Giters

WIPACrepo / lta

Long Term Archive

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Clean up for Component base class

blinkdog opened this issue · comments

commented

The status_loop and work_loop of components calls check_drain_semaphore.

This seemed like a good idea before, but the way we run components makes this obsolete.
Generally:

  • We are running a component as a long-running daemon inside Kubernetes; the system can/will kill it at any time without providing a semaphore
  • We are running a component as a free-standing daemon on some host (i.e.: Bundlers on node007, c8-7, and c9-7); these are also killed without providing any semaphore
  • We are running the components as RUN_ONCE_AND_DIE as part of jobs at NERSC; these die after processing one unit of work and would not use the semaphore.