无法使用docker-compose运行具有副本集的mongo容器

Question

Here's my docker-compose file:

version: '3'
services:

  mongo:
    hostname: mongo
    container_name: search_mongo
    image: mongo:latest
    volumes:
      - ./docker/local/persist/mongo:/data/db
      - ./docker/mongo:/opt/mongo
    ports:
      - "8884:27017"
      - "8885:27018"
    entrypoint: /opt/mongo/entrypoint_mongo.sh

  agent:
    build: .
    image: myapp_search:compose
    depends_on:
      - mongo

Here's my entrypoint_mongo.sh

#!/bin/bash
mongod --port 27018 --replSet rs0 --fork --syslog --smallfiles
mongo --port 27018 --eval "rs.initiate({_id : 'rs0', members : [{_id : 0, host : 'mongo:27018'}]})"
mongo --port 27018 --eval "while(true) {if (rs.status().ok) break;sleep(1000)};"

The issue I am facing is : The mongo container is executing all its steps successfully but its exiting with status 0.

mongo_1          | about to fork child process, waiting until server is ready for connections.
mongo_1          | forked process: 7
mongo_1          | child process started successfully, parent exiting
mongo_1          | MongoDB shell version v3.4.10
mongo_1          | connecting to: mongodb://127.0.0.1:27018/
mongo_1          | MongoDB server version: 3.4.10
mongo_1          | { "ok" : 1 }
mongo_1          | MongoDB shell version v3.4.10
mongo_1          | connecting to: mongodb://127.0.0.1:27018/
mongo_1          | MongoDB server version: 3.4.10
search_mongo exited with code 0

Answer 1

If your last command at the script will never exit (ie stop running), your container keeps running. So, while (true);do sleep 1; done while (true);do sleep 1; done is one fix.

If you want to have solution where you can close container when wanted, you can use something like

while (! test -e /tmp/stop); do sleep 1; done; rm /tmp/stop

So, when you say at command line touch /tmp/stop , it will stop that loop what keeps container running.

Answer 2

Your startup script should not initialise or monitor the replicaset; those should be manual tasks.

You should bear in mind that:

• initiating a replica set is strictly a one-off job; once it is initiated, the MongoDB service, when restarted, will continue being part of the same replica set.
• a replica set normally contains several nodes which should be interchangable; if each of them tries to initialise the replica set on startup, they will throw errors
• restarting a service is normal, expected behaviour; for example when you upgrade to the next version of MongoDB, or after patches to your server host require a reboot, or after a power outage
• if your script tries to initialise an already-initialised replica set each time it starts the MongoDB service, it will throw errors

I strongly recommend that you make three changes:

1. Let your mongo container just run mongo, without the steps to initiate and monitor the replica set.
2. If you want to run a replica set, initiate it carefully and in a controlled manual way; ditto if you want to add / remove nodes, or reconfigure.
3. If you want to monitor the health of your replica set, use a separate tool to do that; let the mongo service just do its ordinary job.