> > However notice that there is 5 minutes gap between the actual data > has been read from the client and send to the tomcat. > By that time tomcat already closed the connection (30 seconds in > server.xml) and then > any request to send the data fails. Actually all sockets shut down. > > > So, seems like you have very slow client, so try to enlarge the > timeouts > or find the reason for slow clients.
Ok, understood. My problem is that it only occurs for the tomcat local to the mod_jk. This is a very weird circumstance. If it was a client problem, I'd think it'd be server-agnostic, but I can confirm only seeing it from one client (which is a really a series of http proxies from the client. One ASP based and the other custom, that's all I know about it). Is there something in the new code that is just making things more sensitive on the mod_jk side, which is why this never showed up in .28? > > Also why are you using recovery_options = 4 ? > This seemed the most logical. 4 seemed natural because why should we close the connection to tomcat if a client write error occurs and I also didn't want 1 and 2 because I figure that it should be able to recover to a different tomcat if any particular tomcat errored out (wouldn't mean ALL tomcats would error). -Tony --------------------------- Manager, IT Operations Format Dynamics, Inc. 303-573-1800x27 abia...@formatdynamics.com http://www.formatdynamics.com --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
