[ 
https://issues.apache.org/jira/browse/IMPALA-5293?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16878019#comment-16878019
 ] 

ASF subversion and git services commented on IMPALA-5293:
---------------------------------------------------------

Commit 111035ef77f0f3331fff6abe3e60185c3d4e9a10 in impala's branch 
refs/heads/master from Abhishek Rawat
[ https://gitbox.apache.org/repos/asf?p=impala.git;h=111035e ]

IMPALA-8673: Add query option to force plan hints for insert queries

IMPALA-5293 enabled the pre-insert clustering by default. This could
cause performance regression and this change provides a query option
for setting default hints for INSERT statement.

New query option 'DEFAULT_HINTS_INSERT_STATEMENT' was added. It also
supports adding multiple supported hints when separated by ':'
  set DEFAULT_HINTS_INSERT_STATEMENT=[clustered|noclustered];
  set DEFAULT_HINTS_INSERT_STATEMENT=[shuffle|noshuffle];
  set DEFAULT_HINTS_INSERT_STATEMENT=
          [clustered|noclustered]:[shuffle|noshuffle];

If a given insert statement already has plan hints in the query text,
the default hints, if any, are all ignored. This is because, if a query
has plan hints specified by the user, we don't want to override it.
When a default hint is set, and there is an INSERT statement without any
plan hints in the query text, the default hints have the same affect as
they would have had, if they were applied as plan hints in the query
text. So these default hints have the same application and restrictions
as the existing plan hints for INSERT statement. The default hints apply
to HDFS and Kudu table formats and are ignored for HBase table format.

Testing:
- Added unit tests in AnalyzeDDLTest for CTAS.
- Added unit tests in AnalyzeStmtsTest for insert statements.
- Added unit tests in PlannerTest validating the plan for various
  scenarios involving different combinations of default hints.

Change-Id: I1c3f213402b8e4d1940f96738ad21edf800fa43a
Reviewed-on: http://gerrit.cloudera.org:8080/13753
Reviewed-by: Impala Public Jenkins <[email protected]>
Tested-by: Impala Public Jenkins <[email protected]>


> Turn insert clustering on by default
> ------------------------------------
>
>                 Key: IMPALA-5293
>                 URL: https://issues.apache.org/jira/browse/IMPALA-5293
>             Project: IMPALA
>          Issue Type: Bug
>          Components: Frontend
>    Affects Versions: Impala 2.8.0
>            Reporter: Lars Volker
>            Assignee: Lars Volker
>            Priority: Major
>              Labels: compatibility, include-in-v3, incompatibility
>             Fix For: Impala 3.0
>
>
> Currently pre-insert clustering needs to be enabled using the {{clustered}} 
> hint. We should enable it by default, allowing it to be disabled using the 
> {{noclustered}} hint.
> It may be necessary to wait until the next compatibility breaking release.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to