Michael Pliskin a écrit :
Hello Nicolas,
NC> Several possibilities :
NC> a) it's an infinite loop, it should then take a lot of CPU : it's a bug
NC> and should be avoided.
NC> b) Apache blocks sometimes while sending the answer to the client.
NC> (status W = it seems to be the case here). You have to define the
NC> Timeout var in httpd.conf to have it close the connection.
It actually looks like the first option - it does eat a lot of CPU,
and I have
Timeout 120
in my httpd.conf. So I of course need to go & debug the problem, the
question is then whether there is a way to limit a request processing
time somehow so Apache would force a handler to terminate even if it
is inside an infinite loop?
Yes, I think it's called a hard timeout. It should simply kill the
process. You can modify mod_neko to add it if you need.
In Apache 1.3 API, try "ap_hard_timeout" / "ap_kill_timeout" around the
call to the VM module. I don't know about Apache 2.x corresponding API.
Best,
Nicolas
--
Neko : One VM to run them all
(http://nekovm.org)
