Log the in progress work as a warning when force terminating.

[jonhyman]

Helps trace through logs to understand what jobs got run more than once if a worker was forced to shutdown.

[mperham]

Should it only log if necessary?

[jonhyman]

Do you mean moving the logging to before the t.raise Shutdown line? I thought if this log statement was executed, some job(s) was going to get run more than once.

[mperham]

I mean log.warn {... } if necessary?

[jonhyman]

I think that any clause we put there will by and large always eval to true. E.g.,

logger.warn { ... } unless @busy.empty?

should still log since we're in the hard_shutdown_in method, meaning that busy is non-empty (otherwise clean_up_for_graceful_shutdown would have called shutdown)

I think it's possible that @busy finishes while this code is being called, so I can update to guard against that but I think it'll almost always eval to true.

Is there some other guard condition I'm not thinking of?

[mperham]

busy can't change while we're executing. Actors are exclusive - the manager manages its own internal state and it only processes one message at a time.

I see what you are saying though. If busy is non-empty, in_progress should be non-empty too. Log in_progress on a separate log statement and I think that's fine.

[jonhyman]

updated