[ 
https://issues.apache.org/jira/browse/LUCENE-3803?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13211622#comment-13211622
 ] 

Uwe Schindler commented on LUCENE-3803:
---------------------------------------

The problem is that you are using the same fixed thread pool for everything. If 
you only have 1 or 2 threads in the pool (as noted in your mail) and one of 
those threads calls searcher.search() from within its own callable, 
searcher.search will try to execute the sub-searches in its own callables. If 
the thread pool does not have enough free threads, it will wait for a thread to 
get available (because another callable finishes). The callable that called 
searcher.search will of course not finish as it waits for itsself to finish -> 
deadlock.

The bad idea with thread pools that leads to your problem is: Don't start 
Callables from within another Callable in the same thread pool. This almost 
always leads to deadlocks.
                
> Using a searcher with an executor service does not work from within a 
> Callable called by that same executor service
> -------------------------------------------------------------------------------------------------------------------
>
>                 Key: LUCENE-3803
>                 URL: https://issues.apache.org/jira/browse/LUCENE-3803
>             Project: Lucene - Java
>          Issue Type: Bug
>          Components: core/search
>    Affects Versions: 3.5
>            Reporter: Benson Margulies
>         Attachments: lucene-tc.tgz
>
>
> The test case I'm about to attach creates an executor service uses invokeAll 
> to run things from it. The 'things' (Callable) that it runs in turn run 
> Lucene searchers using a searcher that has the very same executor service. 
> This turns out to hit some sort of deadlock related to the executor service 
> queue. 
> I won't be terribly surprised if this is really a matter of a fundamental 
> limitation of the executor service, but if nothing else this JIRA might 
> provoke a warning in the javadoc. Or, what do I know, maybe there's a way 
> Lucene could interact with the executor service that gets along with this?

--
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

        

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to