Github user andrewor14 commented on the pull request:
https://github.com/apache/spark/pull/251#issuecomment-38973103
To add to the existing discussion, I think the difference between the two
proposed approaches is in the mechanism, not in the semantics. The semantics of
both approaches are actually very similar in that in both approaches there
cannot be another event posted after the Shutdown event.
My only concern with busy waiting is that if the listener code does take
forever to finish, then the SparkContext will hang, and in the mean time use up
a lot of resources. Since multiple applications may run on the same cluster,
this may end up being costly. (In the broader context, however, this case is
relevant to like <0.1% of all Spark applications, so IDK how important this
difference actually is).
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
