Github user ksakellis commented on the pull request:
https://github.com/apache/spark/pull/3711#issuecomment-69804537
@pwendell @nitin2goyal I had previously thought of adding it to the
DAGScheduler - this would avoid the duplicated code between the
CoarseGrainedSchedulerBackend and MesosSchedulerBackend. The reason I did not
go with this approach is two fold:
a) Not enough information is available in the the DAGScheduler (don't know
the core count). This is more minor and can be added to the callbacks.
b) More importantly, the code path from
CoarseGrainedScheduler.resourceOffers->TaskSchedulerImpl.resourceOffers->DAGScheduler.executorAdded
happens *before* the executor actually registers with the scheduler backed. I
wanted the events to be published *after* the executor registers via the Akka
event.
The consequence is that the events are generated at a lower level but I
think they are more accurate. Thoughts? I'd really like to get this merged so
that it can unblock: https://github.com/apache/spark/pull/3486 (Yarn links in
UI) which i think is pretty important for usability.
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
