1. yep, it does a lot more writes ==> more storage space issues, large file issues, and also more i/o. But it is an easy way to address the problem in a lazy way. In addition to frequent log rotation, I just throw fairly decent hardware at the problem, and Linux is quite good at buffering writes like this... 2. Perl script: see previous post.

-----Original Message----- From: Fred Yankowski [mailto:fred@ontosys.com] Sent: Monday, June 09, 2003 8:18 AM To: sean.upton@uniontrib.com Cc: zope@zope.org Subject: Re: [Zope] Squid vs Apache+mod_proxy+mod_gzip

On Mon, Jun 09, 2003 at 08:55:56AM -0700, sean.upton@uniontrib.com wrote:

...

Another logging option for Squid: we log with log_mime_headers on, which dumps a rather verbose log, but we filter this to combined format for later use and archival.

Interesting idea. I just tried this briefly and I see how it would do the job. In my test case it logged about 8 times more data per GET than without 'log_mime_headers'.

As someone else asked, could you share your script for converting that data into combined log format as expected by log analyzers? It looks like a simple Perl script would do the job, extracting referrer and user-agent data when available from the corresponding headers. Are there any tricks to that?

-- Fred Yankowski fred@ontosys.com tel: +1.630.879.1312 OntoSys, Inc PGP keyID: 7B449345 fax: +1.630.879.1370 www.ontosys.com 38W242 Deerpath Rd, Batavia, IL 60510-9461, USA