Hi, it seems this problem is more related to apache and not to tomcat. There might be a race condition with HTTP keepalive.
I will ask my questions on the apache mailing list. kind regards Janning On Wednesday 25 March 2009 10:51:55 Janning Vygen wrote: > Hi, > > we encounter some strange errors with our apache (v2.2) fronting 4 tomcat > (v5.5) server. > > We are loadbalancing with sticky sessions to our tomcats. Everything runs > fine beside the fact that 0.1% of all requests get a 503 http status code. > The rate raises under load, but it is at least 0.1% (way too much!!) and > occurs even at the early morning hours. The request is not logged on the > tomcat side (no exceptions, no access.log), only on apache access.log with > 503 and error log as stated below. > > 1. This is from our apache log file: > > [Wed Mar 25 07:08:44 2009] [error] [client 74.6.22.94] (70014)End of file > found: proxy: error reading status line from remote server tc1 > [Wed Mar 25 07:08:44 2009] [error] [client 74.6.22.94] proxy: Error reading > from remote server returned by /site.html > > 2. This is from our apache config (prefork engine) > > KeepAlive Off > > ServerLimit 650 > ListenBackLog 50 > StartServers 50 > MinSpareServers 100 > MaxSpareServers 200 > MaxClients 300 > MaxRequestsPerChild 10000 > > ProxyRequests Off > > <Proxy balancer://tomcatcluster> > BalancerMember http://tc1:80 max=100 route=tc1 > BalancerMember http://tc2:80 max=100 route=tc2 > BalancerMember http://tc3:80 max=100 route=tc3 > BalancerMember http://tc4:80 max=100 route=tc4 > Order Deny,Allow > Allow from all > </Proxy> > > ProxyPass / balancer://tomcatcluster/ stickysession=JSESSIONID|jsessionid > ProxyPassReverse / http://tc1/ > ProxyPassReverse / http://tc2/ > ProxyPassReverse / http://tc3/ > ProxyPassReverse / http://tc4/ > ProxyPreserveHost On > > 3. This is our connector from our tomcat configuration > <Connector > port="80" URIEncoding="UTF-8" maxHttpHeaderSize="8192" > maxThreads="200" minSpareThreads="100" maxSpareThreads="200" > enableLookups="false" acceptCount="25" > connectionTimeout="10000" disableUploadTimeout="true" > maxKeepAliveRequests="100" > compression="on" compressionMinSize="2048" > noCompressionUserAgents="gozilla, traviata" > compressableMimeType="text/html,text/xml,text/css" > /> > > I tried to investigate this problem but wasn't even able to reproduce it on > my test engine. Googling it, i found some other people having this problem > but no solution. > > As far as i understand from my investigations "BalancerMember http://tc1:80 > max=100" does not make sense on "prefork". max should be 1 with prefork > model. But i think this is not the reason. > > if max=100 is not effective, there "could" be situations where MaxClients > of 300 all go to one tomcat because of session affinity. But this is rather > theoretical. > > Maybe i can't use keepalive on the tomcat side. But it seems to work. But > honestly: i am clueless. > > Do you have any hints why i do encounter these connection errors? > > Or does anybody has a running configuration with apache2.2, > mod_proxy_balancer and tomcat 5.5. where these problems does not occur? > > kind regards > Janning > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org > For additional commands, e-mail: users-h...@tomcat.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
