Re: [squid-users] What load can SQUID cope with in accelerator

Possibly buffered I/O. If you haven't turned off buffered I/O in your
redirector, do so..

Rob

----- Original Message -----
From: "Ole Moller" <olm@cybercity.dk>
To: "Brett Lymn" <blymn@baesystems.com.au>
Cc: "Duane Wessels" <wessels@squid-cache.org>; "Simon Morley"
<simon@xaraonline.com>; <squid-users@ircache.net>
Sent: Friday, March 16, 2001 8:12 PM
Subject: Re: [squid-users] What load can SQUID cope with in accelerator

At 00:24 16-03-2001, Brett Lymn wrote:
>According to Ole Moller:
> >
> >(instead of expected 1mb/sec). When squid had run for about 10
minutes it=20
> >killed itself with a message of "FATAL: Too many queued redirector=
> > requests".
> >
>
>Sounds like you need to have more redirector children - the log
>messages say you should do this. If you are still having troubles
>then you should look at the coding of your redirector and try to make
>it more efficient so they do their job quicker.

I have the maximum of 30 configured, please remember that when using the
exact same script whithin apache 1 was sufficient. Further more:

# time head -1000 access.log | awk '{printf("%s\n",$7)}' |
../bin/map.zip.pl > /dev/null
         0.06 real 0.00 user 0.00 sys

At a peak of 20 req/s and 30 redirectors to handle it, I dont think the
real problem is the redirector but "something" else.

> >I believe that this not is the real cause of death considering that I
had=20
> >started 30 and one did the job fine when running from apache. Have
you any=
> >=20
> >idea of what could be blocking? disk io?
> >
>
>What does iostat & vmstat say? These should give you some clue as to
>where the bottleneck is.

There is disc activity, but the numbers indicate that they only are
running
about 60% of maximum.

Regards

--
Ole Møller olm@cybercity.dk, Sysadm CyberCity Internet