In the end the problem I was having was with the configure test that was
being run. By ignoring the #if's that could prevent the loop that opens
fd's from happening, I convinced configure to actually see how many fd's
it could open, and use that value. In the end, rl.rlim_max was accurate,
and I seem to have my 20k fd's.
Most of this is in the FAQ, up to the part about configure not seeming to
do the right thing.
-Peter
On Sat, Dec 19, 1998 at 12:27:57AM +0300, Mikhail A. Sokolov wrote:
> As reflected in docs on www.freebsd.org and, of course, in your own set,
> you can play with MAX_FDSETSIZE which is in /sys/sys/types.h. After that,
> recompile (at least, probably whole make world would be ok) kernel, squid
> and here you go. Don't forget about login.conf(5) limits..
>
> This is non-squid related, but guys, maybe we should have it in the FAQ?
>
> On Fri, Dec 18, 1998 at 03:10:02PM -0500, Peter C. Norton wrote:
> # Does squid set its own resource limits, or does it have to obey the
> # shell's max? I have a set of 4 freebsd 2.2.7 systems that are configured
> # with 30000 file descriptors available, however neither ulimit(1) or the
> # shells' builtin ulimit on the system will not ulimit themselves above 8192
> # fd's. Can I recompile squid and somehow force this limit up? I would
> # like to be able to set this to about 25000 fd's. I am already running
> # into occasionaly problems w/ the current limit of 8k.
> #
> # TIA,
> # Peter
> #
> #
> # --
> # Peter C. Norton Time comes into it. / Say it. Say it.
> # [EMAIL PROTECTED] | The Universe is made of stories,
> # http://spacey.dyn.ml.org | not of atoms.
> # |
> # Muriel Rukeyser "The Speed of Darkness"
>
> --
> -mishania
--
Peter C. Norton Time comes into it. / Say it. Say it.
[EMAIL PROTECTED] | The Universe is made of stories,
http://spacey.dyn.ml.org | not of atoms.
|
Muriel Rukeyser "The Speed of Darkness"
