35

I'm trying to run rabbitMQ using docker-compose, but the service is always starting or unhealthy.
rabbit is running fine, so I suspect there is something wrong with my health check.

Running the healthcheck command locally does return a value.

> curl -f http://localhost:5672 AMQP %

But docker-compose ps always says the service is unhealthy (or starting, before it runs out of time).

> docker-compose ps docker-entrypoint.sh rabbi ... Up (unhealthy) 15671/tcp

Here is what my docker-compose.yml file looks like.

# docker-compose.yml version: '2.3' # note: I can't change this version, must be 2.3 volumes: rabbit-data: services: rabbit: hostname: 'rabbit' image: rabbitmq:3.8.5-management healthcheck: test: ["CMD", "curl", "-f", "http://localhost:5672"] interval: 30s timeout: 30s retries: 3 ports: - '5672:5672' - '15672:15672' volumes: - 'rabbit-data:/var/lib/rabbitmq/mnesia/' networks: - rabbitmq networks: rabbitmq: driver: bridge

I have also tried using nc instead of curl in the healthcheck, but got the same result.

healthcheck: test: [ "CMD", "nc", "-z", "localhost", "5672" ]

From https://github.com/docker-library/rabbitmq/issues/326

Improve this question

5 Answers 5

Reset to default
52

You could use the command rabbitmq-diagnostics -q ping in case you just need a basic check.

healthcheck: test: rabbitmq-diagnostics -q ping interval: 30s timeout: 30s retries: 3

More information on how to run more advanced health checks could be found here

Improve this answer
4
  • Thanks a million, that's exactly what I needed! The docs mention each stage builds on the previous. >"This includes the stage 1 check plus ..." Do you know if you have to chain the commands? or does each command implicitly cover the previous one? e.g. which do I need for stage 2? 1. rabbitmq-diagnostics -q ping && rabbitmq-diagnostics -q status 2. rabbitmq-diagnostics -q status Commented Aug 13, 2020 at 11:20
  • 2
    Option 2 should be enough. Every time the status command returns OK, the ping command would have been successful too. But, the status makes some additional checks. Commented Aug 18, 2020 at 14:56
  • Memory alarms are not detected here, ex: "memory resource limit alarm set on ... Publishers will be blocked until this alarm clears" Commented Jul 20, 2022 at 7:22
  • If you have a cluster, then you need to check the node is booting otherwise it blocks the boot. You can use jq and test if the node is not booting for example (Swarm) : rabbitmq-diagnostics --formatter=json is_booting | jq --exit-status -n '.result' || rabbitmq-diagnostics -q check_running Commented Jun 22, 2023 at 10:44
10

I just want to add to above answers about rabbitmq-diagnostics, which worked great, however there was one missing part.

 depends_on: rabbitmq: condition: service_healthy rabbitmq: image: rabbitmq:3.8.17-management healthcheck: test: rabbitmq-diagnostics check_port_connectivity interval: 30s timeout: 30s retries: 10

With that my dependent service waits exactly what is needed to.

Improve this answer
1
9

If rabbitmq-diagnostics -q ping not working for you, check this

healthcheck: test: rabbitmq-diagnostics check_port_connectivity interval: 1s timeout: 3s retries: 30

In my example I also reduced interval.

Improve this answer
2

Your solution is good if you do not own / control the RabbitMQ Compose code but still have to check the service presence and your RabbitMQ image contains the management plugin.

First of all you are using curl which sends an HTTP request which on the other hand rabbitmq is not HTTP.

To fix this you can modify you command as follows:

# Note the port 15672 is the RabbitMQ Management UI port # It responds with 200 OK and the HTML page which is acceptable for a healcheck test: ["CMD", "curl", "-f", "http://localhost:15672"]

Note also that localhost may not work. Instead you should most likely use the RabbitMQ service name declared externally as host name (e.g. http://rabbitmq-service:15672) and ensure your service containing health check is on the same external network with the RabbitMQ server.

As well you have to have curl or netcat package installed on the Docker image your run the container from. This would require to construct a custom dockerfile.

Other answers show lots of approaches the rabbitmq team suggests in case the RabbitMQ service is defined in your Docker Compose file.

Improve this answer
1

I will go with this, I want to raise issues like "Publishers will be blocked until this alarm clears":

 healthcheck: test: rabbitmq-diagnostics -q status && rabbitmq-diagnostics -q check_local_alarms interval: 60s timeout: 30s retries: 3
Improve this answer
1
  • As it’s currently written, your answer is unclear. Please edit to add additional details that will help others understand how this addresses the question asked. You can find more information on how to write good answers in the help center. Commented Jul 20, 2022 at 10:16

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.