> > Luc Foisy wrote: > > It did eventually lead to an OutOfMemory exception ( at > least months ago ) > > But as I continue to watch "top" I have pushed it up to > tomcat using 108 megs of memory (and it doesnt go away). As > this server only has 128 megs of memory, thats pushing it, > and that was in the span of maybe 15 minutes. I didnt > actually push it to the outofmemory state this time around, > as it is quite obviously the same problem I was getting about > 4 month ago when I first switched to the 4.x strain. I was > hoping the newer version would have solved that issue ( and > darned it, I was almost ready to deploy on production server > AGAIN! I don't need the clients ripping me a new umm below > orafice again! :) > > Having that exception printed out in the logs should not lead to any > memory consumption. Useless flushing happens in Jasper, which will be > removed (but not in 4.1.18, since there is no time to do a beta). > You should check for other factors which could cause memory > consumption > (like having too many active sessions), and report if you > find a part of > Tomcat causing the problem.
One active session. Wanna let me know other factors that I could look for??? I have no idea what goes on behind the scenes and really I am not caring much, I just want something that works. Whatever it might not be, the memory use jump happens everytime i get that exception, so whatever it is, is close by. > > Anyways, using the legacy HTTP1.1 connector I have not been > able to reproduce the socketException after multiple people > pounding on it for the last 10 minutes. That makes me happy... > > The legacy HTTP/1.1 connector is bad in many ways (I know, because I > wrote it), and given the Jasper code, it is wrong in that particular > case to swallow the exception. > I don't recommend using it. Swallowed exception or not, I do not see any errors happening in my logs, nor do I see an extreme increase in memory usage. With a single session using the Coyote connector I was able to increase memory usage 100% in less that 15 minutes ( not good when I expect at least 300 sessions simultaneously ) So give yourself and your bad connector a little more credit, it doesn't blow up my server. And if its the same connector used in 3.2.x , I'll be way more than happy to use it since we have been using it for the last year and a half without any problems. -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
