The KeepAliveTimeout is 15 seconds (default). After doing some additional internal investigation and making the issue more well known, it looks like a culprit could be a monitoring job that a well intentioned employee added on another server that makes sure the application is holding up. Asked them to stop until we can get it through QA and verify that it's not at fault. They used the perl LWP::UserAgent in a cron job and send the results to themselves, creative.
In any event, after having to recycle apache because I hit my 200 MaxClients (all 200 shows state="w" in server-status), the problem has not come back. Is there a way to tie the PID back to the results from a netstat? John Moore Bill Barker wrote: >I've been running a nearly identical (except for the hardware) configuration >for quite some time now without problems. I don't have enough information >to give an answer, but if I had to guess, I'd say that your KeepAliveTimeout >is set way too high. Some TIME_WAITs are just the cost of doing business on >the net (since browsers are greedy), but by default Apache does a good job >of managing things. > >The way to tell where to look is by the port number. If the port is 80 or >443, then it is an Apache problem. If it is port 8007 or 8009 it is >probably a Tomcat problem. >----- Original Message ----- >From: "John Moore" <[EMAIL PROTECTED]> >To: "Tomcat-Dev (E-mail)" <[EMAIL PROTECTED]> >Sent: Thursday, January 31, 2002 7:18 PM >Subject: FW: Apache/Tomcat/mod_jk growing TIME_WAIT counts > > >>** Also posted to Tomcat-User >> >>First, I am running an Apache 1.3.22 with Tomcat 3.3 using shared mod_jk, >>mod_ssl and mod_rewrite on a Sun Netra X1 Solaris 8, JDK 1.3.1_02 with os >>and jvm recommended patches applied. >> >>I notice that I have a growing number of processes that are left in a "w" >>state according to server-status. I check using netstat and see that >> >there > >>are sockets hanging out in a TIME_WAIT state. These will hang out for >> >who > >>knows how long but I have some that are days old (when apache was last >>restarted). Is this a bad-client issue, configuration of apache issue, >>apache-tomcat-session issue or something else I'm not considering. >> >>I'm guessing it may be an Apache issue but since I monitor this forum and >>see lots of expertise on combined containers I thought I'd start here. >> >Any > >>advise or direction is greatly appreciated. >> >>John Moore >> >> >> >> >> > > >-- >To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> >For additional commands, e-mail: <mailto:[EMAIL PROTECTED]> >