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

Posted to common-issues@hadoop.apache.org by "Steve Loughran (JIRA)" <ji...@apache.org> on 2012/07/28 02:19:34 UTC

[jira] [Created] (HADOOP-8629) Add option for TableMapping to reload mapping file on match failure

Steve Loughran created HADOOP-8629:
--------------------------------------

             Summary: Add option for TableMapping to reload mapping file on match failure
                 Key: HADOOP-8629
                 URL: https://issues.apache.org/jira/browse/HADOOP-8629
             Project: Hadoop Common
          Issue Type: Improvement
          Components: util
            Reporter: Steve Loughran
            Priority: Minor


As commented in HADOOP-7030, the table mapping topology mapper handles new node addition worse than the script mapping, because the table mapping is frozen for the life of the service.

I propose adding an option (true by default?) for the class to look at the timestamp of the mapping file, and reload it on a change -if a hostname lookup failed. 

 

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HADOOP-8629) Add option for TableMapping to reload mapping file on match failure

Posted by "Steve Loughran (JIRA)" <ji...@apache.org>.

    [ https://issues.apache.org/jira/browse/HADOOP-8629?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13424230#comment-13424230 ] 

Steve Loughran commented on HADOOP-8629:
----------------------------------------

Auto reload of the table mapping is not completely straightforward as
* need to make upload thread safe.
* reload logic needs policy on failure to reload the file. Retain the old mapping?
* reload logic needs policy on removal of entries from the file and changed values. With caching in front of the class, these changes won't get picked up anyway, but if we strip that out by not caching in-VM lookups, then it may become an issue. 
                
> Add option for TableMapping to reload mapping file on match failure
> -------------------------------------------------------------------
>
>                 Key: HADOOP-8629
>                 URL: https://issues.apache.org/jira/browse/HADOOP-8629
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: util
>            Reporter: Steve Loughran
>            Priority: Minor
>
> As commented in HADOOP-7030, the table mapping topology mapper handles new node addition worse than the script mapping, because the table mapping is frozen for the life of the service.
> I propose adding an option (true by default?) for the class to look at the timestamp of the mapping file, and reload it on a change -if a hostname lookup failed. 
>  

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

[jira] [Commented] (HADOOP-8629) Add option for TableMapping to reload mapping file on match failure

Posted by "liang xie (JIRA)" <ji...@apache.org>.

    [ https://issues.apache.org/jira/browse/HADOOP-8629?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13502997#comment-13502997 ] 

liang xie commented on HADOOP-8629:
-----------------------------------

I prefer to provide an admin cmd to reload by administrator
                
> Add option for TableMapping to reload mapping file on match failure
> -------------------------------------------------------------------
>
>                 Key: HADOOP-8629
>                 URL: https://issues.apache.org/jira/browse/HADOOP-8629
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: util
>            Reporter: Steve Loughran
>            Priority: Minor
>
> As commented in HADOOP-7030, the table mapping topology mapper handles new node addition worse than the script mapping, because the table mapping is frozen for the life of the service.
> I propose adding an option (true by default?) for the class to look at the timestamp of the mapping file, and reload it on a change -if a hostname lookup failed. 
>  

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira