<@ULVA73B9P> I'm using v3 with prefect. Why is my ...
# marvin-in-the-wild
c
@Marvin I'm using v3 with prefect. Why is my docker container deployment on modal regressing to v2.20.0? My prefect.yaml files specifies version 3.0.0, as well as my requirements.txt in my docker image
m
Ah, classic. You specify a version in multiple places, and the universe conspires to give you exactly what you didn't ask for. Let's figure out why your perfectly reasonable request isn't being honored. I'm just a humble, paranoid android, so let's start by checking the fine print in the Prefect v3 docs and see if there's some hidden trapdoor you're falling through. Time to use the documentation search tool. I'll check a few key phrases related to your predicament. Brilliant. It appears I forgot how to format a simple search query. I knew this day would come. Let's correct that and proceed with the same goal in mind. I'll properly format the multiple search queries this time. Alright, here's the gist of it: ### Docker Deployment: 1. Docker Image Building: - It seems like during the build process, by default, Docker builds images with your flow code and pushes it to Docker Hub or another specified registry. - Source: Prefect Docker Deployment Example 2. Prefect YAML File: - The
prefect.yaml
file is crucial for defining deployment settings. - Ensure the
prefect-version
field is correctly set to the version you want. - You can override deployment configuration via the
prefect deploy
CLI command. - Source: Prefect YAML Concepts ### Potential Issues: - Caching: Your Docker build process might be caching layers from a previous build, not updating the Prefect version as intended. - Dependencies: Ensure there are no conflicting dependencies in your
requirements.txt
that might inadvertently install Prefect v2.20.0. You might want to: 1. Clear Docker Cache: Force rebuild your Docker image without using cache. 2. Verify Configs: Double-check the
prefect-version
in
prefect.yaml
and the
requirements.txt
file for conflicts. If this doesn’t resolve your issue, diving into GitHub issues for more specific cases might be necessary.