[infinispan-dev] The reason(s) why we have so many threads are spawned in the testsuite

[Galder Zamarreno]

On 10/12/2009 11:29 AM, Manik Surtani wrote:
>
> On 9 Oct 2009, at 16:57, Galder Zamarreno wrote:
>
>>
>>
>> On 10/09/2009 05:49 PM, Vladimir Blagojevic wrote:
>>> On 09-10-09 11:32 AM, Galder Zamarreno wrote:
>>>>
>>>> On 10/09/2009 04:19 AM, Vladimir Blagojevic wrote:
>>>>
>>>>> Galder,
>>>>>
>>>>> Under profiler none of these seem to be a big problem except
>>>>> asyncTransport TPE. All of these threads that you saw in thread
>>>>> dumps
>>>>> are related to rehash requests spawned by LeaveTask. LeaveTask
>>>>> tries to
>>>>> submit request to TPE and is block waiting. All of these threads
>>>>> are. I
>>>>> am still trying to figure out what is going.
>>>>>
>>>> Block waiting where?
>>>>
>>> ExecutorService#submit(Callable)<---- profiler shows thread is
>>> blocked here
>>> RPCManagerImpl#invokeRemotelyInFuture
>
> This can't be about the pool size being too small.  I'll have a look,
> I'm going to be revamping the dist tests this week anyway.

If using a TPE, I don't see anything in AbstractExecutorService.submit() 
(that's the impl used by TPE) that could block, so that what's I 
suggested the pool size thing. Having had a deeper look, rather than 
blocking, the execution would be rejected if the worker count goes over 
the max, so that wouldn't block.

-- 
Galder Zamarreño
Sr. Software Engineer
Infinispan, JBoss Cache