Re: Bug in 1.1.9NOVM

From: <james@dont-contact.us>
Date: Tue, 29 Apr 1997 16:58:19 +0200 (SAT)

On Mon, 28 Apr 1997 james@iafrica.com wrote:

> On Sun, 27 Apr 1997, James R Grinter wrote:
>
> > at a guess, your system's number of fds is > 1024, but Solaris select()
> > can't work with that. If this is so, compile to use poll() instead.
> >
> > >#6 0x1e080 in comm_select (sec=-1) at comm.c:1004
> >
> > you probably never hit this problem with the vm versions because
> > they tend to hold less open file descriptors at any one time.
>
> Thanks, I've recompiled 1.NOVM.10 with -DUSE_POLL
>
> You were right about reaching 1024 decriptors, according to the cachemgr
> it was hovering around 980 and the uptime of the process was at most about
> 20 mins.

I'm using 1.NOVM.10 with -DUSE_POLL and 4096 file descriptors. Squid
died again, just before it died squid was reportedly using 4018 file
descriptors. The maximum for solaris is apparently 4096, what now? I
guess I'll have to use the VM version, which means buying more memory...

                     [James Power]
[UUnet Internet Africa Operations]
Received on Tue Apr 29 1997 - 08:24:59 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:35:03 MST