Skip to content

Conversation

@guowei2
Copy link
Contributor

@guowei2 guowei2 commented Aug 12, 2014

it seems that set command does not run by SparkSQLDriver. it runs on hive api.
user can not change reduce number by setting spark.sql.shuffle.partitions

but i think setting hive properties seems just a role to spark sql.

@AmplabJenkins
Copy link

Can one of the admins verify this patch?

@marmbrus
Copy link
Contributor

Why do you want use hive code for configuration? We have our own configuration that already mirrors into HiveConf.

@guowei2
Copy link
Contributor Author

guowei2 commented Aug 13, 2014

what i want is to use our own configuration. but it use hive code for configuration now.

@chenghao-intel
Copy link
Contributor

+@liancheng, @marmbrus I think this is good catch.SparkSQLCLIDriver will resort to SetProcessor(of Hive) for all of the set command currently, but with this PR, it will goes into the SparkCliDriver, and eventually resort to SetCommand (of SparkSQL).

@guowei2 , can you also add a unit test for that?

@liancheng
Copy link
Contributor

@guowei2 Are you sure set spark.sql.shuffle.partitions=n doesn't work? Would you mind to provide steps to reproduce this issue? 'Cause I just tried it with the most recent master branch (HiveThriftServer2 + beeline):

sparksql lian laptop local spark stages

Stage 0 and stage 3 indicate that the set command works.

@guowei2
Copy link
Contributor Author

guowei2 commented Aug 13, 2014

@liancheng it is ok on (HiveThriftServer2 + beeline).
it doesn't work on bin/spark-sql. so does on my develop IDE with running SparkSQLCLIDriver

@liancheng
Copy link
Contributor

Oh I see where the problem is. So in the current master branch, SET commands never get executed by Spark SQL since SetProcessor is not intercepted in SparkSQLCLIDriver and passed to Spark SQL. Thanks @guowei2 @chenghao-intel!

@liancheng
Copy link
Contributor

test this please

1 similar comment
@marmbrus
Copy link
Contributor

test this please

@SparkQA
Copy link

SparkQA commented Aug 13, 2014

QA tests have started for PR 1904. This patch merges cleanly.
View progress: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/18452/consoleFull

@SparkQA
Copy link

SparkQA commented Aug 13, 2014

QA results for PR 1904:
- This patch FAILED unit tests.
- This patch merges cleanly
- This patch adds no public classes

For more information see test ouptut:
https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/18452/consoleFull

@marmbrus
Copy link
Contributor

Jenkins, test this please.

@SparkQA
Copy link

SparkQA commented Aug 13, 2014

QA tests have started for PR 1904. This patch merges cleanly.
View progress: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/18490/consoleFull

@SparkQA
Copy link

SparkQA commented Aug 14, 2014

QA results for PR 1904:
- This patch PASSES unit tests.
- This patch merges cleanly
- This patch adds no public classes

For more information see test ouptut:
https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/18490/consoleFull

@marmbrus
Copy link
Contributor

Thanks! I've merged this to master and 1.1

asfgit pushed a commit that referenced this pull request Aug 14, 2014
it seems that set command does not run by SparkSQLDriver. it runs on hive api.
user can not change reduce number by setting spark.sql.shuffle.partitions

but i think setting hive properties seems just a role to spark sql.

Author: guowei <[email protected]>

Closes #1904 from guowei2/temp-branch and squashes the following commits:

7d47dde [guowei] fixed: setting properties like spark.sql.shuffle.partitions does not effective

(cherry picked from commit 63d6777)
Signed-off-by: Michael Armbrust <[email protected]>
@asfgit asfgit closed this in 63d6777 Aug 14, 2014
xiliu82 pushed a commit to xiliu82/spark that referenced this pull request Sep 4, 2014
it seems that set command does not run by SparkSQLDriver. it runs on hive api.
user can not change reduce number by setting spark.sql.shuffle.partitions

but i think setting hive properties seems just a role to spark sql.

Author: guowei <[email protected]>

Closes apache#1904 from guowei2/temp-branch and squashes the following commits:

7d47dde [guowei] fixed: setting properties like spark.sql.shuffle.partitions does not effective
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants