I'm fairly certain they are 1.0 requests. -----Original Message----- From: Sexton, George [mailto:[EMAIL PROTECTED]] Sent: Wednesday, August 14, 2002 2:49 PM To: Tomcat Users List Subject: RE: Resend: Tomcat Connection Management/Performance
Just out of curiosity, does your test bed do HTTP 1.0 requests, or HTTP 1.1 requests? Does it explicitly set a Connection: close header in the request if you are doing http 1.1 requests? -----Original Message----- From: Marinko, Jeff [mailto:[EMAIL PROTECTED]] Sent: 14 August, 2002 3:41 PM To: 'Tomcat Users List (E-mail)' Subject: Resend: Tomcat Connection Management/Performance Greetings! I've been running some performance tests and have run into a slight problem. Connections are not being properly closed/timed out, and this causes a great deal of processing by Tomcat and incoming connections to be refused (as the connections reach "acceptCount" in server.xml). I am using a simple tool to do transactions per second (TPS), and for the most part Tomcat and my underlying servlet are performing very well (underlying servlet does not create any connections). However, after running for a while (several thousand requests), I get numerous "connection refused" errors. When I do a "netstat -a" on WinNT 4 or Win2K, I can see numerous connections between the test program/box and Tomcat sitting in a TIME_WAIT status. These connections persist even after I stop the test, and even after the "connectionTimeout" value I specified in server.xml. The numerous connections eventually go away, but not nearly quickly enough. Any advice on how to remedy this problem? Some details: Tomcat 4.0.4, WinNT/2K, JDK 1.3.1 and 1.4 (happens on both). In server.xml: <Connector className="org.apache.catalina.connector.http.HttpConnector" port="8080" minProcessors="100" maxProcessors="100" enableLookups="false" redirectPort="8443" bufferSize="8192" acceptCount="2000" debug="0" connectionTimeout="5000"/> If I'm interpreting the docs right, the "connectionTimeout" should be 5 seconds. Typically, the connections that are the problem don't go away for 30+ seconds, far longer than I can afford. Any suggestions would be great! Jeff -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]> -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]> -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
