hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Weiwei Yang (JIRA)" <j...@apache.org>
Subject [jira] [Comment Edited] (YARN-7875) AttributeStore for store and recover attributes
Date Tue, 06 Feb 2018 12:14:00 GMT

    [ https://issues.apache.org/jira/browse/YARN-7875?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16353787#comment-16353787
] 

Weiwei Yang edited comment on YARN-7875 at 2/6/18 12:13 PM:
------------------------------------------------------------

Hi [~bibinchundatt]

Thanks for working on this JIRA, apologies that I got some delay and was just able to take
a look at the patch. I got some high level comments. I started with a comparison between {{FileSystemNodeLabelsStore}}
and {{FileSystemNodeAttributeStore}}, but I found there are still too many code overlapping,
if you look at the these methods
* getDefaultFSNodeLabelsRootDir
* init
* close
* initSchema
* updateNodeToLabelsMappings vs replaceNodeAttributes
* storeNewClusterNodeLabels vs addNodeAttributes
* removeClusterNodeLabels vs removeNodeAttributes
* ...

their logic are just too similar except some minor difference on some constant values, configuration
properties etc. I can't see how much this refactor work helps. If we really want to refactor
the class to be more generic so labels and attributes store can share most of code, we need
a cleaner hierarchy that now. Does that make sense to you? 

Thanks


was (Author: cheersyang):
Hi [~bibinchundatt]

Thanks for working on this JIRA, apologies that I got some delay and was just able to take
a look at the patch. I got some high level comments. I started with a comparison between {{FileSystemNodeLabelsStore}}
and {{FileSystemNodeAttributeStore}}, but I found there are still too many code overlapping,
if you look at the these methods
* getDefaultFSNodeLabelsRootDir
* init
* close
* initSchema
* updateNodeToLabelsMappings vs replaceNodeAttributes
* storeNewClusterNodeLabels vs addNodeAttributes
* removeClusterNodeLabels vs removeNodeAttributes
* ...
their logic are just too similar except some minor difference on some constant values, configuration
properties etc. I can't see how much this refactor work helps. If we really want to refactor
the class to be more generic so labels and attributes store can share most of code, we need
a cleaner hierarchy that now. Does that make sense to you? 

Thanks

> AttributeStore for store and recover attributes
> -----------------------------------------------
>
>                 Key: YARN-7875
>                 URL: https://issues.apache.org/jira/browse/YARN-7875
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>            Reporter: Bibin A Chundatt
>            Assignee: Bibin A Chundatt
>            Priority: Major
>         Attachments: YARN-7875-WIP.patch
>
>
> Similar to NodeLabelStore need to support NodeAttributeStore for persisting attributes
mapping to Nodes.



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

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


Mime
View raw message