Hello, It sounds like disabled http keep alive (connection cache). Here is the solution<http://blog.griddynamics.com/2011/04/fast-hessian-methods-leads-to.html>for jdk's http client. Unfortunately I have no experience with your Commons Http Client, but cm.closeIdleConnections(0L) looks very suspicious.
Please let me know whether it works for you. Regards On Fri, Jan 13, 2012 at 7:24 PM, Jonty Rhods <jonty.rh...@gmail.com> wrote: > Hi All , > > I am facing problem of too many CLOSE_WAIT. > My env is : > > solr 3.4 in Linux RHEL 5.2. I am getting around 1 million request per day > on application server on my production. > Production server is communicating locally with solr server. > I have 5 core setup and for each core I am using seprate instance of > solrServer with following snippet of code. > > MultiThreadedHttpConnectionManager cm=new > MultiThreadedHttpConnectionManager(); > cm.getParams().setMaxTotalConnections(100); > cm.getParams().setDefaultMaxConnectionsPerHost(100); > cm.closeIdleConnections(0L); > HttpClient httpClient=new HttpClient(cm); > server=new CommonsHttpSolrServer(url, httpClient); > > At starting my server responding fine but as time progress it start > responding slow and then stop working (within a day). Only solution at that > time is to reset the application server. > What I notice is at the time when server start responding slow that time > CLOSE_WAIT increased. When I reset the application server CLOSE_WAIT become > 0. I search and find it is due to CLOSE_WAIT. > > Please suggest me how to resolve this CLOSE_WAIT issue. > > BR > Jonty > -- Sincerely yours Mikhail Khludnev Lucid Certified Apache Lucene/Solr Developer Grid Dynamics <http://www.griddynamics.com> <mkhlud...@griddynamics.com>
