[hibernate-dev] Analyzing the ORM testsuite

Thu Apr 19 18:32:04 EDT 2018

On 19 April 2018 at 22:50, Chris Cranford <chris at hibernate.org> wrote:
> The entire build process under OpenJDK 8 here:
>
> BUILD SUCCESSFUL in 47m 26s
> 251 actionable tasks: 233 executed, 18 up-to-date
>
> The only thing I noticed was that while executing the hibernate-core tests,
> the memory consumption seemed quite high toward the end of the test suite,
> particularly ~3.5GB memory consumed just to run the hibernate-core tests
> alone and that FindBugs complained about some magic numbers on some zip
> files while inspecting hibernate-osgi.
>
> Any idea if any substantial amount of that memory consumption related to the
> test framework tracking the output/errors of those executed tests?

I guess that confirms there's something specific to my environment.
That's a great hint to have, thanks!

Regarding the memory, yes I'm seeing the same. >90% of that is
Mockito, I'm exploring some ideas to reduce its memory usage; I have
some working experiments but I'm not happy with them: they make
maintenance of tests error prone so tomorrow I'll try some
alternatives.

>
> On 04/19/2018 05:20 PM, Sanne Grinovero wrote:
>
> On 19 April 2018 at 21:47, Chris Cranford <chris at hibernate.org> wrote:
>
> Sanne -
>
> Are you running the build task or are you executing other more specific
> tasks and noticing the slowness?
>
> Just running "gradle clean build" from the root. I had the same
> results with OpenJDK 8 and 9, twice. Getting exactly the same failures
> as before.
>
> Thanks!
>
> On 04/19/2018 04:25 PM, Steve Ebersole wrote:
>
> Nothing off the top of my head.  The build time is in normal range on my
> machine as of this morning (8 - 10 minutes).
>
> You mentioned a specific commit.  Is that just when you noticed a change,
> or do you suspect something in that commit?
>
> I'm not at computer right now, but when I get back I will check out that
> commit and see if anything is strange locally.
>
>
>
> On Thu, Apr 19, 2018, 3:11 PM Sanne Grinovero <sanne at hibernate.org> wrote:
>
> Hi all,
>
> I'm trying to analyze the memory usage of current master of Hibernate
> ORM; I started looking because of OOM errors on the PostgreSQL
> testsuite (as mentioned in another thread).
>
> So far I've seen that Mockito is allocating a "lot of stuff"; I might
> be able to improve some things in that area but it's a distraction
> from my goal as clearly we use Mockito even when not running the
> PostgreSQL backed tests and that seems to run fine. But it's polluting
> the data reports so making it hard (and slow!) to verify if there is a
> real issue.
>
> I could use some help to verify some tangential issues though; when
> running the testsuite locally it's taking much longer than what I've
> been used to see for the same testsuite some weeks ago, and I have
> some failures - regularly seeing the same failures.
>
> These are triggered by running a default build, on H2:
>
> Task :hibernate-core:test
>
> org.hibernate.test.criteria.CriteriaLockingTest >
>
> testSetLockModeDifferentFromNONELogAWarnMessageWhenTheDialectUseFollowOnLockingIsTrue
> FAILED
>     org.junit.runners.model.TestTimedOutException
>
> org.hibernate.test.criteria.CriteriaLockingTest >
>
> testSetLockModeNONEDoNotLogAWarnMessageWhenTheDialectUseFollowOnLockingIsTrue
> FAILED
>     java.lang.Exception
>
> org.hibernate.test.criteria.CriteriaLockingTest >
> org.hibernate.test.criteria.CriteriaLockingTest FAILED
>     java.lang.Exception
>
> org.hibernate.test.locking.warning.LockNoneWarmingTest >
> org.hibernate.test.locking.warning.LockNoneWarmingTest FAILED
>     java.lang.Exception
>
> 6307 tests completed, 4 failed, 327 skipped
>
> Yet the puzzling point: on ci.hibernate.org I don't see these failures
> happening, nor the testsuite time seems to have increased of any
> significant amount (it regularly completes in somewhere between 15 and
> 20 minutes). It takes me more than an hour to get this result on a
> machine which would normally complete the testsuite in 8 minutes, and
> I got the same 4 failures when re-running it a second time.
>
> Could other developers also please checkout master - specifically at
> commit 291d4a3eeaa2ade32a42cfbcad5868b2114c34fe - and let me know if
> it works allright for you and if the build time is consistent with the
> usual times for your machine? I need to understand what might be
> different between my workstation and the CI server.
>
> I might even have some great improvements ready to reduce the
> testsuite time, but I'm not comfortable in committing them in this
> context.
>
> Thanks,
> Sanne
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>
>
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/hibernate-dev
>
>