docker swarm和命名卷出现问题

Question

I'm having some issues using named volumes in one of my docker stacks.

Basically I'm running a bitcoin node developed by bcoin. What I'm trying to achieve is to have the containers running the nodes to write and read the blockchain data from a Digital Ocean volume attached to one of my droplets (VMs). I'm guessing I'm having some issues because I normally use volumes one way only, having stuff in my host machine being present in my containers.

This DO volume is present in the directory /mnt/blockchain-data in my manager node (or droplet).

So here's my docker-stack.yml file:

version: "3.6"
services:
  bcoin:
    image: hollarves/bcoin-mueve
    ports:
      #-- Mainnet
      # - "8333:8333"
      # - "8332:8332" # RPC/HTTP
      # - "8334:8334" # Wallet
      #-- Testnet
      - "18333:18333"
      - "18332:18332" # RPC/HTTP
      - "18334:18334"
    networks:
      - bitcoin-private-net
    volumes:
      - blockchain:/data
      - ${PWD}/bitcoin/secrets/bcoin.conf:/data/bcoin.conf
      - ${PWD}/bitcoin/secrets/wallet.conf:/data/wallet.conf
    environment:
      BCOIN_CONFIG: /data/bcoin.conf
    deploy:
      labels:
        - "traefik.docker.network=bitcoin-net"
        - "traefik.bcoin.frontend.rule=Host:bcoin.xxxx.com"
        - "traefik.bcoin.backend=bcoin"
        - "traefik.bcoin.port=18332"
        - "traefik.bcoin.frontend.entryPoints=http"
        - "traefik.bcoin-wallet.frontend.rule=Host:bcoin-wallet.xxxx.com"
        - "traefik.bcoin-wallet.backend=bcoin-wallet"
        - "traefik.bcoin-wallet.port=18334"
        - "traefik.bcoin-wallet.frontend.entryPoints=http"
      placement:
        constraints:
          - node.role == manager

volumes:
  blockchain:
    driver: local
    driver_opts:
      type: bind
      device: :/mnt/blockchain-data

networks:
  bitcoin-private-net:
    external: true

So basically I want whatever is in the /data directory of my containers to be synced with whatever it is in my /mnt/blockchain-data in my host machine. If the container writes some new blocks they should be copied to the /mnt/blockchain-data dir, and if a new bitcoin node is replicated, it should already have it's data directory synced with the files in my host machine.

The main idea is to avoid redownloading the blockchain whenever a node is destroyed and recreated for whatever reason. As well as being able to replicate as many nodes I want without having to download the blockchain for each one of them.

Any ideas on what I might be doing wrong? Thanks!

Answer 1

There are various ways to achieve your goal, especially in a swarm environnement. At first, let's say that docker does not sync volumes among swarm, unless you use a suited driver. Thus, you'll have to use drivers or other solutions, according to whether you need r/w or only read!

Here are a few ideas (not all solutions!) :

• You may mount share drive among the machines hosting your nodes, by exemple NFS . Ensure they are mounted on the same path on all hosts. Then, just bind mound the path to you container /data. Though, this is not the easiest way. But if you need R/W options, this is a suited solution. However, note that this solution can be seen as introducing a SPOOF , which is not that great in production environnement.
• you may use named volume as well, but in a distributed environment, you will have to use a volume driver to be able to sync them. Docker doc has a few exemples . You have a list of volume plugings here . According to which one you choose, you may eliminate the SPOOF problem.
• Eventually, if you don't need r/w possibilities, swarm mode allow you to use secrets / configs . Especially, secrets are obviously suited to such sensitive data as yours.

Hope this helps!