How to retain RabbitMQ user accounts in Docker

Question

I am having trouble persisting dockerized RabbitMQ user accounts set up through the management panel. Upon restart they disappear and I believe it is related to new mnesia databases being created on each restart.

I tried binding a docker volume to /var/lib/rabbitmq:

version: '3.1'

services:

  rabbitmq:
    image: rabbitmq:management-alpine
    volumes:
      - rabbitdata1:/var/lib/rabbitmq/
    ports:
      - "5672:5672"
      - "15672:15672"

volumes:
  rabbitdata1:
    driver: local

When I look at the contents of the mounted directory, I get:

$ docker exec -ti local_rabbitmq_1 /bin/bash
bash-5.0# ls /var/lib/rabbitmq/mnesia/
rabbit@0eceaaa217c8                 rabbit@0eceaaa217c8-plugins-expand
rabbit@0eceaaa217c8-feature_flags   rabbit@0eceaaa217c8.pid

But when I restart the service, it seems like a new instance gets created for the new PID and all changes are lost:

$ docker exec -ti local_rabbitmq_1 /bin/bash
bash-5.0# ls /var/lib/rabbitmq/mnesia/
rabbit@0eceaaa217c8                 rabbit@ac5afbef3c81
rabbit@0eceaaa217c8-feature_flags   rabbit@ac5afbef3c81-feature_flags
rabbit@0eceaaa217c8-plugins-expand  rabbit@ac5afbef3c81-plugins-expand
rabbit@0eceaaa217c8.pid             rabbit@ac5afbef3c81.pid

I also tried setting the RABBITMQ_NODENAME environment variable so that instead of the above rabbit@0eceaaa217c8 and rabbit@ac5afbef3c81 I get a constant string for .pid and mnesia directories but then RabbitMQ would not even restart:

2020-04-10 10:41:34.657 [info] <0.309.0> Running boot step database defined by app rabbit
2020-04-10 10:41:34.685 [error] <0.308.0> CRASH REPORT Process <0.308.0> with 0 neighbours exited with reason: {{failed_to_cluster_with,[foo@190e6343c238],"Mnesia could not connect to any nodes."},{rabbit,start,[normal,[]]}} in application_master:init/4 line 138
2020-04-10 10:41:34.686 [info] <0.44.0> Application rabbit exited with reason: {{failed_to_cluster_with,[foo@190e6343c238],"Mnesia could not connect to any nodes."},{rabbit,start,[normal,[]]}}
{"Kernel pid terminated",application_controller,"{application_start_failure,rabbit,{{failed_to_cluster_with,[foo@190e6343c238],\"Mnesia could not connect to any nodes.\"},{rabbit,start,[normal,[]]}}}"}
Kernel pid terminated (application_controller) ({application_start_failure,rabbit,{{failed_to_cluster_with,[foo@190e6343c238],"Mnesia could not connect to any nodes."},{rabbit,start,[normal,[]]}}})

Is there any other way to retain changes between RabbitMQ docker service restarts?

Maybe there is some other directory that could do the trick?

I checked for other potential candidates but only found /etc/rabbitmq/ and /opt/rabbitmq/ which seem like configuration and installation directories respectively:

bash-5.0# find . -name 'rabbitmq'
./etc/rabbitmq
./var/log/rabbitmq
./var/lib/rabbitmq
./opt/rabbitmq
./opt/rabbitmq/etc/rabbitmq

Answer 1

Create Two folders, data and etc

enabled_plugins

[rabbitmq_management,rabbitmq_prometheus].

rabbitmq.conf

    auth_mechanisms.1 = PLAIN
    auth_mechanisms.2 = AMQPLAIN
    loopback_users.guest = false
    listeners.tcp.default = 5672
    #default_pass = admin
    #default_user = admin
    hipe_compile = false
    #management.listener.port = 15672
    #management.listener.ssl = false
    management.tcp.port = 15672
    management.load_definitions = /etc/rabbitmq/definitions.json

definitions.json

       {
        "users": [
            {
            "name": "admin",
            "password": "admin",
            "tags": "administrator"
            }
        ],
        "vhosts": [
            {
            "name": "/"
            }
        ],
        "policies": [
            {
            "vhost": "/",
            "name": "ha",
            "pattern": "",
            "apply-to": "all",
            "definition": {
                "ha-mode": "all",
                "ha-sync-batch-size": 256,
                "ha-sync-mode": "automatic"
            },
            "priority": 0
            }
        ],
        "permissions": [
            {
            "user": "admin",
            "vhost": "/",
            "configure": ".*",
            "write": ".*",
            "read": ".*"
            }
        ],
        "queues": [
            {
            "name": "job-import.triggered.queue",
            "vhost": "/",
            "durable": true,
            "auto_delete": false,
            "arguments": {}
            }
        ],
        "exchanges": [
            {
            "name": "lob-proj-dx",
            "vhost": "/",
            "type": "direct",
            "durable": true,
            "auto_delete": false,
            "internal": false,
            "arguments": {}
            }
        ],
        "bindings": [
            {
            "source": "lob-proj-dx",
            "vhost": "/",
            "destination": "job-import.triggered.queue",
            "destination_type": "queue",
            "routing_key": "job-import.event.triggered",
            "arguments": {}
            }
        ]
    }

Run Docker

docker run --restart=always -d -p 5672:5672 -p 15672:15672 --mount type=bind,source=E:\docker\rabbit\data,target=/var/lib/rabbitmq/ --mount type=bind,source=E:\docker\rabbit\etc,target=/etc/rabbitmq/ --name rabbitmq --hostname my-rabbit rabbitmq:3.7.28-management

Things would be persisted across restarts

Taken from here

Answer 2

In the documentation for the rabbitmq Docker Hub image it notes (under "Running the daemon"):

One of the important things to note about RabbitMQ is that it stores data based on what it calls the "Node Name", which defaults to the hostname. What this means for usage in Docker is that we should specify -h/--hostname explicitly for each daemon so that we don't get a random hostname and can keep track of our data.

The equivalent Docker Compose setting is hostname:. It defaults to the container ID, which changes every time the container is recreated, which is why you're not seeing data persisted and why the filenames have 12-hex-digit IDs in their names.

services:
  rabbitmq:
    image: rabbitmq:management-alpine
    hostname: rabbitmq                 # <-----
    volumes:
      - rabbitdata1:/var/lib/rabbitmq/
    ports:
      - "5672:5672"
      - "15672:15672"

(The only thing hostname: sets is what the container thinks its own host name is. It has no connection to the networking setup at all. Setting it usually isn't necessary, unless you have software like this that specifically looks at it.)