Hi :) What is the best solution to deploy a custom...
# prefect-community
Hi :) What is the best solution to deploy a custom flow with some python dependencies? I know I need a remote storage (mostly git* or docker). I didn't got the key differences between DockerRun (run_config) or Docker (storage). For simple python dependencies: should I use the official prefect image and set the python_dependencies via the storage object? So that i don't even need an dockerfile. Is there an example anywhere? I think it's an absolut basic task... We use a single machine setup at the moment. Best regards Michael
And how do I have to deploy the created image?
ā€¢ DockerStorage puts your flow in a docker image. DockerRun can use docker storage but can also use other storage so if your flow is using Github storage it'll pull from there then execute it in the docker image you've specified. ā€¢ For dependencies, DockerStorage with the
is a simple solution. The downside is that everytime your flow changes you have to rebuild a docker image. If you use a DockerRun referring to an image you've built (use the prefect image as a base and install your own dependencies on top of it) then you can use Github storage to build your flow (which is much faster)
If you are using a single machine, you can just use a LocalAgent and install the dependencies on that machine manually as well.
If you're using
you would deploy your image using typical docker build/push commands then refer to the tag in your code. If you're using
you would deploy your image by setting the
Thank you very much for your answer! šŸ™‚ Ok, if I use different flows with different dependencies, i guess the best solution is a own docker image, which will load the code from github storage. Are there any example configurations?
And: How to provide all my files (using some classes etc.)
You would just setup github storage as normal then a docker run pointing to an image you've built as in https://docs.prefect.io/orchestration/flow_config/run_configs.html#dockerrun
There's not a full example I can point you to though.
To provide all your files just copy them into your docker image using a Dockerfile
Ok. Thanks a lot. Will try it today.