The ratio (have to look at a graph), but might have done the trick because you were not moving objects out of survivor generations to old/tenur generations fast enough....ala creating a lot of new objects (tons of them) and not having enough memory in the new space and the objects couldn't be moved into old space fast because of the kickout algorithm or something..memory leak maybe or just a ton of "new" objects and not enough of them being released to old space because of so many being created at the same times. I've had weird stuff happen before with different generation sizings. I've even had it happen that the VM would have long lock or "pause" times trying to figure out what it could move into old as new objects were being created so much and not dying quick enough to just be dumped straight from eden, so they would dump into survivor space and then seemed (according to memory graphs) not to move into the tenur and kept the survivor space pretty full constantly. Basic ally if the VM runs a garbage collection on the new space and can't free enough memory then runs a full and can not free enough memory it throws an out of memory error depending on the percentage free....I'm not exactly what it does if enough is free yet the survivor space isn't staying free enough...might keep trying to run full collections or maybe if something happens that a certain amount of time passes it simply throws an out of memory error...? However it's probably mute If you get an error talking about PermGen space as the cause then you have a limit on PermGen which is too small or the default isn't large enough. PermGen is where your loaded classes and static information go. So, if you have a lot of classes and have a lot of static variables and much data in these static variables (strings etc) then you're going to get out of memory errors if the space you or the VM have allocated isn't large enough to hold the information. Basically you may have a threshold where some users are logged in and working....they haven't accessed all of the classes yet. Some other users log in later on in the day....bam all the classes in your app load and then all of your static information gets put into Perm space. OOME and you can't do anything because 1) You can't load any more classes and 2) You can't create the static data for the classes trying to load....thus the VM can't continue to operate with your classes and instru ctions because it can't load the request classes and information. Wade
----- Original Message ---- From: Stephen Caine <[EMAIL PROTECTED]> To: Tomcat Users List <users@tomcat.apache.org> Sent: Thursday, December 22, 2005 20:00:59 Subject: Re: java.lang.OutOfMemoryError: PermGen space Chuck, > > Sorry, but that's wrong. It's the ratio between the size reserved > for newly allocated objects (New Generation) and that used for long- > lived objects (Tenured or Old Generation)... I stand corrected. In either case though, changing the ratio had a dramatic effect on our 'out of memory' problems. Any insights as to why? Stephen Caine CommonGround Softworks, Inc. --------------------------------------------------------------------- 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]
