Hi, On Sun, 2 Nov 2008, Michael van Elst wrote:
On Sat, Nov 01, 2008 at 08:55:48PM -0500, Al - Image Hosting Services wrote:I would guess that it does try to map the whole file, but if the kernel is built with SOSEND_NO_LOAN, shouldn't this problem go away?yes, and as you say, it already did.It doesn't wait in sokva, so did I create a new issue?Apparently. Find out what lighttpd is now waiting for any you can probably see the new bottleneck.
Well, what is so troubling is that it still takes a reboot to get lighttpd running again after it stops responding.
Is there something else in the kernel that would be overwhelmed?There can be many things, usually not in the kernel but in the application itself. My guess is that you run out of file descriptors on a very loaded webserver.
I built a kernel with 16,000 file descriptor and I quadrupled the number of file descriptors that lighttpd would use a while ago. It seemed to stop it from give error 500, but it doesn't seem to help on this server. I spread the load among 3 servers and that seemed to fix the problem for now, but I am still troubled by the fact that it requires a reboot before it will start answering requests. Also, I have noticed the amount of memory being used by [ioflush] and that it is way too high. This server will crash anyway if not rebooted every couple of days.
Best Regards, Al