[jira] [Commented] (HIVE-12999) Tez: Vertex creation is slowed down when NN throttles IPCs

["Gopal V (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-12999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15131681#comment-15131681 ] 

Gopal V commented on HIVE-12999:
--------------------------------

Made sure that we don't call mkdirs() for each input path, instead it's de-duped to once per alias.

> Tez: Vertex creation is slowed down when NN throttles IPCs
> ----------------------------------------------------------
>
>                 Key: HIVE-12999
>                 URL: https://issues.apache.org/jira/browse/HIVE-12999
>             Project: Hive
>          Issue Type: Bug
>          Components: Tez
>    Affects Versions: 1.2.0, 1.3.0, 2.0.0, 2.1.0
>            Reporter: Gopal V
>            Assignee: Gopal V
>         Attachments: HIVE-12999.1.patch
>
>
> Tez vertex building has a decidedly slow path in the code, which is not related to the DAG plan at all.
> The total number of RPC calls is not related to the total number of operators, due to a bug in the DagUtils inner loops.
> {code}
> 	at org.apache.hadoop.fs.FileSystem.mkdirs(FileSystem.java:1877)
> 	at org.apache.hadoop.hive.ql.exec.Utilities.createTmpDirs(Utilities.java:3207)
> 	at org.apache.hadoop.hive.ql.exec.Utilities.createTmpDirs(Utilities.java:3170)
> 	at org.apache.hadoop.hive.ql.exec.tez.DagUtils.createVertex(DagUtils.java:548)
> 	at org.apache.hadoop.hive.ql.exec.tez.DagUtils.createVertex(DagUtils.java:1151)
> 	at org.apache.hadoop.hive.ql.exec.tez.TezTask.build(TezTask.java:388)
> 	at org.apache.hadoop.hive.ql.exec.tez.TezTask.execute(TezTask.java:175)
> {code}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)