You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hive.apache.org by Sergio Pena <se...@cloudera.com> on 2016/11/23 20:02:37 UTC
Review Request 54042: HIVE-15114: Remove extra MoveTask operators
from the ConditionalTask
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/54042/
-----------------------------------------------------------
Review request for hive.
Bugs: HIVE-15114
https://issues.apache.org/jira/browse/HIVE-15114
Repository: hive-git
Description
-------
This patch removes an extra MoveTask from the CONDITIONAL TASK added to the query plan. It would be used only as a Blobstorage optimization.
In HDFS, this two tasks are just simple atomic renames, but on Blobstorage, a rename is too expensive.
Diffs (updated)
-----
itests/hive-blobstore/src/test/queries/clientpositive/conditional_task_optimization.q PRE-CREATION
itests/hive-blobstore/src/test/results/clientpositive/conditional_task_optimization.q.out PRE-CREATION
itests/hive-blobstore/src/test/results/clientpositive/insert_into.q.out 00ad136ae40a5ed2689ad8ed2f882bbcd238324d
ql/src/java/org/apache/hadoop/hive/ql/optimizer/GenMapRedUtils.java 88d5afadd690c3dce276b72a37608d30ddeea271
ql/src/test/org/apache/hadoop/hive/ql/optimizer/TestGenMapRedUtilsCreateConditionalTask.java PRE-CREATION
Diff: https://reviews.apache.org/r/54042/diff/
Testing (updated)
-------
# Testa added
mvn test -Dtest=TestBlobstoreCliDriver -Dqfile=conditional_task_optimization.q
# All tests
mvn test -Dtest=TestBlobstoreCliDriver
mvn test -Dtest=TestBlobstoreNegativeCliDriver
Thanks,
Sergio Pena
Re: Review Request 54042: HIVE-15114: Remove extra MoveTask operators
from the ConditionalTask
Posted by Sergio Pena <se...@cloudera.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/54042/
-----------------------------------------------------------
(Updated Nov. 28, 2016, 9:52 p.m.)
Review request for hive.
Bugs: HIVE-15114
https://issues.apache.org/jira/browse/HIVE-15114
Repository: hive-git
Description
-------
This patch removes an extra MoveTask from the CONDITIONAL TASK added to the query plan. It would be used only as a Blobstorage optimization.
In HDFS, this two tasks are just simple atomic renames, but on Blobstorage, a rename is too expensive.
Diffs (updated)
-----
itests/hive-blobstore/src/test/queries/clientpositive/conditional_task_optimization.q PRE-CREATION
itests/hive-blobstore/src/test/results/clientpositive/conditional_task_optimization.q.out PRE-CREATION
itests/hive-blobstore/src/test/results/clientpositive/insert_into.q.out 00ad136ae40a5ed2689ad8ed2f882bbcd238324d
ql/src/java/org/apache/hadoop/hive/ql/optimizer/GenMapRedUtils.java 88d5afadd690c3dce276b72a37608d30ddeea271
ql/src/test/org/apache/hadoop/hive/ql/optimizer/TestGenMapRedUtilsCreateConditionalTask.java PRE-CREATION
Diff: https://reviews.apache.org/r/54042/diff/
Testing
-------
# Testa added
mvn test -Dtest=TestBlobstoreCliDriver -Dqfile=conditional_task_optimization.q
# All tests
mvn test -Dtest=TestBlobstoreCliDriver
mvn test -Dtest=TestBlobstoreNegativeCliDriver
Thanks,
Sergio Pena
Re: Review Request 54042: HIVE-15114: Remove extra MoveTask operators
from the ConditionalTask
Posted by Aihua Xu <ax...@cloudera.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/54042/#review156792
-----------------------------------------------------------
ql/src/java/org/apache/hadoop/hive/ql/optimizer/GenMapRedUtils.java (line 1465)
<https://reviews.apache.org/r/54042/#comment227087>
You need to make sure task1 and task2 are both moving file or table. Right now, seems task1 can be file and task2 can be table.
ql/src/java/org/apache/hadoop/hive/ql/optimizer/GenMapRedUtils.java (line 1539)
<https://reviews.apache.org/r/54042/#comment227089>
Seems again you need make sure both are MoveFile or MoveTable.
