impala-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alex Behm (Code Review)" <ger...@cloudera.org>
Subject [Impala-ASF-CR] IMPALA-2905: Handle coordinator fragment lifecycle like all others
Date Wed, 21 Sep 2016 18:37:50 GMT
Alex Behm has posted comments on this change.

Change subject: IMPALA-2905: Handle coordinator fragment lifecycle like all others
......................................................................


Patch Set 6:

(1 comment)

http://gerrit.cloudera.org:8080/#/c/4402/6/be/src/exec/push-pull-sink.cc
File be/src/exec/push-pull-sink.cc:

Line 31: constexpr int32_t QUEUE_DEPTH = 16;
> We need to carefully consider the implications on memory management. Some e
Thinking about this a little more, I think the accumulation of batches without copying the
tuple data in certain situations is wrong. It breaks our current memory-transfer contract.
We have a similar problem in the nested-loop-join where we lazily copy the tuple data for
all accumulated batches when a batch with "needs_to_return" is encountered.

I think we need to rethink the design to fit our current memory transfer model.


-- 
To view, visit http://gerrit.cloudera.org:8080/4402
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: Ibb0064ec2f085fa3a5598ea80894fb489a01e4df
Gerrit-PatchSet: 6
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Henry Robinson <henry@cloudera.com>
Gerrit-Reviewer: Alex Behm <alex.behm@cloudera.com>
Gerrit-Reviewer: Marcel Kornacker <marcel@cloudera.com>
Gerrit-HasComments: Yes

Mime
View raw message