Or, configuration. Consider what I ran into the last few weeks. We were doing very high load testing of Tomcat and getting the dreaded OutOfMemory error simply hitting the Tomcat examples. I immediately thought this to be bug as well. Well, after using OptimizeIt on the code, I found out that the culprit is that we were creating so many sessions that we used up all of the memory before we hit the session timeout. This is a tuning problem, not a bug. I could 1) Throw more memory at Tomcat (we had -Xmx256m) 2) Decrease the session timeout. The default is good in most cases of not-so-extreme traffic 3) Ponder on whether load testing 500-600 req/s, with each request being a completely new session is a valid test scenario. (Sometime it is: Think getting slash-dotted)
I also think that a lot of the more bizarre OOM errors being reported on the user list currently, when reloading contexts for instance, have to do with the newer GC model of new/old/permanent generation divisions - the permanent generation is filling even though old and new have room to spare. This is also tuning. (Well, the user did say that Sun is doing a "fix" for that, so I guess it could be either) That all said, OptimizeIt is well worth the money spent. It vastly decreased the amount of time spent tracking down my OOM error. (After 3 or so days of staring at heap dumps, we just got OptimizeIt, and found the problem almost immediately). >On Wed, 3 Dec 2003, Adam Fisk wrote: > >> I've heard mention on this list many times of these OutOfMemoryErrors >> not being bugs. I work on a Java app that experiences very high network >> traffic load, however, and it's been my experience that >> OutOfMemoryErrors are, in fact, always bugs regardless of how tempting >> it is to chalk it up to something else. >> >> What makes everyone so certain it's not a bug or multiple bugs? Since >> you don't get stack traces and at best can pin down the thread name for >> OutOfMemoryErrors, they take a lot of time to track down. In my >> experience, though, they tend to result from an unaccounted for >> degenerate request coming that causes the code to, well, consume all the >> available memory (i.e., a bug). >> >> -Adam >> >> >> Shapira, Yoav wrote: >> >> > Howdy, >> > This belongs on the user, not dev, list. It's most likely a simple >> > issue (not a bug) with you not allocating enough memory to the JVM. >> > Please pursue this issue on the user list. >> > >> > Yoav Shapira >> > Millennium ChemInformatics Jeff Tulley ([EMAIL PROTECTED]) (801)861-5322 Novell, Inc., The Leading Provider of Net Business Solutions http://www.novell.com --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
