You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@tez.apache.org by "Chenren Shao (JIRA)" <ji...@apache.org> on 2019/04/03 02:21:00 UTC

[jira] [Created] (TEZ-4059) TezSplitGrouper does respect tez.grouping.max-size if the originalSplit result in bigger splits

Chenren Shao created TEZ-4059:
---------------------------------

             Summary: TezSplitGrouper does respect tez.grouping.max-size if the originalSplit result in bigger splits
                 Key: TEZ-4059
                 URL: https://issues.apache.org/jira/browse/TEZ-4059
             Project: Apache Tez
          Issue Type: Bug
    Affects Versions: 0.9.1
         Environment: Tez 0.9.1, Hive 2.3.4, Hadoop 2.8.5 (EMR 5.20.0),
            Reporter: Chenren Shao
         Attachments: split-generation-log.txt

In using Tez 0.9.1, Hive 2.3.4, Hadoop 2.8.5 (EMR 5.20.0), I found if we have below scenario, TezSplitGrouper is unable to respect tez.grouping.max-size.

On line 238, Let us say, a customer-defined inputformat generate a M number of splits, but its split size is greater than tez.grouping.max-size, then line 248 is executed, and a newDesiredNumSplits is calculated based on tez.grouping.max-size, N. (N is greater than M), desiredNumSplits is set to be N.

However, when it goes to line 273, since the N is greater than M now, it executes line 274, it uses originalSplits anyway and basically ignores the M setting.

 



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