[ 
https://issues.apache.org/jira/browse/IGNITE-5986?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16119944#comment-16119944
 ] 

Andrew Mashenkov edited comment on IGNITE-5986 at 8/9/17 2:06 PM:
------------------------------------------------------------------

Patch prevent node failure on stop attached.
But we still have an issue with query stopping and index closing order.


was (Author: amashenkov):
Patch prevent node failure on stop.
But we still have an issue with query stopping and index closing order.

> Fix failed test in .NET suite.
> ------------------------------
>
>                 Key: IGNITE-5986
>                 URL: https://issues.apache.org/jira/browse/IGNITE-5986
>             Project: Ignite
>          Issue Type: Test
>          Components: cache, platforms
>    Affects Versions: 2.1
>            Reporter: Andrew Mashenkov
>            Assignee: Andrew Mashenkov
>              Labels: MakeTeamcityGreenAgain
>             Fix For: 2.2
>
>         Attachments: Fix_Lucene_index_closing_bug_.patch
>
>
> Preface:
> I've found we use closeQuiet() method in LuceneIndex.close() that suppress 
> exceptions. 
> So, this is why we have no issues here before.
> I'd think when index is closing, it is expected that all queries being 
> cancelled and at the moment all QueryCursors (which use IndexReaders) has 
> been closed already.
> But I' observe it is not true in some cases.
> Looking at GridCacheProcessor.stopCache() method we call onCacheStop()-> 
> unregisterCache -> table.onDrop() -> luceneIdx.close().
> Looks like someone forget to close Cursor or\and Cursor wasn't closed during 
> query cancelation.
> We have to investigate this and 



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to