David, if it's fast for you to reproduce, would it be possible for you to try the latest Jetty 6.1.24 and see if the issue still exists? http://dist.codehaus.org/jetty/
Seems like we should upgrade to 6.1.24 anyway (there were quite a few fixes in 6.1.23) http://jira.codehaus.org/secure/ReleaseNote.jspa?projectId=11160&version=16044 -Yonik http://www.lucidimagination.com On Thu, May 27, 2010 at 5:43 PM, Smiley, David W. <dsmi...@mitre.org> wrote: > I'd like to warn people about the default configuration of Jetty in the Solr > trunk release (not present in Solr 1.4 and prior). There is a difference in > the jetty configuration which is for the latest Solr to use the > QueuedThreadPool (as seen in jetty.xml). Previously, it had used a > BoundedThreadPool implementation that I've heard is considered deprecated > presently. I have a multi-core setup where Jetty is serving up lots of Solr > cores 9+ and when our client does a distributed search (3 of them at a time > actually), it triggers a condition in which the query takes 50 plus seconds > to respond. During this time, the machine is effectively idle, seemingly > waiting for something. To fix this, go back to the former BoundedThreadPool > implementation or don't use Jetty. FWIW this has triggered us to swtich to > Tomcat. > > Sorry but I have sunk so much resources into tracking down this nasty problem > that I can't spend much more on further figuring out why QueuedThreadPool is > failing us. > > ~ David Smiley > Author: http://www.packtpub.com/solr-1-4-enterprise-search-server/ > > > > >
