Github user sryza commented on the pull request:
https://github.com/apache/spark/pull/6652#issuecomment-109445990
On the other side, I can also envision situations where it would be really
helpful to request a higher number of preferred locations. Consider a large
YARN cluster where an app gets executors on a small subset of the nodes. The
app is idle for a while after running an initial job, so the initial set of
executors go away. The app wakes up later and wants to use the output of its
first stage for another job. If the scheduler could request tasks on the nodes
housing the outputs of the first stage, with SPARK-4352, the
ExecutorAllocationManager would be able to request executors on these nodes.
This probably makes sense as later work, but still thought it could be
worth bringing up as part of the discussion.
---
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]
