Hallo, concerning deploying dagster with Docker. There a very good repository from you with three containers, the code server, dagit and dagster daemon. https://github.com/dagster-io/dagster/tree/master/examples/deploy_docker I was able to run this setup locally. It functions well. What I don't understand is the fact that for each job run a new container is created. At least part of this process is defined in the docker-compose file, naming the image and passing it as variable CURRENT_DAGSTER_IMAGE. How can I disactivate this behavior and simply run the jobs with dagit from the code server without creating a container each time? Another question I have is concerning the location_name variable with a value of "example_user_code" in workspace.yml. Where exacly is this value ("example_user_code") defined (dagit expects this value)? Isn't the location suppose to be the location of the python file in the code server? When I log in into my code container there no such directory as "example_user_code" - can you help out please.

To run jobs from the user code server without creating new docker containers you'll want to switch to using the DefaultRunLauncher in your dagster.yaml file - I think you might be able to even just remove the

run_launcher

section from your current dagster.yaml file. Also see this document on RunLaunchers for a more in-depth understanding of how they work and what your options are. The location_name in the workspace.yaml is kind of just a tag used to differentiate different user code servers. That way if you're running multiple user code containers (a common deployment pattern for isolating environments and teams) Dagit can differentiate between them. This documentation around running gRPC servers gives more details about how to specify actual directory locations / packages / modules from your user code container to load code from

Thanks Zach!