Hi

We run Apache 1.3.20 on a Windows NT4 box (dual Xeon 550 w 512Mb RAM), and
have a problem where our web server stalls. That is, it will cease to
respond to any requests.
This happens quite regularly and is becoming a point of concern. It will
usually last at least 15 minutes and probably longer. When the server isn't
stalled performance is good even at peak usage times, which indicates that
while our service is busy it is by no means overloaded.
Now, the problem seems to have something to do with wa.exe, because quite
often what I will do to cure the problem is to kill the wa.exe processes
using Kill.exe from the NT resource kit. This is perhaps not an ideal
method, but having the service ground to a halt is embarassing, and the web
server does start to work again once I have kille dthsoe processes.
When the server is in that "stalled" state, looking at NT task manager
doesn't yield any clues that I can see. That is, there is over 250Mb of free
physical RAM, and processor usage under 50%. However, the process list does
contain lots of wa.exe processes (around about 2 dozen, say) whereas in the
normal running of things wa.exe processes come and go and there are never
normally more than half a dozen open at any one time.
Now, I am looking into configuration of Apache and while I'm not expert at
the mechanics of how web servers work, I can't see anything wrong with the
configuration. I thought it might be something to do with wa.exe processes
not timing out properly, coupled with frustrated users who hit their stop
buttons & then try the wa.exe request again, leaving 2 or more processes
going and further adding to the problem. However, that is really just
guesswork on my part and since it *seems* that wa.exe plays a big part in
it, I was wondering if anyone here had come across this before?

--
Spencer Warhurst
Listserv Service Manager