Hi, is it possible to add some additional init containers in to attach it to the workloads ie run configuration so that each

Op

, when it starts a Kubernetes Pod, it can run with additional init containers. This is specially needed for one of the most used security tool, Hashicorp Vault and Mutating web-hooks to inject credentials in the running container.

You can probably add them in the

dagster-k8s/config

tag under

pod_spec_config

So this thing bothers me, by using this style, we’re forcing devs to bring in lot of yaml config in python code, if this were utilized somehow from the Helm templates itself, it can become just a baseline configuration for each occurrence. Right now each job def will requires helm configuration outside of helm chart. It would be good to understand why this approach was chosen instead of allowing such configurations to be controlled by

values.yaml

file or

helm

templates

Indeed, the Dagster team is already aware of this and wants to review the way they do k8s configuration. In the meanwhile you can use the "op factory" pattern to abstract away the configuration from your developers -> https://docs.dagster.io/concepts/ops-jobs-graphs/ops#op-factory