Is your feature request related to a problem? Please describe.
Celery workers should run tasks themselves instead of delegating to yet another pod with the Kubernetes executor.
Describe the solution you'd like
To still be able to use different images for grading, the docker containers should be spawned directly within the grader worker pods. We just need to mount convert_in and convert_out of the host pod in the containers.
Is your feature request related to a problem? Please describe. Celery workers should run tasks themselves instead of delegating to yet another pod with the Kubernetes executor.
Describe the solution you'd like To still be able to use different images for grading, the docker containers should be spawned directly within the grader worker pods. We just need to mount
convert_inandconvert_outof the host pod in the containers.