Github user koeninger commented on the issue:
https://github.com/apache/spark/pull/15102
I'm not concerned about people deleting partitions before messages have
been processed, because they can take care of that problem themselves, by
not deleting things until consuming has caught up. This is fairly intuitive
IMHO.
I am concerned about shipping an implementation that can't handle partition
changes at all, because that's a big step backwards.
On Sep 18, 2016 11:41 PM, "Shixiong Zhu" <[email protected]> wrote:
> We do need to handle it comparing completely different topicpartitions,
> because it's entirely possible to have a job with a single topicpartition
> A, which is deleted or unsubscribed, and then single topicpartition B is
> added, in the space of one batch.
>
> Yeah, it's also possible that when a batch is running, a single
> topicpartition C is created then deleted. But the Kafka source doesn't
even
> know C. The root issue is currently the approach is polling metadata, and
> any metadata changes between two pollings are missing.
>
> â
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub
> <https://github.com/apache/spark/pull/15102#issuecomment-247910429>, or
mute
> the thread
>
<https://github.com/notifications/unsubscribe-auth/AAGAB4dbMY6oP4EYl062Dd3chUtOk520ks5qrhKQgaJpZM4J9QvR>
> .
>
---
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]
