RabbitMQ container is stopped by macOS suspension and wake-up

[amills-vibeirl]

Describe the bug

I have this docker-compose file below. Mongo and Redis always survive when I close my Macbook and reopen. But rabbit always dies. I am not sure if this is intended or not, but it's certainly not convenient for devs that end-up running things locally quite often. Is there a good reason why it always dies? Can't do a suspend? Obviously I don't know all the goings-on when the machine suspends, but I am curious.

version: '3.9'

services:

  redis:
    image: redis:latest
    ports:
      - "6379:6379"

  rabbitmq:
    image: rabbitmq:3-management
    container_name: "vibe-rabbitmq"
    volumes:
      - rabbitmq_data:/var/lib/rabbitmq
    ports:
      - "5672:5672"
      - "15672:15672"
      - "5552:5552"
    environment:
      - RABBITMQ_PLUGINS=rabbitmq_stream,rabbitmq_stream_management
      - RABBITMQ_DEFAULT_USER=rabbitmq
      - RABBITMQ_DEFAULT_PASS=rabbitmq

  mongo:
    image: mongo:6.0.12
    container_name: "vibe-mongo"
    restart: always
    ports:
      - "27017:27017"
    volumes:
      - mongo_data:/data/db

volumes:
  mongo_data:
  rabbitmq_data:

any feedback appreciated, but seems like unintended behavior.

Reproduction steps

1. docker-compose up
2. close macbook pro
3. reopen macbook pro

mongo --> alive
redis --> alive
rabbit --> dead

Expected behavior

expect rabbit to survive

Additional context

n/a

[michaelklishin]

You haven't provided any logs or details. What "dies" in your setup is not RabbitMQ, it is a container, and that can happen for many different reasons.

The only relatively unusual thing about OS suspension's effects on RabbitMQ is that when time suddenly jumps (this is the case when the OS is waken up), the Erlang runtime used to fire a lot of timers in rapid succession. This means a CPU usage spike. Docker or something else can potentially kill a container based on its CPU usage (that's a guess).

Anecdotal evidence suggests that this is a lot less of a problem with recent versions of Erlang.

[michaelklishin]

I cannot reproduce. Here are the steps:

1. Start a RabbitMQ 3.12.10 node on Erlang 26.2.1 using the generic UNIX package (in the foreground)
2. Apple logo => sleep => wait for two minutes
3. Wake the machine up
4. The node is up and running
5. Pull the charging cable in a Mac that has its lid closed
6. Wait for 4 minutes
7. Wake the machine up
8. The node is up and running

In fact, on many occasions I have left nodes up and running in the foreground for several hours or even a couple of days. These nodes never terminated.

Maybe Docker image maintainers would know what relevant logs can help. The image configures RabbitMQ to log into standard output.

Whether docker-compose logs is how those logs can be accessed between OS suspension and wake up, I don't know. This is a question for a Docker user community.