Hi all, I am new to dagster. I am deploying dagste...
# ask-community
y
Hi all, I am new to dagster. I am deploying dagster OSS to K8s (specifically AKS) using helm chart ver 1.4.7. I want to run my own gRPC server by disabling subchart deployment using external configmaps to point to workspace.yaml file. I have come to a realization that I want to validate that it is correct: • If I have multiple pipelines (jobs) that focus on a given workload (for example ETL) , this code repo maps to 1 one dagster code definition location ,right? • Each code definition (code repo) maps to one instance (deployment) of gRPC server, correct?
Is it possible to have one deployment of gRPC server exposing multiple code definitions?
j
have you checked out the page about Definitions and Code Locations? https://docs.dagster.io/concepts/code-locations
y
Yep! I have. I wanted to validate my assumptions. One code definition -> One gRPC server?
j
yep!
y
I am wondering why there is a one to one relationship between a gRPC server and a user deployment (One code definition)? I am trying to understand the architectural design here
The way I thought it worked was deploy a replica set gRPC server, configure my user deployments on gRPC replica set as a configmap and mount my code definitions via an azure file share. That way every time I need to deploy a code change / add new code definitions. All I need to do is: • Update workspace.yaml configmap • Update gRPC list of code definitions