Add CLUSTER_NETWORK_HOST environment variable to configure network.host

[jfut]

Hello,

This PR add CLUSTER_NETWORK_HOST environment variable to configure network.host for Elasticsearch Cluster.

Example: CLUSTER_NETWORK_HOST=_local_,192.168.1.111

If CLUSTER_NETWORK_HOST is not set, then 0.0.0.0.0 is used as before.

Problem:

If an Elasticsearch container can recognize multiple IP addresses, it may respond the local IP of the docker network and other Elasticseach containers(node) are not be able to join the cluster.

• Docker Host 1: elasticsearch(es01) + network_mode: host

  elasticsearch:
    image: wazuh/wazuh-elasticsearch:3.13.1_7.8.0
    hostname: elasticsearch
    extra_hosts:
      - "elasticsearch:192.168.1.111"
      - "es01:192.168.1.111"
      - "es02:192.168.1.112"
    restart: always
    ports:
      - "9200:9200"
      - "9300:9300"
    environment:
      - "ES_JAVA_OPTS=-Xms1g -Xmx1g"
      - ELASTIC_CLUSTER=true
      - CLUSTER_NETWORK_HOST=_local_,192.168.1.111
      - CLUSTER_NODE_MASTER=true
      - CLUSTER_MASTER_NODE_NAME=elasticsearch
    ulimits:
      memlock:
        soft: -1
        hard: -1
    mem_limit: 2g
    network_mode: host

• Docker Host 2: es02 + network_mode: host

  es02:
    image: wazuh/wazuh-elasticsearch:3.13.1_7.8.0
    hostname: es02
    extra_hosts:
      - "elasticsearch:192.168.1.111"
      - "es01:192.168.1.111"
      - "es02:192.168.1.112"
    restart: always
    ports:
      - "9200:9200"
      - "9300:9300"
    environment:
      - "ES_JAVA_OPTS=-Xms1g -Xmx1g"
      - ELASTIC_CLUSTER=true
      - CLUSTER_NETWORK_HOST=_local_,192.168.1.112
      - CLUSTER_MASTER_NODE_NAME=elasticsearch
      - CLUSTER_NODE_NAME=es02
    ulimits:
      memlock:
        soft: -1
        hard: -1
    mem_limit: 2g
    network_mode: host

In the following case, I expect elasticsearch(es01) to respond with 192.168.1.111, but it responds with 172.25.0.1. Therefore, es02 will fail to join the cluster.

Docker Host1 status:

# ip a | grep "inet "
    inet 127.0.0.1/8 scope host lo
    inet 192.168.1.111/16 brd 192.168.255.255 scope global noprefixroute br0
    inet 172.25.0.1/24 brd 172.25.0.255 scope global br-15fe5c2fd2dc
    inet 172.17.0.1/16 brd 172.17.255.255 scope global docker0

# curl http://localhost:9200/_cat/nodes?v
ip         heap.percent ram.percent cpu load_1m load_5m load_15m node.role master name
172.25.0.1           17          93  34    2.19    2.07     1.85 dilmrt    *      elasticsearch

es02 log:

followup connection failed.

es02_1  | {"type": "server", "timestamp": "2020-08-23T02:38:35,411Z", "level": "WARN", "component": "o.e.d.HandshakingTransportAddressConnector", "cluster.name": "elasticsearch", "node.name": "es02", "message": "[connectToRemoteMasterNode[192.168.1.111:9300]] completed handshake with [{elasticsearch}{5MsimsjLTBS2ms_AImfQ_A}{nfEq1iyyQHiOIuNX4HVenA}{172.25.0.1}{172.25.0.1:9300}{dilmrt}{ml.machine_memory=6442450944, ml.max_open_jobs=20, xpack.installed=true, transform.node=true}] but followup connection failed",

After applying this patch:

• Docker Host1: elasticsearch(es01) + network_mode: host + CLUSTER_NETWORK_HOST=_local_,192.168.1.111

# curl http://localhost:9200/_cat/nodes?v
ip            heap.percent ram.percent cpu load_1m load_5m load_15m node.role master name
192.168.1.111           15          94  18    1.92    2.06     1.83 dilmrt    *      elasticsearch

And es02 successfully joins the cluster.

# curl http://localhost:9200/_cat/nodes?v
ip            heap.percent ram.percent cpu load_1m load_5m load_15m node.role master name
192.168.1.112            8          93   0    0.19    0.12     0.07 dilrt     -      es02
192.168.1.111           16          94  18    2.04    1.66     1.47 dilmrt    *      elasticsearch

[xr09]

LGTM! Thanks for the contribution @jfut 👍