[Mulgara-dev] Sporadic exception from MulgaraTransactionFactory.acquireMutex

Morten Grouleff mgr at trifork.com
Sun Aug 24 08:32:47 UTC 2008


Thanks for all the advice.

On reviewing your input I realised that I was too trigger-happy sending the original mail. I do not have as many 
InterruptedException cases as I have failures on the client side, so they're not the primary reason I'm getting failures 
in this test scenario. Sorry for that. Actually, I'm unable to reproduce the InterruptedException at all now.

Also forgot to mention that we're running fedora release-3.0.0, build from the HEAD of that svn-tag.

I think we should just let it rest...

Life is hard, and then you die wrote:
> On Fri, Aug 22, 2008 at 02:27:27PM -0700, Life is hard, and then you die wrote:
> [snip]
>>> On Fri, Aug 22, 2008 at 2:57 PM, Morten Grouleff <mgr at trifork.com> wrote:
>>>> I'm unsure if this is a Mulgara or Fedora  problem, but my best bet is Mulgara.
>>>>
>>>> We're using the Mulgara embedded inside Fedora. When running many itql queryes in parallel through the "risearch"
>>>> interface in fedora, we sometimes get this stacktrace in the fedora server log, but only a few times out of the 1849
>>>> calls. The queries are identical, apart from a parameter. If I run the same sequence of queries sequentially, no
>>>> exceptions occur.
>> [snip]
>>>> Caused by: java.lang.InterruptedException
>>>>         at java.lang.Object.wait(Native Method)
>>>>         at org.mulgara.resolver.MulgaraTransactionFactory.acquireMutex(MulgaraTransactionFactory.java:276)
>>>>         ... 47 more
>>>>
>>>> Is it a part of Mulgara that interrupts another thread? Or should
>>>> I look elsewhere for the "interrupting party"?
>> I think the most likely culprit would be the transaction timeouts -
>> when those fire they interrupt the thread if it's holding the
>> write-lock in an attempt knock them out of whatever they're doing.
>> (this actually occurs for any heuristic-rollback and also for a
>> session-close).
>>
>> Check what your current timeouts are (either set explicitly on
>> DatabaseSession or XAResource, or the defaults which usually come from
>> the mulgara-config.xml); and can you get some idea of how long a
>> thread is waiting before it gets interrupted? How many threads are you
>> running?
> 
> A couple things I forgot: 1) if it is timing out, you should see a
> warn-level log message. 2) You might try upping the logging level to
> debug for the following classes and see if that helps narrow things
> down:
>   org.mulgara.resolver.DatabaseSession
>   org.mulgara.resolver.MulgaraTransactionManager
>   org.mulgara.resolver.MulgaraTransactionFactory
>   org.mulgara.resolver.MulgaraInternalTransactionFactory
>   org.mulgara.resolver.MulgaraInternalTransaction
> 
> Oh, and lastly: are you running a verion of Fedora/Trippi with Edwin
> Shin's latest fixes? (see the thread starting at
> http://lists.mulgara.org/pipermail/mulgara-general/2008-July/000467.html)
> The random closing of session's could cause that interrupt too.
> 
> 
>   Cheers,
> 
>   Ronald
> 
> _______________________________________________
> Mulgara-dev mailing list
> Mulgara-dev at mulgara.org
> http://mulgara.org/mailman/listinfo/mulgara-dev

-- 
Mvh.
Morten Grouleff
--------------------------------------------------------------------
Trifork A/S,  Margrethepladsen 4, DK-8000 Aarhus C, Denmark.
JAOO 2008: Sep. 28. - Oct. 3. 2008  http://jaoo.dk/   See you there?



More information about the Mulgara-dev mailing list