Better Error Messaging for Users when docker-compose up Fails

[brandenchan]

Is your feature request related to a problem? Please describe.
I was running docker-compose up on a Macbook Pro. Docker Desktop was allotted 4GB of memory. The REST API would continually fail to load the ElasticsearchDocumentStore and hence also the Retriever. The Elasticsearch instance would keep starting, crashing and restarting. There was no error message as to why this was happening.

After some deep exploration with @tstadel , we found that this machine was able to start up the REST API and the Elasticsearch containers separately. But when started up together using docker-compose up, the Elasticsearch container would crash because of insufficient memory. Since the restart: on-failure is set in docker-compose.yml it would keep trying to restart.

After we changed the restart policy to:

deploy:
  restart_policy:
    condition: none

Error code 137 would be returned when ES crashed.

Calling docker container inspect haystack_elasticsearch_1 after crash might show "State"/"OOMKilled": true

Describe the solution you'd like
It would be great to have some kind of error message displayed to the user about why docker-compose up is not working in this situation. It would also be good to explain that you might need to allot more memory to Docker to ensure the containers don't crash, and that you can also change the number of workers to reduce memory consumption.

[masci]

A fix like it's described in this issue is not possible: we can't deduce why one of the containers orchestrated is killed or stuck in crashloop.

What we can do is providing hardware requirements for running the demo, adding the documentation label.

[masci]

We've since added an healthcheck for the ES container, the crash-loop can't happen anymore and the failure would be now evident, closing.