> > >>> The only thing I found suspicious is that PermGen is at 99% in both of >>> them. >> >> What should be, accoring to your experience, a correct permgen size ?
> (You also have a tonne of non-Tomcat threads running around, but none >>> appear hung; I presume your webapps created these and will dispose of them >>> properly at an appropriate time.) >> >> Could you list these threads ? From within TDA, the only thing I saw is 10 Quartz process in waiting state when all other categories have only one instance. > >>> If you are filling up the PermGen, you should be getting log entries for >>> the OOMEs being thrown - unless something in your webapps is catching them >>> and throwing them away. >> >> What is OOMEs ? I will look forward for more informations about this point. > PermGen exhaustion is frequently caused by poor application design, hanging >>> on to references to classes that should be discarded. This link (from the >>> Tomcat FAQ) contains some interesting comments: >>> >>> http://opensource.atlassian.com/confluence/spring/pages/viewpage.action?pageId=2669 >> >> Our application is spring-hibernate-oracle based. > >>> <http://opensource.atlassian.com/confluence/spring/pages/viewpage.action?pageId=2669> >>> >>> You might try increasing the PermGen size, but that is likely only to >>> delay the inevitable until you fix the memory leak. You can monitor the >>> PermGen growth with JConsole with little impact on performance. >> >> I will check about Jconsole. Is there a special thing to know about WebApp or Tomcat configuration about Jconsole connection ? Regards, Olivier
