Hey folks - on 0.14.5, whenever we roll a new user code repo, it looks like

dagit

picks up the changes, but on running jobs, we hit some unexpected errors if the pipeline notably changed. Some examples: 1.

Received unexpected config entry "user_ids" at path root:inputs.

We hit this when we added a new

user_ids

input to the job. 2.

An exception was thrown during execution that is likely a framework error, rather than an error in user code.

Followed by a

KeyError

with a new output from a new operator. Whenever this happens, force-reloading the repo in

dagit

always fixes the issue. I'm wondering why

dagit

appears to have picked up the changes, but executing the job fails.

a bit confused, are you saying that these errors disappear when you reload the repo? Or that when the job runs, that the changes aren't picked up?

When we try running the job, it looks like the changes aren't fully picked up. We then reload the repo via the UI and we rerun the job and it works. I guess I'm wondering if in our CI we need to force-reload the repo via Dagit when we update the user code to ensure future pipeline runs don't error out.

Ah I see. Yea I don't think the user code server automatically reloads upon every job launching. I think there was some work to make this happen at some point, but I don't know exactly where things ended up. Cc @sean