Worker handles SIGINT and SIGKILL.

[williams-jack]

Feature/Problem Description

Workers populated by a larger Kubernetes system should be able to handle being shut down (e.g., in a scale down scenario). This means ensuring that a job that has not been completed is able to be regraded and picked up again ASAP.

Solution (Changes Made)

• GracefulKiller class to manage signal handling.
• Observer to look for a .jobcompleted or .killjob file to signify the termination of a job.
  • The former case is needed to ensure the thread (wrapped by a Future) does not run forever.
• Re-enqueue function for when a job is killed prematurely.

[blerner]

Observer to look for a .jobcompleted or .killjob file to signify the termination of a job. Is this still present? I don't see any use of it any more...

[williams-jack]

Observer to look for a .jobcompleted or .killjob file to signify the termination of a job. Is this still present? I don't see any use of it any more...

All instances were in __main__.py and stop_worker.py. Everything's gons from there and I ran a grep on all the code -- we should be all set.

[blerner]

Then why are they still in .gitignore?