> -----Ursprungliche Nachricht----- > Von: Henri Gomez [mailto:[EMAIL PROTECTED] > Gesendet: Donnerstag, 22. April 2004 12:38 > An: Tomcat Developers List > Betreff: Re: AW: AW: java.io.IOException: Read bytes failed in Tomcat > 3.3.2 with Coyote Connectors > > > Hans Schmid wrote: > > > Hi Henri, > > > > thanks for the quick reply. > > > > This is what is used (same config as the former AJP13Connector) > > > > <CoyoteConnector > > processorClassName="org.apache.jk.server.JkCoyoteHandler" > > port="11019" > > maxThreads="2048" > > maxSpareThreads="80" > > minSpareThreads="40" /> > > > > old was: > > > > <Ajp13Connector port="11019" > > shutdownEnable="true" > > maxThreads="2048" > > maxSpareThreads="80" > > minSpareThreads="40" /> > > > > served from 2 Apache Servers each with 1024 maxthreads. Threads > should be > > enough. > > > > > > I saw 8 occurrences of the stacktrace within 2 Days, so it does > not occur > > often. > > I can go back to my old server.xml with the AJP13Connectors as > well, but I > > would > > love to run Coyote for a while in order to prepare a Tomcat 5 upgrade. > > > > We have up to 50 concurrent users and up to 800000 req/day :) > > But not now. this is in November :) > > > > > > BTW. I am missing the shutdownEnabled feature :( > > (One addional AJP12Connector and port per Tomcat Instance required) > > 2048 threads on both Tomcat and Apache2 ? > > I'm using only 150 threads on both end
1024 on each of the 2 Apaches and 2048 on Tomcat. We ran into problems with 1024/512/512 setup on peek times, but this was probably related to a bug in mod_jk1.2.5, where apache 1.3 left open file handles when doing graceful restarts. This led to fork Errors on the Apache side and to hung Tomcat AJP13Connector threads. This is fixed in mod_jk 1.2.6-dev and we applied this patch successfully. Just as Info: We have 3 Tomcats running in parallel, which we restart hourly. Every hour we shutdown the oldest Tomcat, start it up again, symlink a different worker.properties on our Apache and gracefully restart them. This way we use the lb worker to switch load to a newly started Tomcat and let old sticky sessions slowly finish on the older instances. This all is done due to a memory leak in a third party lib we can not influence :( More ideas? Thanks a lot, Hans > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]