Github user jerryshao commented on the pull request:
https://github.com/apache/spark/pull/8976#issuecomment-146465163
From my understanding for some batch workloads like ETL, traditional
process pattern is to read data from HDFS, process and output to HDFS, in such
process pattern data locality is quite important, instead dynamic allocation is
not quite useful, so normally user will not enable dynamic allocation, in such
scenario node locality will be a problem because there's no way to specify the
container locality based on the input data.
With SPARK-4352, to reactivate this feature is quite simple, besides if
dynamic allocation is enable the logic is the same, so instead of removing it,
reactivating this parameter might be another choice.
Just my opinion. Please point out if there's different concern.
---
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]
