[jboss-dev] Re: JBoss Branch_4_2 testsuite failures

[Brian Stansberry]

Aleksandar Kostadinov wrote:
> Brian Stansberry wrote:
> <snip>
>> These are failures on tests of asynchronous replication. With async the
>> test client doesn't know how long it will take for a session to
>> replicate, so it sleeps 300ms before switching from one server to the
>> other.  There are hundreds of these switches made in the course of a
>> testsuite run. If there's some hiccup during a replication (e.g. a bit
>> longer than normal gc run delays it), the test client can switch before
>> the replicated state arrives and you get a transient failure.
>>
>> I bumped the sleep time a bit (and made it configurable via a system
>> property.)
> 
> Did you documented how to use this system property and where?
> I ask you as I need to document another one ;)

(NOTE: I haven't even tried overriding the default via system property; 
just added the property to the testsuite class with the assumption I'd 
use it at some later point.)

Property is jbosstest.cluster.failover.sleep, value is # of ms the test 
driver thread should sleep.  Gets used just about everywhere in the web 
tier clustering tests.

To set it you basically want to make ant aware of it as an ant property; 
any ant property in the jbosstest namespace gets passed through as a 
system property to the test driver.  So, to your local.properties file, 
try adding:

jbosstest.cluster.failover.sleep=300

I *think* something like this should work as well:

./build.sh -Djbosstest.cluster.failover.sleep=300 <your ant target>


-- 
Brian Stansberry
Lead, AS Clustering
JBoss, a division of Red Hat
brian.stansberry at redhat.com