Wsrtjtyk

Came up with a solution that suits my needs, leaving it here in case anybody else needs it.

Summary

The idea is to take a standard rabbitmq container with management plugin enabled and use it to create the required configuration, then export and use it to start new containers. The below solution creates a derived docker image but it also works to just mount the two files at runtime (e.g. using docker compose).

References

• the info I started from


• complete rabbitmq.config example

Components

• official rabbitmq image, management plugin version (rabbitmq:management)


• 
  

  custom image based on the original one, with this Dockerfile (using version 3.6.6):

  
  
  

  FROM rabbitmq:3.6.6-management
  
  ADD rabbitmq.config /etc/rabbitmq/
  
  ADD definitions.json /etc/rabbitmq/
  
  RUN chown rabbitmq:rabbitmq /etc/rabbitmq/rabbitmq.config /etc/rabbitmq/definitions.json
  
  CMD ["rabbitmq-server"]
  
  

  
  


• 
  rabbitmq.config just tells rabbitmq to load definitions from the json file


• 
  definitions.json contains the users, vhosts, etc. and can be generated by the export function of the management web interface

rabbitmq.config example:

[

  {rabbit, [

    {loopback_users, }

  ]},

  {rabbitmq_management, [

    {load_definitions, "/etc/rabbitmq/definitions.json"}

  ]}

].

definitions.json example:

{

 "rabbit_version": "3.6.6",

 "users": [

  {

   "name": "user1",

   "password_hash": "pass1",

   "hashing_algorithm": "rabbit_password_hashing_sha256",

   "tags": ""

  },

  {

   "name": "adminuser",

   "password_hash": "adminpass",

   "hashing_algorithm": "rabbit_password_hashing_sha256",

   "tags": "administrator"

  }

 ],

 "vhosts": [

  {

   "name": "/vhost1"

  },

  {

   "name": "/vhost2"

  }

 ],

 "permissions": [

  {

   "user": "user1",

   "vhost": "/vhost1",

   "configure": ".*",

   "write": ".*",

   "read": ".*"

  }

 ],

 "parameters": ,

 "policies": ,

 "queues": ,

 "exchanges": ,

 "bindings": 

}

Alternave version

Deriving a new docker image is just one solution and works best when portability is key, since it avoids including host-based file management in the picture.

In some situations using the official image and providing configuration files from storage local to the host might be preferred.

The rabbitmq.config and definitions.json files are produced the same way, then mounted at runtime.

Notes:

• I'm assuming they have been placed in /etc/so/ for the sake of these examples


• files need to either be world readable or owned by the rabbitmq user or group (numerical id inside the docker container is 999), this needs to be handled by the host's sysadmin

docker run example:

    docker run --rm -it 

        -v /etc/so/rabbitmq.config:/etc/rabbitmq/rabbitmq.config:ro 

        -v /etc/so/definitions.json:/etc/rabbitmq/definitions.json:ro 

        rabbitmq:3.6-management

docker compose example:

    version: '2.1'

    services:

        rabbitmq:

            image: "rabbitmq:3.6-management"

            ports:

                - 5672:5672

                - 15672:15672

            volumes:

                - /etc/so/rabbitmq.config:/etc/rabbitmq/rabbitmq.config:ro

                - /etc/so/definitions.json:/etc/rabbitmq/definitions.json:ro

I would like to add that sudo's response helped me a lot. But that it still missed a command to be added to the Dockerfile.

The rabbitmq.config and definitions.json file should be owned by the rabbitmq user & group. So after adding the files run chown.

The full Dockerfile in my case was the following:

FROM rabbitmq:3-management-alpine

ADD definitions.json /etc/rabbitmq/

ADD rabbitmq.config /etc/rabbitmq/

RUN chown rabbitmq:rabbitmq /etc/rabbitmq/rabbitmq.config /etc/rabbitmq/definitions.json



EXPOSE 4369 5671 5672 15671 15672 25672



CMD ["rabbitmq-server"]

The rabbitmq.config file has the following content being a merge from the default image's config and the added definitions loading:

[

    { rabbit, [

        {loopback_users, },

        { tcp_listeners, [ 5672 ]},

        { ssl_listeners, [ ]},

        { hipe_compile, false } 

    ]},

    { rabbitmq_management, [

        { load_definitions, "/etc/rabbitmq/definitions.json"},

        { listeners, [

            { port, 15672 },

            { ssl, false } 



        ]}

    ]}

].

