Hard to trace server overload/freeze/hang on apache2 after a few hours of usage

[l.ritter@…]

I'm running an apache/trac server (debian etch) hosting about 8 projects of which two are most frequently used. The server runs remotely, which means that once its overloaded, the reasons are quite hard to catch, so I hope you can advise me here how to better trace the problem.

A while ago, the server occasionally managed to use up all memory resources once in 3 weeks and needed to be restarted. Since an upgrade from 0.10 to RC1 however, I need to restart apache2 about twice a day because the server clogs very fast. It is hard to say whether Trac is the culprit here, but this is new behavior in a server configuration that has hardly changed in half a year, except for the Trac upgrade.

Trac runs through mod_python. There are two additional plugins for Trac installed, batch modification and xmlrpc. The log indicates that there have been a lot of apache2 processes running that needed to be killed by the daemon because they ran too long.

So far, running a watch "ps aux | grep apache" for approx. one hour yielded only a slowly paced change in apache process count, but nothing threatening.

How can I narrow down this problem? Is it possible to exclude that it is not Trac causing that issue?

[l.ritter@…]

of course i mean: "Is it possible to exclude that it IS Trac causing that issue?"

[anonymous]

0.11rc1 is known to have memory leakage issues : #6614