Hi, given that I have k8s or docker work pool, is this possible to incorporate worker into the container itself, set it's type to 'process' and keep container running and ready to accept requests even after run_deployment is done? I am going to run thousands of jobs on such containers with AI models and spinning up them again from the scratch would take a lot of time since loading model weights is pretty heavy task

Hey. Yes. I have a Hashicorp Nomad cluster that runs multiple docker containers for different workpools as just Process workers.

hi igor, if you want to avoid dynamic infra provisioning you’d be free to define a process worker as a k8s deployment or simply serve a flow (or task) as a long lived container