Bug#219378: Runaway number of children
On Thu, 6 Nov 2003, Fabio Massimo Di Nitto wrote:
>
> Hi Dave,
> I have been doing some load tests on my machine but i cannot
> reproduce this problem. To "stress" apache i used the following command in
> all my tests:
>
> ab -n 100000 -c 100 http://localhost/
>
> The amount of processes were never exceeding the values specified in the
> config for:
>
> MinSpareServers
> MaxSpareServers
> StartServers
> MaxClients
> MaxRequestsPerChild
>
I don't mean that they were exceeding these values, it just seems that it
too rapidly approaches MaxClients. Under normal load there's maybe 2 or 3
children actually in use at any time, so even 20 is a lot.
These are the setting I'm using:
MinSpareServers 5
MaxSpareServers 10
StartServers 5
MaxClients 150
MaxRequestsPerChild 100
yet it seems after just a request or 2 from a child, a new one is spawned
and the old one remains. I left MaxClients as 150 to be able to handle
cases of high load but that doesn't mean I actually *want* 150 running!
Maybe something changed upstream? Or is this really unexpected behavior?
I dont see why a mostly idle machine would need 60 apache children eating
up memory (and database handles!)
Reply to: