Sounds like a PHP resource leak. A temporary fix would be sessions configured to time-out sooner rather than trying to recycle existing sessions open that aren't in use. We've seen a lot of code using JRUN on Win32 and Python (PHP) - on Freebsd and Redhat that have been waiting too long for the Ack back from a client (that never turns up) - resulting in time-out ..... waiting and ... waiting until ... bam ! No more resources availbable for more sessions available.
I know that people are working to resolve the problems, But i never had these continual problems on the old set-up! Did we really go backwards on the memory available, when the old system was already struggling?
Doog, No, there was no drop in the RAM available from what i've read, just that this new software deals with it differenlty and causes the issues we're all now having. Read the post in DSC topics, it's quite concise.