Graylog on Docker — Part 2: MongoDB

Now that we’ve setup our Elasticsearch cluster in part one, we can move to the second stage which is setting up a MongoDB replica set.

Configure MongoDB replica set

On our master machine (docker swarm), we will be add a new docker compose which will be used for the MongoDB deployment.

mongo.yaml

version: "3.8"

services:
  mongo-1:
    image: mongo
    volumes:
      - /data/mongo/mongo1:/data/db
    command: 'mongod --oplogSize 128 --replSet dbrs'
    networks:
      - my-overlay-2
    ports:
      - 27017:27017
    deploy:
      placement:
        constraints:
          - node.labels.type == master
      replicas: 1
      restart_policy:
        condition: on-failure
  mongo-2:
    image: mongo
    volumes:
      - /data/mongo/mongo2:/data/db
    command: 'mongod --oplogSize 128 --replSet dbrs'
    networks:
      - my-overlay-2
    ports:
      - 27018:27017
    deploy:
      placement:
        constraints:
          - node.labels.type == worker-1
      replicas: 1
      restart_policy:
        condition: on-failure
  mongo-3:
    image: mongo
    volumes:
      - /data/mongo/mongo3:/data/db
    command: 'mongod --oplogSize 128 --replSet dbrs'
    networks:
      - my-overlay-2
    ports:
      - 27019:27017
    deploy:
      placement:
        constraints:
          - node.labels.type == worker-2
      replicas: 1
      restart_policy:
        condition: on-failure
volumes:
  mongodb:
    driver: "local"
networks:
  my-overlay-2:
    external: true

Now let’s walk through it:

• We create 3 containers mongo-1,mongo-2 and mongo-3
• Each container is always placed on a different node using constraints. I’ve configured my master node with the master label, and each of the worker nodes with a worker label.
• MongoDB files are stored under the /data/mongo directory on each node. I’ve used NFS to mount this directory on a different server for larger storage but you can manually create these directories or each node, or change them entirely based on your needs.
• The containers use the same network my-overlay-2 as the rest of the containers described on Part 1. This is crucial since we need all the components of the Graylog cluster (Elasticsearch, MongoDB, Graylog) to talk to each other.
• The command mongod --oplogSize 128 --replSet dbrs tells our mongo instances that we want to initiate a replica set called dbrs. You can change this to your desired name.
• I’ve used different host ports for each container (27017, 27018, 27019) since initially I didn’t plan to place each container on a different node. If you’re planning to omit the constrains so the containers can be placed on any node, you should keep this port configuration. Otherwise, you can switch to 27017:27017 for all the containers.

Start the Stack

That’s pretty much it. Once you’ve made the changes to the compose file, save it somewhere on the master node and run it:

docker stack deploy -c mongo.yaml mongo

Initialize the replica set

Once your containers are up and running, you should check the logs to make sure everything is okay. You can do so by this command:

docker logs -f <container_name>

If everything is done correctly, you should see something like waiting to Initialize ... in the logs. If that’s the case, you need to run this command from a machine that has access to your cluster:

mongosh --host '172.17.93.171:27017' --eval 'rs.initiate({ _id: "dbrs", members: [{ _id: 0, host : "172.17.93.171:27017" }, { _id: 1, host : "172.17.93.172:27018" }, { _id: 2, host : "172.17.93.170:27019" }]})'

• You should have mongosh installed on the node that’s running this.
• Replace the IPs with your docker swarm node IPs.
• If you’ve changed the replica set name in the docker-compose file, make sure to change it from dbrs to your name.

Once this command is executed, you should see a message regarding the successful replica set initialization.

And that’s it! You now have a working MongoDB replica set on docker swarm! 🥳🥳🥳

In the next post, we’re going to finish up the cluster by deploying Graylog.