树莓派5上Docker网络间歇性异常问题排查求助

环境信息 #

• 设备：树莓派5 8GB
• 运行环境：树莓派版Docker，所有服务均以容器化方式运行
• 涉及服务：Nginx Proxy Manager（NPM）、Jenkins、自定义服务my-service-container-name
• 网络配置：所有服务接入同一Docker外部网络main-net

问题现象 #

1. 初始部署完成后，NPM通过容器名my-service-container-name:4173反向代理该服务，工作正常。
2. 通过Jenkins更新并重新部署my-service-container-name容器后，NPM返回502错误。
3. 从NPM容器内ping该服务的IP172.20.0.4提示目标主机不可达，但ping同网络的Jenkins容器IP172.20.0.7通信正常。
4. my-service-container-name容器状态显示正常（docker ps无异常），且通过宿主机IP192.168.0.254:3500可正常访问该服务。

容器网络与状态信息 #

网络main-net详情 #

$ docker network inspect main-net
[
    {
        "Name": "main-net",
        "Id": "488966e42093e1abe082c19e83443955855952df79207c331c0a4f08e12b2bb7",
        "Created": "2024-01-24T12:13:31.125832336+09:00",
        "Scope": "local",
        "Driver": "bridge",
        "EnableIPv6": false,
        "IPAM": {
            "Driver": "default",
            "Options": null,
            "Config": [
                {
                    "Subnet": "172.20.0.0/16",
                    "Gateway": "172.20.0.1"
                }
            ]
        },
        "Internal": false,
        "Attachable": false,
        "Ingress": false,
        "ConfigFrom": {
            "Network": ""
        },
        "ConfigOnly": false,
        "Containers": {
            "2cc86409aed17c249b1d7724e2eca5a5010cf7f97e2cfb15d23ca2677114b085": {
                "Name": "my-service-container-name",
                "EndpointID": "8d781f8a59bdb5abd52a74712143c4bb7c9bf77392a59ea8ab40e5eb3b9cde54",
                "MacAddress": "02:42:ac:14:00:03",
                "IPv4Address": "172.20.0.4/16",
                "IPv6Address": ""
            },
            "a74e49c05c1482a42141d34d194a5e2879d207dff1ec631acd7a0570643bc4f3": {
                "Name": "nginx-proxy-manager",
                "EndpointID":       "30b9b6b80dfa890dcbae4415a655918015ad198e42169ef4a835adad782de732",
                "MacAddress": "02:42:ac:14:00:03",
                "IPv4Address": "172.20.0.5/16",
                "IPv6Address": ""
            },
            "ce25283e3f332aa925eb9ca35f1d4a7a5322b803a741334e045d4b35b4ea1a49": {
                "Name": "jenkins",
                "EndpointID": "df1da37a87db198f1e0fceeaac3488466118f4b68f6a8e9911d412532c5ac863",
                "MacAddress": "02:42:ac:14:00:02",
                "IPv4Address": "172.20.0.7/16",
                "IPv6Address": ""
            }
            ... More
        },
        "Options": {},
        "Labels": {
            "com.docker.compose.network": "main-net",
            "com.docker.compose.project": "host",
            "com.docker.compose.version": "2.24.1"
        }
    }
]

NPM容器ping服务IP结果 #

[root@docker-a74e49c05c14:/app]# ping 172.20.0.4
PING 172.20.0.4 (172.20.0.4) 56(84) bytes of data.
From 172.20.0.5 icmp_seq=10 Destination Host Unreachable
From 172.20.0.5 icmp_seq=11 Destination Host Unreachable
From 172.20.0.5 icmp_seq=12 Destination Host Unreachable

NPM容器ping JenkinsIP结果 #

PING 172.20.0.7 (172.20.0.7) 56(84) bytes of data.
64 bytes from 172.20.0.7: icmp_seq=1 ttl=64 time=0.171 ms
64 bytes from 172.20.0.7: icmp_seq=2 ttl=64 time=0.094 ms
64 bytes from 172.20.0.7: icmp_seq=3 ttl=64 time=0.096 ms

my-service-container-name容器状态 #

CONTAINER ID   IMAGE                             COMMAND                  CREATED        STATUS        PORTS                                                                                  NAMES
2cc86409aed1   my-service-container-name          "docker-entrypoint.s…"   25 hours ago   Up 13 hours   0.0.0.0:3500->4173/tcp, :::3500->4173/tcp                                              my-service-container-name

相关配置文件 #

vite.config.ts #

export default defineConfig({
    plugins: [react(), svgrPlugin(), tsconfigPaths()],
    server: {
        port: 3000,
        host: '0.0.0.0',
    },
    preview: {
        port: 4173,
        host: '0.0.0.0',
    },
})

docker-compose.yml #

version: '3.9'
services:
  front:
    build:
      context: .
      dockerfile: Dockerfile
    restart: unless-stopped
    ports:
      - '3500:4173'
    networks:
      - main-net
networks:
  main-net:
    external: true

Dockerfile #

FROM node:20-alpine
RUN apk add yarn
WORKDIR /hcc
COPY package.json .
COPY yarn.lock .
RUN yarn
COPY . .
RUN yarn build
EXPOSE 4173
CMD ["yarn", "preview"]

可能的原因与排查方向 #

1. Docker网络端点未正确更新
   容器重新部署后，Docker可能未在main-net中正确刷新容器的网络路由条目。即使docker network inspect显示容器已接入网络，实际网络栈中可能不存在该容器的路由。可尝试：

   • 重启NPM容器，触发网络路由刷新
   • 移除并重新添加my-service-container-name到main-net（docker network disconnect main-net my-service-container-name再docker network connect main-net my-service-container-name）

2. Docker DNS缓存问题
   若NPM使用容器名解析服务地址，Jenkins重建容器后IP变更，但NPM的DNS缓存未更新。可尝试：

   • 在NPM容器内手动解析容器名（nslookup my-service-container-name），查看是否指向正确IP
   • 重启NPM容器清除DNS缓存

3. main-net的Attachable属性限制
   当前网络Attachable: false，虽然初始部署正常，但容器重建时可能因该设置导致无法正确注册到网络。可执行命令修改属性后重新部署服务：

   docker network update --attachable true main-net
   

4. 树莓派Docker架构兼容性问题
   树莓派为ARM架构，Docker bridge驱动可能存在容器重建后跨容器通信的bug。可尝试：

   • 重启Docker服务（sudo systemctl restart docker）
   • 删除并重新创建main-net网络

5. 服务启动状态未就绪
   容器显示Up但内部服务可能未完全启动，导致NPM转发失败。可在docker-compose中添加健康检查：

   services:
     front:
       # ... 其他配置
       healthcheck:
         test: ["CMD", "wget", "--no-verbose", "--tries=1", "--spider", "http://localhost:4173"]
         interval: 10s
         timeout: 5s
         retries: 3
   

内容的提问来源于stack exchange，提问作者Jessie