[jira] [Commented] (ACCUMULO-2535) large cluster failed to shutdown cleanly

["Mike Drob (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ACCUMULO-2535?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13977392#comment-13977392 ] 

Mike Drob commented on ACCUMULO-2535:
-------------------------------------

I think the major compaction should have been interrupted, so I'm not sure what the underlying issue really is.

> large cluster failed to shutdown cleanly
> ----------------------------------------
>
>                 Key: ACCUMULO-2535
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-2535
>             Project: Accumulo
>          Issue Type: Bug
>          Components: tserver
>    Affects Versions: 1.4.4
>         Environment: large cluster
>            Reporter: Eric Newton
>
> A large 1.4.4 cluster failed to shutdown cleanly.  It was stuck trying to unload 7 non-root !METADATA tablets.  Master was sending "unload tablet" messages constantly, but the tserver refused to unload the tablets.
> Examining the tserver logs of one server, it began a full major compaction, then received the unload message, and finished the compaction.  The tablet was never unloaded.
> The cluster had ~400 metadata tablets overall.
> System recovered after a hard shutdown of the remaining 8 servers (7 metadata table servers and 1 root tablet server).



--
This message was sent by Atlassian JIRA
(v6.2#6252)