This could also be caused by recent OOMs we've been seeing with the Sun JDK on Linux. Have you been seeing any junit runs that fail with a GC/OOM error? Are you still seeing a slowdown? I'm trying to resolve the OOM problems, but it's slow going....
-Donald On 10/2/10 10:48 PM, Donald Woods wrote: > I'm not seeing any slowdown with our nightly builds on Linux. > > Trunk - Derby w/ IBM 6 SDK > Oct 1 - 45m:14s > Sept 24 - 52m:33s > > The changes for DBCP started going in on Sept. 28 in r1002321. > > If you happened to pickup r1003657, then svn up to the latest to revert > the temp enhancer script changes. > > > -Donald > > > On 10/2/10 7:05 AM, Miłosz Tylenda wrote: >> Hi all, >> >> I noticed the JDBC tests got much slower recently. When I compare the times >> with logs I have from last days of August (unfortunately I don't have any >> from September), the slowdown is around 3x. A similar slowdown is visible on >> Hudson builds too if you compare 2.0 and 2.1 builds. >> >> A log message which now is >> >> openjpa.jdbc.JDBC - Using dictionary class >> "org.apache.openjpa.jdbc.sql.DerbyDictionary". >> >> used to be >> >> openjpa.jdbc.JDBC - Using dictionary class >> "org.apache.openjpa.jdbc.sql.DerbyDictionary" (Apache Derby 10.5.3.0 - (8029 >> 17) ,Apache Derby Embedded JDBC Driver 10.5.3.0 - (802917)). >> >> The direct cause is we now pass a null connection to >> DBDictionaryFactory.newDBDictionary(JDBCConfiguration conf, String dclass, >> String props, Connection conn). I suspect this is related to the recently >> introduced DBCP datasources or JDBC 4 delegates but no luck with findings so >> far. >> >> Does someone hear a ringing bell? Maybe you have test suite logs from >> September and can find out when this regression happened? >> >> Regards, >> Milosz >> >> >
