this may be a stupid question, but I’ve tried to build a deployment with:

prefect deployment build flows/hello_world.py:hello_world --name "Hello world dev" --tag kubernetes --tag dev --infra kubernetes-job --infra-block base-dev

and I got this:

File "/Users/tessellir/Projects/il-da-orchestrator/.venv/lib/python3.8/site-packages/prefect/blocks/core.py", line 501, in load
    block_type_slug, block_document_name = name.split("/", 1)
ValueError: not enough values to unpack (expected 2, got 1)
An exception occurred.

I’ve ofc created a block in Prefect cloud of type KubernetesJob called base-dev, what am I missing?

-ib kubernetes-job/base-dev works for me, but it should indeed work as you tried to build it

it did the trick, thanks!

I assume you mean "infer 'kubernetes-job' from 'base-dev' ", right? If the name of an infra-block is unique across all infrastructure type, yes. But I don't know whether that's a constraint right now

now I’m trying adding a local storage to the deployment with:

prefect deployment build flows/hello_world.py:hello_world --name "Hello world dev" --tag kubernetes --tag dev --infra kubernetes-job --infra-block kubernetes-job/base-dev --storage-block local-file-system/runtime-folder

and I have the block in the image on cloud, but I get this

File "/Users/tessellir/Projects/il-da-orchestrator/.venv/lib/python3.8/site-packages/prefect/filesystems.py", line 114, in put_directory
    shutil.copytree(from_path, local_path, dirs_exist_ok=True)
NameError: name 'from_path' is not defined
An exception occurred.

Local storage can't be used when using KubernetesJob infrastructure, according to the docs: https://orion-docs.prefect.io/concepts/infrastructure/#kubernetes-flow-runner (link is okay, title has yet to be modified). Actually, I think you're trying something similar to what I am. You also want to store your flow inside a docker image, that is then executed as a kubernetes job?

Hey, I succeeded in having prefect 2.0.0 run as KubernetesJob. Maybe these steps help, if not, sorry for hijacking your thread :-) 1. Create a Docker image containing all my dependencies 2. Create a GCS Storage Block in prefect UI called my_gcs_block 3. Deploy an Agent on kubernetes 4. Create a depoyment with prefect deploymemt build <pathtoflow.py:flowname> --name myname_deployment -i kubernetes-job -sb gcs/my-gcs-block 5. In the resulting deployment.yaml, I manually adjusted the Kubernetes Job settings. Specifically, env, image ans service_account_name 6. prefect deploymemt apply... 7. prefect deploymemt run (note that execute does not work, only run) I think its good to define the KubernetesJob in the deployment.yaml rather than in the UI, because it helps me to better integrate it with our infrastructure as code strategy. Not sure if this is helpful - if you want me to post my deployment configs or more details about some individual step, let me know

Hi @Andreas Nigg it helps lot thank you. I’m now wondering about the best way to edit env, image and service_account_name programmatically for automated deployment. Currently using sed but a it’s a bit hacky

For me it's also sed at the moment. Maybe for automated deployment, first defining a KubernetesJob block in the prefect UI (or with the python client) and then using this precreated block with "prefect deploymemt build -ib <block-name>" would work well in such a scenario?

Hey you idea si great I haven’t been discovered Kubernetes job as a block before your message ! Thanks !

@Mathijs Carlu that is exactly what I want to do sir, have you managed to find a workaround?

@Riccardo Tesselli yes, you can find it over here