Am I right in thinking that this means any dagster code to be deployed has to be compatible with this version?

I don't think your Dagster code needs to be compatible with this version - the dagster-cloud action is just calling an API to tell Dagster Cloud about your new container, it shouldn't be directly interacting with your code at all. For context, we use that action in a Python 3.8 environment, but our code runs 3.9. But I personally would like to see the python version updated for this for consistency's sake

Ah, thanks. I was expecting dagster to be on 3.10 by now!

I”m not a cloud eng so not very familiar with what this does, but 3.8 is the lowest Python we support-- whenever we have to choose a single Python version that will be executing user code I believe we use 3.8.

Hm. I've definitely found that I had a syntax error from 'match' when trying to deploy. Swapping to if/else resolved the issue. That keyword is 3.10, which made me think this was a version issue.

Hi Brendan This Dockerfile is not used for your dagster code - it is just used to run the github action itself. You can pick which version of Python you want to run in Dagster Cloud. In our serverless action there's an env called

PYTHON_VERSION

that you set to the version you want here (

major.minor

only): https://github.com/dagster-io/dagster-cloud-action/blob/b38fdd8492435a226aa89ec14c765884a0860fb4/github/serverless/deploy.yml#L16 I've linked the latest github action yaml above - if you have an earlier version you can still set the python version here if you're using fast deploys, or here if you're not using the fast deploys.

Ah! Thanks! That's exactly what I wanted.

Ah you're right. We should add a

--python-version

flag to the

deploy-python-executable

command here: https://github.com/dagster-io/dagster-cloud-action/blob/b38fdd8492435a226aa89ec14c765884a0860fb4/src/gitlab_action/deploy.py#L35. I'll work on that. In the meantime you could directly call

dagster-cloud

in your

script:

section, with the additional flag. However it does get a little verbose, specially if you handle the branch deployments.