k8s_job_executor

throws error

__new__() got an unexpected keyword argument 'value'

after hitting Launch Run Dagster version

0.14.3

----------------- context: I have a single-op job that works perfectly with K8s run launcher. I define requirements in

@job()

However, I’m moving my resource requirements into

@op()

and changing to

@job(executor_def=k8s_job_executor)

so that I can launch dynamic ops on separate pods with those requirements.

Hi Liezl - try tagging your base image with the dagster version you want to use instead of latest: https://hub.docker.com/r/dagster/user-code-example/tags I think by using latest you are ending up with a different dagster-cloud version than the other dagster libraries and they're supposed to stay in lockstep

k8s executor works now but I’m getting an input not found error when using k8s executor + dynamic op with 1 extra arg (

dynamic_args: Dict[str, Any]

) dagster version

0.14.12

This looks to me like you want to configure a different IO manager that persists the output of ops - the default filesystem io manager isn't going to work when each op is running in its own pod. In GCP there's an example that sets up a GCS IO manager here: https://docs.dagster.io/deployment/guides/gcp#using-gcs-for-io-management And more about IO managers in general here: https://docs.dagster.io/concepts/io-management/io-managers