Re: Squid 2.0 Patch 2 - Traffic and resource counters

Hi,

> On Tue, 20 Oct 1998, Daryl Collins wrote:
>
> > Hi all,
> >
> > > Hi Squid users!
> > >
> > > I've installed Squid 2.0 patch2.
> > > client_http.kbytes_out = 512546
> > >
> > > server.all.kbytes_in = 551342
> > > server.http.kbytes_in = 550658
> > > ----------// //-------------
> > > as i know, server.* should be < client.* because of caching.
> >
> > I've seen this in all versions since 1.2.beta25 - the amount coming into
> > the proxy is usually significantly less than what it is sending out (which
> > is good!).
>
> I don't get your point, if cache gets less then sends it is ok (the more
> the better hit ratio), but our cache somehow gets MORE information then
> sends to clients. See above statistics - 6MB difference during short run.

Sorry I wasnt terribly clear - under 1.2.beta25 and below, we see huge
bandwidth savings from the proxy, but with 2.0pre1 and above, we see a
negative savings as you are seeing.
  
> > I found that by putting the fwdAbortFetch code (which was removed between
> > 1.2.beta25 and 2.0pre1) solved the problem.
>
> You mean that code fwdAbortFetch cures that behaviour, I'm experiencing?

Yes, looking at the patch between 1.2.beta25 and 2.0pre1 and manually
patching the fwdAbortFetch code back into 2.0pre1, I see the bandwidth
savings that I usually get under 1.2.beta25.

I'm surprised that more people arent seeing this problem. I'm seeing it
under solaris 2.5.1

Daryl
Received on Tue Oct 20 1998 - 07:47:29 MDT