This repo contains definitions of various task types designed to run as independent Kubernetes Jobs managed by the JobHandler deployment. The primary components required by the JobHandler are the files (shown relative to this repo root path)
[task_type]/Dockerfile[task_type]/worker/init.py[task_type]/worker/task.py[task_type]/worker/prestop.py[task_type]/worker/jobconfig_spec.tpl.yaml
The Dockerfile is used to build the image that will define the Pod container that the Job creates once launched by Kubernetes. By convention, the container will execute the command python init.py by the pod's lifecycle postStart hook, and then python task.py will be executed by the container to accomplish the desired task. If the Job is terminated by Kubernetes, triggering the preStop lifecycle hook, the prestop.py script should be used to report the task failure to the JobHandler.
Task containers are provided information by the JobHandler via a Kubernetes ConfigMap mounted as a file in the container's filesystem: /home/worker/configjob.yaml. The structure of this YAML file is defined by the JobHandler using a template file. The metadata key in this data structure contains the information common to all task types. The spec key is defined individually by each task type via the jobconfig_spec.tpl.yaml file to satisfy its unique requirements.