Persist docker volume on swarm

Question

So im building a swarm of Elasticsearch nodes and ideally i would like to see two thing happen.

1. Make each node save all its data on a folder on the host.
2. Even if the stack is destroyed ones a new container is initiallized it should be able to pick up where the one before was killed using the same volume.

This is what im doing:

docker volume create --opt type=none --opt device=/mnt/data --opt o=bind --name=elastic-data

docker-compose.yml

version: '3'
services:
  elastic-node1:
    image: amazon/opendistro-for-elasticsearch:0.8.0
    environment:
      - cluster.name=elastic-cluster
      - bootstrap.memory_lock=false
      - "ES_JAVA_OPTS=-Xms32g -Xmx32g"
      - opendistro_security.ssl.http.enabled=false
      - discovery.zen.minimum_master_nodes=1
    volumes:
      - elastic-data:/mnt/data
    ports:
      - 9200:9200
      - 9600:9600
      - 2212:2212  
    ulimits:
      memlock:
        soft: -1
        hard: -1
    networks:
      - elastic-net
    deploy:
      mode: replicated
      replicas: 1

volumes:
  elastic-data:
    external: true

And then i would start the stack, post some data, remove the stack and staring it again but the data is not being retained.

docker stack deploy --compose-file docker-compose.yml opendistrostack

Im a little bit confused about volumes and im not being able to find a good documentation with a detail explanation for each use case. Could you point me on the right direction?

Thanks.

Answer 1

As it is docker swarm support only local volume driver. You will have always fresh data whenever the container is created on a new host.

Common technique is to use a shared volume/fs. I'd suggest to implement GlusterFS, it is distribuite and high scalable fs, very easy to get started and well documented for swarm use cases.

Furthermore you can checkout some 3rd party volume drivers in docker store.