I've found a solution. It could help if someone is trying to do the same.

As, I said in my comment to @ibexit, I'm running two dockerized ES clusters and two separate Liferay portals (not in containers) on the same machine (development mode).

I changed th transport address in Liferay OSGi config file, since it must match the transport tcp port where ES is running:

transportAddresses="127.0.0.1:9301"
logExceptionsOnly="false"
operationMode="REMOTE"
indexNamePrefix="myprefix-"
clusterName="liferay-cluster"

I also added the property network.publish_host=127.0.0.1 in my ES clusters (without this property Liferay was not able to detect ES nodes)

Here is my docker-compose.yml:

Using ES 6.1.4

version: '2.2'
services:
  es01:
    image: docker.elastic.co/elasticsearch/elasticsearch:6.1.4
    container_name: es01
    environment:
      - node.name=es01
      - cluster.name=liferay-cluster
      - bootstrap.memory_lock=true
      - "ES_JAVA_OPTS=-Xms512m -Xmx512m"
      - transport.tcp.port=9301
      - network.publish_host=127.0.0.1
    ulimits:
      memlock:
        soft: -1
        hard: -1
    volumes:
      - esdata01:/usr/share/elasticsearch/data
    ports:
      - "9201:9200"
      - "9301:9301"
    networks:
      - esnet
  es02:
    image: docker.elastic.co/elasticsearch/elasticsearch:6.1.4
    container_name: es02
    environment:
      - node.name=es02
      - cluster.name=liferay-cluster2
      - bootstrap.memory_lock=true
      - "ES_JAVA_OPTS=-Xms512m -Xmx512m"
      - transport.tcp.port=9302
      - network.publish_host=127.0.0.1
    ulimits:
      memlock:
        soft: -1
        hard: -1
    ports:
      - "9202:9200"
      - "9302:9302"
    volumes:
      - esdata02:/usr/share/elasticsearch/data
    networks:
      - esnet

volumes:
  esdata01:
    driver: local
  esdata02:
    driver: local

networks:
  esnet:

network.publish_host did the trick !

Per default, elasticsearch binds the transport and http ports to localhost (local) only. So your ports exposed by docker are not working. You need to bind to a specific ip or using 0.0.0.0 for all or site as explained here: https://www.elastic.co/guide/en/elasticsearch/reference/current/modules-network.html#network-interface-values

Please have in mind, that enabling this will startup the node in production mode followed by several bootstrap checks. Please see the docs if you need more informations on this topics or search SO.

My working local setup with docker compose and two elasticsearch nodes in docker containers and Liferay running on host. I am using elasticsearch 6.8.2 images modified according to Liferay docs available from docker hub with url: https://hub.docker.com/repository/docker/ktorek/liferay7-elasticsearch.

I am using gradle workspace. So I've configured: configs/local/osgi/configs/com.liferay.portal.search.elasticsearch6.configuration.ElasticsearchConfiguration.config:

operationMode=REMOTE
clusterName=docker-cluster
transportAddresses=127.0.0.1:9300,127.0.0.1:9301

I've killed a lot of time with transportAddresses configuration as it's documented to use square brackets and square quotes transportAddresses=["192.168.1.1:9300","192.168.1.2:9300"] but it does not work. Configuration listed above contains actual working configuration syntax.

My docker-compose.yml:

version: '3.7'
services:

  es01:
    container_name: "es01"
    image: ktorek/liferay7-elasticsearch:latest
    environment:
      - node.name=es01
      - node.data=true
      - cluster.name=docker-cluster
      - xpack.security.enabled=false
      - bootstrap.memory_lock=true
      - "ES_JAVA_OPTS=-Xms512m -Xmx512m"
      - "discovery.zen.ping.unicast.hosts=es02"
    ports:
      - "9300:9300"
      - "9200:9200"
    networks:
      - mynetwork
    volumes:
      - es01-data:/usr/share/elasticsearch/data
    ulimits:
      memlock:
        soft: -1
        hard: -1

  es02:
    container_name: "es02"
    image: ktorek/liferay7-elasticsearch:latest
    environment:
      - node.name=es02
      - node.data=true
      - cluster.name=docker-cluster
      - xpack.security.enabled=false
      - bootstrap.memory_lock=true
      - "ES_JAVA_OPTS=-Xms512m -Xmx512m"
      - "discovery.zen.ping.unicast.hosts=es01"
    ports:
      - "9301:9300"
      - "9201:9200"
    networks:
      - mynetwork
    volumes:
      - es02-data:/usr/share/elasticsearch/data
    ulimits:
      memlock:
        soft: -1
        hard: -1
networks:
  mynetwork:
    name: mynetwork
    driver: bridge
    ipam:
      config:
        - subnet: 172.30.29.0/24