Hi, Mark created https://issues.apache.org/jira/browse/SOLR-5022
I think at this time it is hard to find out if it’s a serious production issue or just a test issue. The problem is currently that its hard to reproduce consistently and when it reproduces, the JVM is completely dead, not responding to anything (no kill, no debugger attachment, no jmap,...). The issue appears more often with tests.modifier=3, but this is not really telling you "it’s a test relict". There are no "bad" classloaders involved (I checked the dump, all looks fine, do duplicate classes,...), so it looks like something is creating interned strings all the time, which does not play well with our test infrastructure that runs everything in the same JVM, so interned strings never get cleaned up. A fix would be to create a module for running those tests in good-old one JVM per test suite. Until we know *what and *where* the issue is, I see this as blocker. Uwe ----- Uwe Schindler H.-H.-Meier-Allee 63, D-28213 Bremen http://www.thetaphi.de eMail: [email protected] > -----Original Message----- > From: [email protected] [mailto:[email protected]] On Behalf Of Yonik > Seeley > Sent: Tuesday, July 09, 2013 2:43 PM > To: Lucene/Solr Dev > Subject: Re: 4.4 release planning > > On Tue, Jul 9, 2013 at 7:07 AM, Erick Erickson <[email protected]> > wrote: > > So is the plan to address the Permgen issues before releasing and fix > > it both on the branch and 4x and trunk? > > IMO, it depends if people think it's a test-only issue or not, a hadoop-only > issue or not, etc. > All bug fixes should also go to the branch unless the risk is too high. > > -Yonik > http://lucidworks.com > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] For additional > commands, e-mail: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
