👋 Hey all 👋 I couldn't find any existing instances of this question, I'm trying to figure out how to deploy user code separately from Dagit, where all Dagster components (except for user code) are deployed by Helm. I managed to run a gRPC server locally with my repository, and connect a local Dagit instance to it using the

workspace.yaml

file. Now, if I want to do something similar in a k8s cluster, I guess I need to: 1. Deploy the gRPC server in its own Deployment + Service 2. Configure Dagit's workspace config-map to

load_from

the service However, AFAIK if the

userDeployments

value is enabled, then the chart expects to create a new deployment + service for that. In other words, I don't see a way to point it to an already-running service. Am I missing anything? Is there an easy way to deploy Dagster on k8s and still maintain user deployments independently?

cc @rex ya I’m not sure if this is possible with the current way the helm chart is set up

hi @assaf , thanks for reaching out! supporting this out of the box is on our roadmap for 0.11.0 https://github.com/dagster-io/dagster/issues/3018. one option in the interim is that we can add a “dagit command override” that lets you point dagit at an arbitrary workspace definition file instead of

/dagster/helm/dagster/templates/configmap-workspace.yaml

which could then allow you to point at more flexible locations

Thanks for the replies, I appreciate it.

@rex is planning to work on this next week and can let you know what an eta looks like. Are there any requirements that you have that aren’t yet in the issue? Just want to make sure that the design solves your case 🙂