多区域Cassandra集群跨数据中心节点无法互联求助

在两个拥有公网IP（分别为111.111.111.x和222.222.222.x）的主机上，通过Docker容器部署Cassandra集群，每个数据中心（DC）部署3个节点，但跨DC的节点无法互相发现。同一主机部署多DC时可正常互联，nc测试显示对方主机7000端口可连通，Cassandra日志无报错。

DC1（111.111.111.x主机）docker-compose配置 #

version: '3.3'

services:
  cass1:
    image: cassandra
    container_name: cass1
    hostname: cass1
    healthcheck:
        test: ["CMD", "cqlsh", "-e", "describe keyspaces" ]
        interval: 5s
        timeout: 5s
        retries: 60
    networks:
      - cassandra
    ports:
      - "9042:9042"
      - "7000:7000"
    environment: &environment
        HEAP_NEWSIZE: 128M
        MAX_HEAP_SIZE: 2048M
        CASSANDRA_SEEDS: "cass1, cass2, 222.222.222.222"
        CASSANDRA_CLUSTER_NAME: ks
        CASSANDRA_DC: DC1
        CASSANDRA_RACK: West
        CASSANDRA_ENDPOINT_SNITCH: GossipingPropertyFileSnitch
        CASSANDRA_NUM_TOKENS: 256

  cass2:
    image: cassandra
    container_name: cass2
    hostname: cass2
    healthcheck:
        test: ["CMD", "cqlsh", "-e", "describe keyspaces" ]
        interval: 5s
        timeout: 5s
        retries: 60
    networks:
      - cassandra
    ports:
      - "9043:9042"
      - "7001:7000"
    environment: *environment
    depends_on:
      - cass1

  cass3:
    image: cassandra
    container_name: cass3
    hostname: cass3
    healthcheck:
        test: ["CMD", "cqlsh", "-e", "describe keyspaces" ]
        interval: 5s
        timeout: 5s
        retries: 60
    networks:
      - cassandra
    ports:
      - "9044:9042"
      - "7002:7000"
    environment: *environment
    depends_on:
      - cass2

networks:
  cassandra:

DC2（222.222.222.x主机）docker-compose配置 #

version: '3.3'

services:
  cass1:
    image: cassandra
    container_name: cass1
    hostname: cass1
    healthcheck:
        test: ["CMD", "cqlsh", "-e", "describe keyspaces" ]
        interval: 5s
        timeout: 5s
        retries: 60
    networks:
      - cassandra
    ports:
      - "9042:9042"
      - "7000:7000"
    environment: &environment
        HEAP_NEWSIZE: 128M
        MAX_HEAP_SIZE: 2048M
        CASSANDRA_SEEDS: "cass1, cass2, 111.111.111.111"
        CASSANDRA_CLUSTER_NAME: ks
        CASSANDRA_DC: DC2
        CASSANDRA_RACK: East
        CASSANDRA_ENDPOINT_SNITCH: GossipingPropertyFileSnitch
        CASSANDRA_NUM_TOKENS: 256

  cass2:
    image: cassandra
    container_name: cass2
    hostname: cass2
    healthcheck:
        test: ["CMD", "cqlsh", "-e", "describe keyspaces" ]
        interval: 5s
        timeout: 5s
        retries: 60
    networks:
      - cassandra
    ports:
      - "9043:9042"
      - "7001:7000"
    environment: *environment
    depends_on:
      - cass1

  cass3:
    image: cassandra
    container_name: cass3
    hostname: cass3
    healthcheck:
        test: ["CMD", "cqlsh", "-e", "describe keyspaces" ]
        interval: 5s
        timeout: 5s
        retries: 60
    networks:
      - cassandra
    ports:
      - "9044:9042"
      - "7002:7000"
    environment: *environment
    depends_on:
      - cass2

networks:
  cassandra:

DC1节点nodetool status输出 #

Datacenter: DC1
===============
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
--  Address     Load        Tokens  Owns (effective)  Host ID                               Rack
UN  172.18.0.4  108.81 KiB  256     60.1%             3ceecd38-0c30-4189-9a52-cb23465ca0ca  West
UN  172.18.0.3  113.77 KiB  256     70.4%             f1f6f63a-7cc5-4883-8b90-5159c6b6613c  West
UN  172.18.0.2  108.74 KiB  256     69.5%             94395942-8efa-447b-9961-2d80375bce35  West

nc端口连通性测试结果 #

nc -v 222.222.222.222 7000
Connection to 222.222.222.222 7000 port [tcp/*]
succeeded!

排查与修复建议 #

• 修正节点地址宣告配置：当前Cassandra节点向集群广播的是Docker内部IP（如172.18.0.x），跨DC节点无法访问该私有IP。需在每个节点的环境变量中添加：

  CASSANDRA_LISTEN_ADDRESS: 0.0.0.0  # 让Cassandra监听所有容器内部地址
  CASSANDRA_BROADCAST_ADDRESS: 111.111.111.x  # DC1节点填所在主机公网IP，DC2填222.222.222.x
  

  该配置让节点对外宣告可被跨DC访问的公网IP，而非内部私有IP。

• 调整种子节点列表：当前种子包含本地容器hostname（cass1、cass2），跨DC环境下无法解析这些本地名称。种子节点应仅保留公网IP的种子节点，例如：

  CASSANDRA_SEEDS: "111.111.111.x,222.222.222.x"
  

  确保两个DC的种子列表一致，且仅包含可跨网络访问的公网IP。

• 验证端口映射完整性：除7000端口（Gossip通信）外，需确保9042（CQL）、7001（SSL Gossip，若启用）端口正确映射。但当前nc测试已确认7000连通，主要问题仍集中在地址宣告和种子配置。

修改后重启所有节点，再次执行nodetool status即可查看跨DC节点是否加入集群。

内容的提问来源于stack exchange，提问作者stroiker