[FIX] Add retries to docker-compose.yml, to wait for MongoDB to be ready

[tiangolo]

The current docker-compose.yml file requires manual steps to deploy a Rocket Chat stack, it cannot be done in a single deterministic command or integration with CI systems.

The current instructions require to wait for MongoDB to be ready, and then starting the mongodb-init-replica service manually, and then the rocketchat service manually too.

That means that, if starting from scratch, with a fresh MongoDB container, at the firsts docker-compose up -d, those two services will fail to start and stay down.

The proposed change makes those two services wait for MongoDB to be ready, by retrying 30 times, waiting 5 seconds before each retry. That way, when the service mongo is ready, mongo-init-replica will run, initialize the replica set and exit normally (with exit code 0). And then rocketchat will be able to start with the initialized MongoDB.

---

This would make the instructions for Docker Compose here: https://rocket.chat/docs/installation/docker-containers/docker-compose/

Be almost only docker-compose up -d.

And after a bit, everything will be ready.

The docs PR is here: https://github.com/RocketChat/docs/pull/1050

[CLAassistant]

All committers have signed the CLA.

[tiangolo]

I'm not sure what to do about the test errors as I didn't modify anything related to them.

[geekgonecrazy]

Thanks for the contribution! That's a nice 1 line check. Can you sign the CLA?

[tiangolo]

I did sign it! With the interactive CLAassistant

[tiangolo]

The companion docs PR is here: https://github.com/RocketChat/docs/pull/1050

[tiangolo]

I'm not sure about the process, I have signed the CLA about 3 times, and then the CLA Bot comments saying that I signed it.

And a couple days later it edits the comment and I appear again as if I hadn't signed it. Am I doing something wrong?

[geekgonecrazy]

Looks like all is good from CLA standpoint. 👍

[tiangolo]

Cool! Thanks

[tiangolo]

I see that the Circle CI build is always "Pending — Your job is on hold on CircleCI!", is there anything that I have to do on my side to "un-pause" it?

[rodrigok]

@tiangolo that's expected, it's a job we added to be able to trigger the docker image build manually when we need to test the PR, you can see it's not marked as required here.

[tiangolo]

Get it. Thanks!

[Sing-Li]

@tiangolo have you tested this with any particular version of mongo? On some mongo version, issuing the rs.initiate() EARLY will make the DB totally irresponsive to all subsequent requests 😞

[tiangolo]

I'm using it on a CI/CD environment/application with mongo:3.2, which is what was used when I created the PR. But I haven't updated it to Mongo 4.0.