CGIIP eats memory

[Cecil]

OE 10.1B DB/WebSpeed/Messenger
OS RHE 4
WEB Apache

I am a unusual problem on our live production server that everything runs very slow/pauses, and the problem is the CGIIP WebSpeed Messenger is trying to utilise every scrap of available memory and the server starts paging to disk.

So our server is then unresponsive for about 15-20 mins then I am able to kill cgiip process which is causing the memory issue. However I looked at every log file I can find and I am unable to find any warning or errors.

I rasied a progress tech support call, and they are unable to help me as they have never seen this problem before.

This problem is happening about once a week and the website code and configuration has not been touched since DEC 2007 so I have no idear why it's happening.

Have you seen a simular problem?

 

[Ell]

I've never seen CGIIP do it, but bad code will cause the _progress processes to hang and bang hard on memory. Things like bad dynamic queries or infinite loops are generally the culprit in our environment. Also, accidental full table scans can be an issue. Check the newest batch(es) of code and make sure no one has a huge bug.

 

[Cecil]

The problem was found due to the CGI messengers' internal function of setting cookie "SERVER-SESSION-ID". I found that under certain conditions that it's looping and generating a never ending cookie value which is causing high CPU cycles and using large amounts of memory.

Solution: Don't allow the messenger to generate the SERVER-SESSION-ID cookie but control your own Session cookie inside your own code.

Currently I am using 10.1B hoped this has been fixed in later versions of OE.