The definitions file can be exported from the management interface in the overview tab.

So you would first create a normal 'empty' rabbitmq container. Define whatever users, exchanges and queues you like. Then enter the management interface, export the definitions and create your own image using the file as described above.

Downloading the definitions is the easiest way to get the right password hashes in the definitions file for your own passwords. If you do not wish to do that you should follow the instructions as noted here (https://www.rabbitmq.com/passwords.html) to generate the correct hashes.

The newest version of the RabbitMQ image on Dockerhub has in-built functionality for changing the default username / password from "guest" / "guest" to something else.

Simply set the environment variables "RABBITMQ_DEFAULT_USER" and "RABBITMQ_DEFAULT_PASS" when starting the image.

As a docker command, you would run the image like this:

docker run 

-e RABBITMQ_DEFAULT_USER=test-user 

-e RABBITMQ_DEFAULT_PASS=test-user 

-p 5672:5672 

rabbitmq

In my case sleep 5 solution above did not work because RabbitMQ startup time was much longer and not predictable. Posting solution which waits until RabbitMQ is up and running:

• 
  

  Dockerfile

  
  
  

  FROM rabbitmq:3-management
  
  ADD init.sh /
  
  ADD config_rabbit.sh /
  
  RUN chmod +x /init.sh /config_rabbit.sh
  
  ENTRYPOINT ["/init.sh"]
  
  

  
  


• 
  

  init.sh

  
  
  

  #!/bin/bash
  
  
  
  # Launch config script in background
  
  # Note there is no RabbitMQ Docker image support for executing commands after server (PID 1) is running (something like "ADD schema.sql /docker-entrypoint-initdb.d" in MySql image), so we are using this trick
  
  /config_rabbit.sh &
  
  
  
  # Launch
  
  /docker-entrypoint.sh rabbitmq-server
  
  

  
  


• 
  

  config_rabbit.sh

  
  
  

  #!/bin/bash
  
  
  
  # This script needs to be executed just once
  
  if [ -f /$0.completed ] ; then
  
    echo "$0 `date` /$0.completed found, skipping run"
  
    exit 0
  
  fi
  
  
  
  # Wait for RabbitMQ startup
  
  for (( ; ; )) ; do
  
    sleep 5
  
    rabbitmqctl -q node_health_check > /dev/null 2>&1
  
    if [ $? -eq 0 ] ; then
  
      echo "$0 `date` rabbitmq is now running"
  
      break
  
    else
  
      echo "$0 `date` waiting for rabbitmq startup"
  
    fi
  
  done
  
  
  
  # Execute RabbitMQ config commands here
  
  
  
  # Create user
  
  rabbitmqctl add_user USER PASSWORD
  
  rabbitmqctl set_permissions -p / USER ".*" ".*" ".*"
  
  echo "$0 `date` user USER created"
  
  
  
  # Create queue
  
  rabbitmqadmin declare queue name=QUEUE durable=true
  
  echo "$0 `date` queues created"
  
  
  
  # Create mark so script is not ran again
  
  touch /$0.completed
  
  

  
  

Here is an example of how I add an unprivileged user gg RUN useradd -d /home/gg -m -s /bin/bash gg
RUN echo gg:gg | chpasswd
RUN echo 'gg ALL=(ALL) NOPASSWD:ALL' >> /etc/sudoers.d/gg
RUN chmod 0440 /etc/sudoers.d/gg

I had to make a few changes to the script in the accepted answer to get it working based on the comments above.

Dockerfile

FROM rabbitmq



# Define environment variables.

ENV RABBITMQ_USER user

ENV RABBITMQ_PASSWORD user



ADD init.sh /init.sh

EXPOSE 15672



# Define default command

CMD ["/init.sh"]

init.sh

#!/bin/sh

( sleep 10 && 

rabbitmqctl add_user $RABBITMQ_USER $RABBITMQ_PASSWORD && 

rabbitmqctl set_user_tags $RABBITMQ_USER administrator && 

rabbitmqctl set_permissions -p / $RABBITMQ_USER  ".*" ".*" ".*" ) & 

rabbitmq-server