What are people’s production deployment strategies...
# best-practices
m
What are people’s production deployment strategies for Prefect flows currently? Do people make use of Kubernetes/Azure Container Instannces etc. or also just VM’s that run everything? We are currently struggling quite a bit, also regarding whether we should be versioning YAML files, and how. Because currently we version CLI commands in a Makefile that get run by CI/CD
r
Similar to https://github.com/radbrt/orion_flows but on aws eks , gh actions deploys any new/changed flows that creates jobs within k8s from running agent(s) on eks
🚀 1
gratitude thank you 1
m
That looks very interesting! I was indeed looking to see if something like that existed, so indeed you can have a ‘local’ .prefectignore, because I noticed for us it was basically copy pasting the entire repo for every build. Additionally, you mention for new/changed flows, I wonder how you achieve that? Or at least could you point where in the repo you achieve this
r
Thats easy in gh actions using https://github.com/tj-actions/changed-files
for me I used the folder option and created a matrix of all new/changed flow or deployment folders - then loop to (re)deploy
c
Shoutout to @Henning Holgersen for the
orion_flows
repo 🔥
🙌 3