- Aihua Xu
On Nov. 23, 2016, 8:02 p.m., Sergio Pena wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/54042/
> -----------------------------------------------------------
>
> (Updated Nov. 23, 2016, 8:02 p.m.)
>
>
> Review request for hive.
>
>
> Bugs: HIVE-15114
> https://issues.apache.org/jira/browse/HIVE-15114
>
>
> Repository: hive-git
>
>
> Description
> -------
>
> This patch removes an extra MoveTask from the CONDITIONAL TASK added to the query plan. It would be used only as a Blobstorage optimization.
> In HDFS, this two tasks are just simple atomic renames, but on Blobstorage, a rename is too expensive.
>
>
> Diffs
> -----
>
> itests/hive-blobstore/src/test/queries/clientpositive/conditional_task_optimization.q PRE-CREATION
> itests/hive-blobstore/src/test/results/clientpositive/conditional_task_optimization.q.out PRE-CREATION
> itests/hive-blobstore/src/test/results/clientpositive/insert_into.q.out 00ad136ae40a5ed2689ad8ed2f882bbcd238324d
> ql/src/java/org/apache/hadoop/hive/ql/optimizer/GenMapRedUtils.java 88d5afadd690c3dce276b72a37608d30ddeea271
> ql/src/test/org/apache/hadoop/hive/ql/optimizer/TestGenMapRedUtilsCreateConditionalTask.java PRE-CREATION
>
> Diff: https://reviews.apache.org/r/54042/diff/
>
>
> Testing
> -------
>
> # Testa added
> mvn test -Dtest=TestBlobstoreCliDriver -Dqfile=conditional_task_optimization.q
>
> # All tests
> mvn test -Dtest=TestBlobstoreCliDriver
> mvn test -Dtest=TestBlobstoreNegativeCliDriver
>
>
> Thanks,
>
> Sergio Pena
>
>
Re: Review Request 54042: HIVE-15114: Remove extra MoveTask operators
from the ConditionalTask
Posted by Sergio Pena <se...@cloudera.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/54042/#review156798
-----------------------------------------------------------
ql/src/java/org/apache/hadoop/hive/ql/optimizer/GenMapRedUtils.java (line 1465)
<https://reviews.apache.org/r/54042/#comment227096>
File or table is just how the MoveTask will handle the destination. In both cases, the MoveTask gets the destination directory and it moves the data to it.
In a normal conditional task, MoveTask1 moves to a directory followed by a MoveTask2 that moves to a table directory.
With the patch, MoveTask1 moves to a table directory directly.
Do you see any problem with that merge approach?
- Sergio Pena
On Nov. 23, 2016, 8:02 p.m., Sergio Pena wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/54042/
> -----------------------------------------------------------
>
> (Updated Nov. 23, 2016, 8:02 p.m.)
>
>
> Review request for hive.
>
>
> Bugs: HIVE-15114
> https://issues.apache.org/jira/browse/HIVE-15114
>
>
> Repository: hive-git
>
>
> Description
> -------
>
> This patch removes an extra MoveTask from the CONDITIONAL TASK added to the query plan. It would be used only as a Blobstorage optimization.
> In HDFS, this two tasks are just simple atomic renames, but on Blobstorage, a rename is too expensive.
>
>
> Diffs
> -----
>
> itests/hive-blobstore/src/test/queries/clientpositive/conditional_task_optimization.q PRE-CREATION
> itests/hive-blobstore/src/test/results/clientpositive/conditional_task_optimization.q.out PRE-CREATION
> itests/hive-blobstore/src/test/results/clientpositive/insert_into.q.out 00ad136ae40a5ed2689ad8ed2f882bbcd238324d
> ql/src/java/org/apache/hadoop/hive/ql/optimizer/GenMapRedUtils.java 88d5afadd690c3dce276b72a37608d30ddeea271
> ql/src/test/org/apache/hadoop/hive/ql/optimizer/TestGenMapRedUtilsCreateConditionalTask.java PRE-CREATION
>
> Diff: https://reviews.apache.org/r/54042/diff/
>
>
> Testing
> -------
>
> # Testa added
> mvn test -Dtest=TestBlobstoreCliDriver -Dqfile=conditional_task_optimization.q
>
> # All tests
> mvn test -Dtest=TestBlobstoreCliDriver
> mvn test -Dtest=TestBlobstoreNegativeCliDriver
>
>
> Thanks,
>
> Sergio Pena
>
>