impala-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Thomas Tauber-Marshall (Code Review)" <ger...@cloudera.org>
Subject [Impala-ASF-CR] IMPALA-3524: Don't process spilled partitions with 0 probe rows
Date Fri, 27 Jan 2017 19:35:23 GMT
Thomas Tauber-Marshall has posted comments on this change.

Change subject: IMPALA-3524: Don't process spilled partitions with 0 probe rows
......................................................................


Patch Set 9:

(1 comment)

http://gerrit.cloudera.org:8080/#/c/5389/6/be/src/exec/partitioned-hash-join-node.cc
File be/src/exec/partitioned-hash-join-node.cc:

Line 677:     if (output_unmatched_batch_iter_->AtEnd()) {
> It's set in BufferedTupleStream::GetNextInternal() (this code really needs 
So then, if I understand correctly, you mentioned a build partition with >8mb of data,
since that's the default block size, but the tests in this patch are custom cluster tests
that set the read_size to 200kb, so I should be able to trigger this with a build partition
thats >200kb? As well, when you said "uses multiple buffers" you're referring to multiple
'BufferedBlockMgr::Block's?

If that's correct, then the tests that I already have cover this situation, though I've never
seen them fail even without the call to TransferResourceOwnership. Do you know if ASAN should
be able to detect this?


-- 
To view, visit http://gerrit.cloudera.org:8080/5389
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I175b32dd9031e51218b38c37693ac3e31dfab47b
Gerrit-PatchSet: 9
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Thomas Tauber-Marshall <tmarshall@cloudera.com>
Gerrit-Reviewer: Dan Hecht <dhecht@cloudera.com>
Gerrit-Reviewer: Thomas Tauber-Marshall <tmarshall@cloudera.com>
Gerrit-Reviewer: Tim Armstrong <tarmstrong@cloudera.com>
Gerrit-HasComments: Yes

Mime
View raw message