Just went through some headbanging re: environment variables and docker compose. The documentation was really useful on how to get env vars into images, so things went like this: • Set values in a .env file • Add the values to the

run_launcher.config.env_vars

in

dagster.yaml

• Build/fire up the docker compose • Exec into the running user code container • See that the env vars are there (yay) • Go to dagit, fire off a run (where ops are looking for something in the env) • Failed jobs - the env I expect isn’t there The issue turned out to be a misconception I had about who’s launching the run. I was assuming that the user code image was launching from itself, but because I was using the docker-compose.yaml that set up a daemon and used the queued run launcher. So I needed to pass those envs through to the daemon container in docker-compose.yaml. So my feedback is: please update the docs to be clear about where/how envs are passed, esp in the queued run coordinator context (and esp in the example docker-compose.yaml). Thanks yall!

Hey there! I'm glad the docs were helpful at the start (woo!), but agreed they could be a bit clearer about how/where they're passed in. Adding this to our docs to-do list now.

https://dagster.slack.com/archives/C01U5LFUZJS/p1675100051270359 I agree 😄