actually this commit doesn't help ( but I don't think it causes anything either ),
I will rollback this commit
the idea was :
from the profiler, I see all CustomRunner instances ( one test class each ) live in JVM
and not get GCd at all, as well as thousands of ExtendedFrameworkMethod instance ( one
test case each ), which takes lots of memory.
So, this commit just intended to free these objects after each test run, but I missed it
is also referenced by org.junit.runners.ParentRunner#fFilteredChildren, which means, if
the Runner is not GCd, then these FrameworkMethods will not be GCd
On Apr 25, 2013, at 2:41 AM, Steve Ebersole <steve(a)hibernate.org> wrote:
This seems to be tied to the commits for
https://hibernate.atlassian.net/browse/HHH-8127
Strong, these were your commits. Can you give us any insight here?
The
ci.hibernate.org tests were passing before your work :
http://ci.hibernate.org/job/hibernate-orm-master/52/
They started failing after your commit, and have not completed since :
http://ci.hibernate.org/job/hibernate-orm-master/53/
On Tue 23 Apr 2013 06:10:10 PM CDT, Steve Ebersole wrote:
> Any other thoughts one what to do here?
>
> Sanne, not sure if you saw my earlier response, but heap size is
> already set to 1024m, not 2048m. The only change I pulled over from
> that was to add jvmArgs '-XX:MaxPermSize=256m' as well. Do you want
> me to back that out as well?
>
>
> On Mon 22 Apr 2013 05:01:08 PM CDT, Steve Ebersole wrote:
>> You mean for tests? Right now I see:
>>
>> test {
>> ...
>> maxHeapSize = "1024m"
>> }
>>
>>
>> On Sat 20 Apr 2013 02:20:56 PM CDT, Sanne Grinovero wrote:
>>> Hi Steve,
>>> from the Jenkins console I couldn't figure it out either; I looked on
>>> the server and it seems the system kernel killed a java process
>>> because it needed more memory.
>>>
>>> The system has plenty of memory available now that no tests are
>>> running (about 1700MB), but from system logs it seems it regularly
>>> kills some java ORM process; this is not a regular out-of-heap size so
>>> it might be worth a short to revert the gradle memory options from
>>>
>>> maxHeapSize = "2048m"
>>>
>>> to
>>>
>>> maxHeapSize = "1024m"
>>>
>>> I had applied that change myself when we had the OOMs before because
>>> of the leak we profiled ...
>>> Also I'll keep in mind at the next machine refresh that we might want
>>> more memory.
>>>
>>> Cheers,
>>> Sanne
>>>
>>>
>>> On 20 April 2013 17:24, Steve Ebersole <steve(a)hibernate.org> wrote:
>>>> Can someone tell me why
>>>>
http://ci.hibernate.org/job/hibernate-orm-master/48/ is showing as
>>>> failing?
>>>>
>>>> I looked in console and test output and see nothing (although it only
>>>> runs the hibernate-c3p0 tests).
>>>> _______________________________________________
>>>> hibernate-dev mailing list
>>>> hibernate-dev(a)lists.jboss.org
>>>>
https://lists.jboss.org/mailman/listinfo/hibernate-dev
_______________________________________________
hibernate-dev mailing list
hibernate-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/hibernate-dev
-------------------------
Best Regards,
Strong Liu <stliu at hibernate.org>
http://about.me/stliu/bio