digital_slide_archive
digital_slide_archive copied to clipboard
DigitalSlideArchive
Refactor deployment to just use docker-compose
The deploy_docker.py scripts works, but is cryptic to maintain. Now that docker-compose can do almost everything we need, we should switch to that. There needs to be a migration path as well.
We have a docker-compose setup in the devops directory, but this currently doesn't handle running HistomicsTK or other docker algorithms and has some other limitations.
There is a work-in-progress of this on the https://github.com/DigitalSlideArchive/digital_slide_archive/tree/docker-compose branch.
The docker-compose script can handle everything that the deploy_docker.py script can do. If there is still substantial value in a command-line starting script, it would probably be better to emit a docker-compose yaml file and then defer to docker-compose for actions.
deploy_docker.py has some complex options, such as using a localhost version of rabbitmq, mongodb, and memcached. It would be useful to know if any of these are being used and need to be maintained, or if we can discard some of the deployment options.
We are preferring the docker-compose deployment. At this point, we should mark the deploy_docker script as deprecated.