You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@tez.apache.org by "Siddharth Seth (JIRA)" <ji...@apache.org> on 2017/05/02 01:07:04 UTC
[jira] [Commented] (TEZ-3705) Modify DeletionTracker and deletion
threads to be initialized only if enabled for tez_shuffle
[ https://issues.apache.org/jira/browse/TEZ-3705?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15991930#comment-15991930 ]
Siddharth Seth commented on TEZ-3705:
-------------------------------------
{code}if (ShuffleUtils.isTezShuffleHandler(getConfig())) {{code}
Not sure why this condition is required in ContainerLauncherManager?
Is there an expectation for multiple deletion trackers to exist? IAC, I think it makes sense to move the thread creation into the Impl itself, instead of having the base class create threads. Let individual implementations create whatever threads they need.
Can we rename dagDelete - to something like tez.dag.delete.intermediate.data.on.completion (or a slightly shorter version of the same). tez.dag.delete can be confusing. (Also the parameter names).
Otherwise, looks good to me.
> Modify DeletionTracker and deletion threads to be initialized only if enabled for tez_shuffle
> ---------------------------------------------------------------------------------------------
>
> Key: TEZ-3705
> URL: https://issues.apache.org/jira/browse/TEZ-3705
> Project: Apache Tez
> Issue Type: Sub-task
> Reporter: Kuhu Shukla
> Assignee: Kuhu Shukla
> Attachments: TEZ-3705.001.patch, TEZ-3705.002.patch, TEZ-3705.003.patch
>
>
> Right now the Deletion Tracker and the associated threads are initialized irrespective of the deletion service being used/enabled or not. This can be made better.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)