[jira] [Commented] (ACCUMULO-542) with a large root tablet, tablet servers will get stuck loading other metadata tablets

["Keith Turner (Commented) (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ACCUMULO-542?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13255969#comment-13255969 ] 

Keith Turner commented on ACCUMULO-542:
---------------------------------------

For 1.5 we may want to consider switching the old code the reads directly from the root tablet to user a scanner.  Also we noticed the code was reading everything in the root tablet after the tablet of interest.  This is completely unnecessary.  
                
> with a large root tablet, tablet servers will get stuck loading other metadata tablets
> --------------------------------------------------------------------------------------
>
>                 Key: ACCUMULO-542
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-542
>             Project: Accumulo
>          Issue Type: Bug
>          Components: tserver
>    Affects Versions: 1.4.0, 1.3.5
>         Environment: large cluster with 390+ metadata data tablets
>            Reporter: Eric Newton
>            Assignee: Eric Newton
>            Priority: Critical
>             Fix For: 1.3.6
>
>
> Trying to bring up a large cluster, metadata tablets failed to load. These metadata tablets were trying to read their entries from the root tablet.
> Workaround is to double the memory used for scanning the !METADATA table:
> {noformat}
> shell> config -t !METADATA -s table.scan.max.memory=2M
> {noformat}
> Nothing needs to be done: the tablet servers will recover as soon as this value is changed.

--
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