[Orca-users] Re: orcollator dying with long lines in squid log

[Blair Zajac]

Good to hear that orcallator.se is not dumping core any more.

As for all the cache values, I've never had orcallator process this type
of data nor have I analyzed any myself, so I cannot help if what you're
seeing is correct or not.

You may wonder why then is this code in orcallator.  Adrian Cockcroft
put it in his percollator.se and I made sure the same code was in
orcallator.se.

There are no proper entries in orcallator.cfg for the cache values.
If you come up with some plot {...} entries, please email them to me.

Thanks,
Blair

Paul Haldane wrote:
> 
> On Mon, 15 Nov 1999, Blair Zajac wrote:
> 
> > Thanks for the bug catch regarding getc(stdin) => getc(www_fd).  I've
> > made this change to the existing code.  Unless you see anything else,
> > I'm going to push this new version tomorrow.
> 
> That's now running quite happily - hasn't died since putting that fix in.
> 
> I've just started looking properly at the data generated and I'm puzzled
> by a couple of things.  These may just be stuff that I'm looking at from
> the wrong angle.
> 
> 1) we always get the same value for %cch_req and %cch_qry - this may be
> because of how our cache hierarchy is - I haven't really thought about
> this enough (and it's a while since I had to look at a squid access.log
> file in any depth :->).
> 
> 2) orcallator counts the number of client requests - stores them in the
> variable squid_client_http but never does anything with this value.  I'm
> not sure that we need this as the same info can be got (I think) from the
> httpop graphs.
> 
> %cch_req [and %cch_qry] represent the percentage of requests coming from
> other caches (as opposed to requests direct from users) - is that right?
> 
> Paul
> 
> > Paul Haldane wrote:
> > >
> > > Almost - small problem in measure_web - see below - I thought I'd cc'd you
> > > when I responded to the list, but looks like I didn't
> > >
> > > Paul
> > >
> > > On Mon, 8 Nov 1999, Blair Zajac wrote:
> > >
> > > > Paul,
> > > >
> > > > How is that version of orcallator.se I sent you?  Any problems, core
> > > > dumps, etc?  Please let me know, as I'd like to push out a new release
> > > > of it.
> > > >
> > > > Thanks,
> > > > Blair
> > >
> > > From: Paul Haldane <Paul.Haldane at newcastle.ac.uk>
> > > To: orca-discuss at onelist.com
> > > Subject: Re: [orca-discuss] orcollator dying with long lines in squid log
> > > Date: Thu, 4 Nov 1999 10:08:23 +0000 (GMT)
> > >
> > > From: Paul Haldane <Paul.Haldane at newcastle.ac.uk>
> > >
> > > That seems to work fine (but see below).  I wasn't sure how to tell
> > > orcallator.se just to read an existing log file, so I just took the
> > > accesslog function and stuck it into my little test program - that
> > > stopped the failure, and that's without your extra code in measure-web
> > > for discarding the remainder of long lines.
> > >
> > > Moved to your version of measure_web and it apparently hangs -
> > > shouldn't the getc(stdin) (both of them) in measure_web be getc(www_fd)?
> > >
> > > Paul
> > >
> > > On Wed, 3 Nov 1999, Blair Zajac wrote:
> > >
> > > > Please try the enclosed version of orcallator.se.  It should not
> > > > dump core on the long lines.  Why don't you try it on a short
> > > > test log file to see if it works?
> > > >
> > > > Blair
> > > >
> > > >
> > > > Paul Haldane wrote:
> > > > >
> > > > > From: Paul Haldane <Paul.Haldane at newcastle.ac.uk>
> > > > >
> > > > > Hi - I've been looking in more detail at my problem with orcollator (1.20)
> > > > > and long lines in the squid access.log file (previous mentioned on the
> > > > > orcahelp list).
> > > > >
> > > ...