[infinispan-dev] [patch] Unit test to highlight potential performance and cache eviction issues in distributed mode

[Manik Surtani]

Hi Juha

I saw the JIRA you created, thanks for this - I'll take a closer look this week.

Cheers
Manik

On 17 Dec 2009, at 12:52, Juha Heljoranta wrote:

> Hi,
> 
> In case the my previous mail didn't get through I'd still hope to get my patch 
> getting merged.
> 
> The test should highlight an issue where cache.get() "fails" by taking 15 
> second to execute.
> 
> When I enable trace level logging this here is what jgroups says:
> 14:52:36.206 [main] TRACE org.jgroups.blocks.GroupRequest - call did not 
> execute correctly, request is [req_id=1260622341191
> caller=nine-22069
> entries:
> nine-12394: sender=nine-12394, retval=null, received=false, suspected=false
> nine-39772: sender=nine-39772, retval=SuccessfulResponse, received=true, 
> suspected=false
> 
> I've tried to locate where things go wrong but failed to nail it down. I 
> suspect that either there is some sort of race condition in infinispan or the 
> bug is in jgroups.
> 
> Regards,
> Juha Heljoranta
> 
> 
> On Friday 11 December 2009 21:20:34 Juha Heljoranta wrote:
>> Hi,
>> 
>> I improved the patch slightly. The tests pass now just fine (tested with
>> trunk r1296).
>> 
>> Test suite will now print a warning if cache put()/get() takes longer than
>> 2 s and fails if they take longer than 20 s. I usually see few warnings
>> about get() taking 15 s, an issue which might warrant some investigating.
>> 
>> Could you apply the patch?
>> 
>> Regards,
>> Juha Heljoranta
>> 

--
Manik Surtani
manik at jboss.org
Lead, Infinispan
Lead, JBoss Cache
http://www.infinispan.org
http://www.jbosscache.org