[
https://issues.apache.org/jira/browse/SPARK-5645?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kay Ousterhout updated SPARK-5645:
----------------------------------
Fix Version/s: 1.3.0
Assignee: Kay Ousterhout (was: Kostas Sakellis)
> Track local bytes read for shuffles - update UI
> -----------------------------------------------
>
> Key: SPARK-5645
> URL: https://issues.apache.org/jira/browse/SPARK-5645
> Project: Spark
> Issue Type: New Feature
> Components: Spark Core
> Reporter: Kostas Sakellis
> Assignee: Kay Ousterhout
> Fix For: 1.3.0
>
>
> Currently we do not track the local bytes read for a shuffle read. The UI
> only shows the remote bytes read. This is pretty confusing to the user
> because:
> 1) In local mode all shuffle reads are local
> 2) the shuffle bytes written from the previous stage might not add up if
> there are some bytes that are read locally on the shuffle read side
> 3) With https://github.com/apache/spark/pull/4067 we display the total number
> of records so that won't line up with only showing the remote bytes read.
> I propose we track the remote and local bytes read separately. In the UI show
> the total bytes read and in brackets show the remote bytes read for a
> shuffle.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
