Github user andrewor14 commented on the pull request:
https://github.com/apache/spark/pull/11272#issuecomment-187320271
I would like to backport a variant of this patch into 1.6.1. This was
clearly working in 1.5 but then became unusable in 1.6. This means people who
upgrade to 1.6 will no longer be able to use dynamic allocation in Spark, a
huge regression.
As with all backports there is some risk that we might introduce a
regression between 1.6.0 and 1.6.1. It's *especially* bad if we introduce a
regression between maintenance releases, however, because otherwise users will
lose confidence in upgrading.
Therefore, it would be best to keep the changed surface area minimal, i.e.
to ensure that this patch *only* affects users of Spark on Mesos with external
shuffle service. I haven't looked at this patch in great detail but @bbossy
@dragos it would be great if you could keep that in mind when reviewing /
addressing review comments.
Thanks for fixing this critical issue!
---
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]
