Valid point, Donald.

I had another look at Derby manual [1] and I think we are quite safe if we also 
specify derby.locks.deadlockTimeout [2] and make it less than 
derby.locks.waitTimeout. This should assure we get deadlocks and timeouts as 
before but faster.

Regards,
Milosz

[1] 
http://db.apache.org/derby/docs/10.2/devguide/devguide-single.html#cdevconcepts28436
[2] 
http://db.apache.org/derby/docs/10.2/tuning/tuning-single.html#rtunproper10607


> The tests may not break, but the results could be completely different 
> and passing for the wrong reason.
> 
> Seems that we should also set derby.locks.deadlockTrace
>     http://db.apache.org/derby/docs/10.0/manuals/tuning/perf79.html
> and review the trace results, since a lot of the tests just catch an 
> exception today as the expected test result and don't examine it to 
> ensure it was the expected type, given a locktimeout will throw a 
> SQLException.
>     http://db.apache.org/derby/docs/10.0/manuals/tuning/perf82.html
> 
> 
> -Donald
> 
> 
> Miłosz Tylenda wrote:
> > Hi All!
> > 
> > Have anybody experimented with setting default Derby timeout in our test 
> > suite? I have set the timeout to 5 seconds and the test suite execution 
> > time felt down from 34 to 19 minutes on my system. Part of my 
> > openjpa-persistence-jdbc pom.xml looks like this:
> > 
> >                     
> >                         
> >                             openjpa.Log
> >                             DefaultLevel=${openjpa.loglevel}
> >                         
> >                         
> >                             openjpa.ConnectionDriverName
> >                             org.apache.commons.dbcp.BasicDataSource
> >                         
> >                         
> >                             derby.stream.error.file
> >                             target/derby.log
> >                         
> >                         
> >                             derby.locks.waitTimeout
> >                             5
> >                         
> >                         
> >                             openjpa.ConnectionProperties
> >                             
> > DriverClassName=${connection.driver.name},Url=${connection.url},Username=${connection.username},Password=${connection.password},${dbcp.args}
> >                         
> >                     
> > 
> > We could consider setting this permanently. Hope it will not affect 
> > timeout-related tests.
> > 
> > Greetings,
> > Milosz
> > 
> > 
> 

Reply via email to