[jira] [Commented] (SPARK-11278) PageRank fails with unified memory manager

["Vivek Gupta (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SPARK-11278?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15614954#comment-15614954 ] 

Vivek Gupta commented on SPARK-11278:
-------------------------------------

We are facing an issue with Spark 1.6.0 whereby performance degrades severely  (with extensive Shuffle spill) using the Unified Memory Manager (spark.memory.fraction = 0.9 and spark.memory.storageFraction = 0.0).

However same application works with improved performance having switched to legacy mode (spark.memory.useLegacyMode=true, spark.shuffle.memoryFraction=0.9, spark.storage.memoryFraction=0.0).

Is this something related with this issue?

> PageRank fails with unified memory manager
> ------------------------------------------
>
>                 Key: SPARK-11278
>                 URL: https://issues.apache.org/jira/browse/SPARK-11278
>             Project: Spark
>          Issue Type: Bug
>          Components: GraphX, Spark Core
>    Affects Versions: 1.5.1
>            Reporter: Nishkam Ravi
>            Assignee: Andrew Or
>            Priority: Critical
>         Attachments: executor_log_legacyModeTrue.html, executor_logs_legacyModeFalse.html
>
>
> PageRank (6-nodes, 32GB input) runs very slow and eventually fails with ExecutorLostFailure. Traced it back to the 'unified memory manager' commit from Oct 13th. Took a quick look at the code and couldn't see the problem (changes look pretty good). cc'ing [~andrewor14][~vanzin] who may be able to spot the problem quickly. Can be reproduced by running PageRank on a large enough input dataset if needed. Sorry for not being of much help here.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org