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

[Manik Surtani]

On 7 Dec 2009, at 15:50, Juha Heljoranta wrote:

> Hi,
> 
> Thanks for clarifying. The eviction policy seems to make sense.
> 
> I still get test failure when running a test with L1 cache disabled. I added 
> an assertion check for the cache put/get operations to detect timeout. And 
> this is what I'm constantly getting:
> java.lang.AssertionError: Cache get took 15002 milliseconds (timeout=2000) 
> key=L1DisabledAsymmetricClusterTest-28.
> 
> Could you try to re-run the test? I've enclosed an improved version of the 
> patch. Also, if you still cannot reproduce the issue, please try to increase 
> cacheEntriesNum in AbstractAsymmetricClusterTest.

I too have changed the patch a bit to work better in the test framework - care to have a look and merge with your changes?

-------------- next part --------------
A non-text attachment was scrubbed...
Name: asymmetric1.patch.gz
Type: application/x-gzip
Size: 1962 bytes
Desc: not available
Url : http://lists.jboss.org/pipermail/infinispan-dev/attachments/20091207/cf287505/attachment-0002.gz 
-------------- next part --------------

> In case increasing of the cacheEntriesNum doesn't affect it might help to 
> change TestCacheManagerFactory.createClusteredCacheManager() to 
> GlobalConfiguration.getClusteredDefault() in AbstractAsymmetricClusterTest 
> temporarily and see does it make any difference.
> 
> I'm running the test via Eclipse TestNG plugin. 
> The JVM options are:
> -Dbind.address=127.0.0.1 -Djava.net.preferIPv4Stack=true
> My development machine is a few years old laptop with dual core processor and 
> 4 GiB ram. The Java version is:
> java version "1.6.0_0"
> OpenJDK Runtime Environment (IcedTea6 1.6) (fedora-30.b16.fc11-x86_64)
> OpenJDK 64-Bit Server VM (build 14.0-b16, mixed mode)
> 
> Regards,
> Juha Heljoranta
> 
> 
> On Monday 07 December 2009 16:40:39 Manik Surtani wrote:
>> Hi Juha
>> 
>> I had a look, essentially it is only the 2 eviction-enabled tests that
>> fail.  And this is because you do not have a cache store configured.  The
>> eviction thread will kick in and pull stuff out of memory - and at this
>> point you will need to have this overflow into a cache store.  When using
>> DIST and the key is mapped to the local node, the DistributionManager will
>> *not* attempt a remote lookup as the key is assumed to be available
>> locally (either in memory or in a cache store).
> <asymmetric-v2.patch>

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