观测性配置。

访问日志

Kong 日志配置

# kong.conf
proxy_access_log = /dev/stdout
proxy_error_log = /dev/stderr
admin_access_log = /dev/stdout
admin_error_log = /dev/stderr
log_level = notice

APISIX 日志配置

# HTTP 日志
curl -i -X PUT http://localhost:9180/apisix/admin/routes/1 \
  -H "X-API-KEY: edd1c9f034335f136f87ad84b625c8f1" \
  -d '{
    "uri": "/api/*",
    "plugins": {
      "http-logger": {
        "uri": "http://log-server/logs",
        "batch_max_size": 100,
        "max_retry_count": 3
      }
    },
    "upstream_id": "1"
  }'

# Kafka 日志
curl -i -X PUT http://localhost:9180/apisix/admin/routes/1 \
  -H "X-API-KEY: edd1c9f034335f136f87ad84b625c8f1" \
  -d '{
    "uri": "/api/*",
    "plugins": {
      "kafka-logger": {
        "broker_list": {
          "kafka:9092": 9092
        },
        "kafka_topic": "access-logs",
        "key": "remote_addr"
      }
    },
    "upstream_id": "1"
  }'

指标收集

Prometheus 集成

# Kong
curl -i -X POST http://localhost:8001/plugins \
  -d "name=prometheus"

# 访问指标
curl http://localhost:8001/metrics

# APISIX
# config.yaml
plugin_attr:
  prometheus:
    enable_exporter: true
    exporter_addr:
      ip: "0.0.0.0"
      port: 9091

# 访问指标
curl http://localhost:9091/apisix/prometheus/metrics

Prometheus 配置

# prometheus.yml
scrape_configs:
  - job_name: 'kong'
    static_configs:
      - targets: ['kong:8001']
  - job_name: 'apisix'
    static_configs:
      - targets: ['apisix:9091']

分布式追踪

Jaeger 集成

# APISIX
curl -i -X PUT http://localhost:9180/apisix/admin/routes/1 \
  -H "X-API-KEY: edd1c9f034335f136f87ad84b625c8f1" \
  -d '{
    "uri": "/api/*",
    "plugins": {
      "opentelemetry": {
        "sampler": {
          "name": "always_on"
        },
        "exporter": {
          "address": "otel-collector:4317"
        }
      }
    },
    "upstream_id": "1"
  }'

SkyWalking 集成

# APISIX
curl -i -X PUT http://localhost:9180/apisix/admin/routes/1 \
  -H "X-API-KEY: edd1c9f034335f136f87ad84b625c8f1" \
  -d '{
    "uri": "/api/*",
    "plugins": {
      "skywalking": {
        "sample_ratio": 1,
        "service_name": "APISIX-Gateway",
        "service_instance_name": "apisix-instance-1",
        "endpoint_addr": "skywalking-oap:11800"
      }
    },
    "upstream_id": "1"
  }'

健康检查

# Kong
curl http://localhost:8001/status

# APISIX
curl http://localhost:9180/apisix/status \
  -H "X-API-KEY: edd1c9f034335f136f87ad84b625c8f1"

Grafana 仪表板

导入仪表板

# Kong 仪表板 ID: 7424
# APISIX 仪表板 ID: 11719

关键指标

指标	说明
Request Rate	请求速率
Latency	响应延迟
Error Rate	错误率
Upstream Latency	上游延迟
Bandwidth	带宽使用

小结

可观测性要点：

• 访问日志：HTTP、Kafka
• 指标收集：Prometheus
• 分布式追踪：Jaeger、SkyWalking
• 健康检查：状态监控

下一章我们将学习生产实践。