Docker swarm cluster and elasticsearch, using cons

Question

i was hoping someone here might be able to give me some input with a problem i'm having.

I have a Docker swarm cluster with 3 nodes and want to run the ELK stack but i am not sure how to store my data.

version: '3'
services:
  master01:
    image: elasticsearch:5.2.2
    ports:
      - 9200:9200
      - 9300:9300
    networks:
      - es
    volumes:
      - /es/data:/usr/share/elasticsearch/data
    command: >
      elasticsearch
      -E network.host=_eth0_
      -E node.master=true
      -E discovery.zen.ping.unicast.hosts=es_master01,es_master02,es_master03
      -E discovery.zen.minimum_master_nodes=3
      -E cluster.name=ElasticCluster
      -E node.name=es_master01
      -E transport.tcp.port=9300
      -E http.port=9200
      -E node.max_local_storage_nodes=3
    deploy:
      replicas: 1

  master02:
    image: elasticsearch:5.2.2
    ports:
      - 9201:9200
      - 9301:9300
    networks:
      - es
    volumes:
      - /es/data:/usr/share/elasticsearch/data
    command: >
      elasticsearch
      -E network.host=_eth0_
      -E node.master=true
      -E discovery.zen.ping.unicast.hosts=es_master01,es_master02,es_master03
      -E discovery.zen.minimum_master_nodes=3
      -E cluster.name=ElasticCluster
      -E node.name=es_master02
      -E transport.tcp.port=9300
      -E http.port=9200
      -E node.max_local_storage_nodes=3
    deploy:
      replicas: 1

  master03:
    image: elasticsearch:5.2.2
    ports:
      - 9202:9200
      - 9302:9300
    networks:
      - es
    volumes:
      - /es/data:/usr/share/elasticsearch/data
    command: >
      elasticsearch
      -E network.host=_eth0_
      -E node.master=true
      -E discovery.zen.ping.unicast.hosts=es_master01,es_master02,es_master03
      -E discovery.zen.minimum_master_nodes=3
      -E cluster.name=ElasticCluster
      -E node.name=es_master03
      -E transport.tcp.port=9300
      -E http.port=9200
      -E node.max_local_storage_nodes=3
    deploy:
      replicas: 1

  logstash:
    image: logstash:5.2.2
    ports:
      - 5000:5000
    networks:
      - es
    command: >
      logstash -e 'input { tcp { port => 5000 } } output { elasticsearch { hosts => "master01:9200" } }'
    deploy:
      replicas: 1

  kibana:
    image: kibana:5.2.2
    ports:
      - 5601:5601
    environment:
      SERVER_NAME: "kibana"
      SERVER_HOST: "0"
      ELASTICSEARCH_URL: "http://elastic:changeme@master01:9200"
      ELASTICSEARCH_USERNAME: "elastic"
      ELASTICSEARCH_PASSWORD: "changeme"
      XPACK_SECURITY_ENABLED: "true"
      XPACK_MONITORING_ENABLED: "true"
    networks:
      - es
    depends_on:
      - master01
    deploy:
      replicas: 1

networks:
  es:
    driver: overlay

It actually works apart from the fact that my master01,02,03 are created randomly and can be moved around randomly on the 3 nodes meaning that they will replicate their data to the new node when they can't find it after being remade on a new node. Over time this means my data exists x3.

I haven't been able to use constraints properly to bind the 3 elastic services to a node each, and i can't really seem to find anything that works when searching.

I've tried using environment: "constraint:node==node1" but it doesn't seem to have any effect at all when building using my compose file.

I've searched around and found some examples on how to do it with docker service create, but i can't seem to find a functioning syntax.

First time posting here, so if i did something wrong, be gentle please.

Answer 1

I may have found a solution myself.

http://embaby.com/blog/using-glusterfs-docker-swarm-cluster/

Gluster might be the perfect solution, i will post results tomorrow if it solves my problem.