I've kind of asked this question before - What does JVM flag CMSClassUnloadingEnabled actually do?, but this time it's slightly different!
If I start a Java VM with the -XX:+CMSPermGenSweepingEnabled flag set, the following message is printed:
Please use CMSClassUnloadingEnabled in place of CMSPermGenSweepingEnabled in the future
The -XX:+CMSPermGenSweepingEnabled flag is often used to mitigate against PermGen OutOfMemory errors, however I have read elsewhere that people have found that by following the above advice they have still had these errors, but by sticking with -XX:+CMSPermGenSweepingEnabled as well as -XX:+CMSClassUnloadingEnabled their VM has stayed up longer between restarts.
Does -XX:+CMSClassUnloadingEnabled really supersede -XX:+CMSPermGenSweepingEnabled or is there still some benefit in having them both?
Thanks in advance
Rich ps: I know that the root cause of perm gen issues is still usually Classloader leaks, this is more about the message that the JVM produces if we use the above options.
From what I can tell, the only thing that the 1.6 JVM does with CMSPermGenSweepingEnabled is to print that message - it's not processed the way it is in 1.5. For CMSClassUnloadingEnabled to have any impact, UseConcMarkSweepGC must also be set.
来源：https://stackoverflow.com/questions/3717937/cmspermgensweepingenabled-vs-cmsclassunloadingenabled