From owner-freebsd-isp Fri Jul 25 22:30:11 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.5/8.8.5) id WAA15317 for isp-outgoing; Fri, 25 Jul 1997 22:30:11 -0700 (PDT) Received: from super-g.inch.com (super-g.com [207.240.140.161]) by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id WAA15312 for ; Fri, 25 Jul 1997 22:30:08 -0700 (PDT) Received: from localhost (spork@localhost) by super-g.inch.com (8.8.5/8.8.5) with SMTP id BAA24624; Sat, 26 Jul 1997 01:35:15 -0400 (EDT) Date: Sat, 26 Jul 1997 01:35:14 -0400 (EDT) From: spork X-Sender: spork@super-g.inch.com To: "Jeffrey J. Mountin" cc: Jim Shankland , freebsd-isp@FreeBSD.ORG Subject: Re: analog and Apache? In-Reply-To: <3.0.3.32.19970725221756.00c885a4@mixcom.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-isp@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk On Fri, 25 Jul 1997, Jeffrey J. Mountin wrote: > Uh, you may not want to do this. Or you might ;) Our number one concern was that if someone is paying us for quality hosting, we wish to do anything we can within our budget to make sure that some other ISP's customers get reasonable access to the site. The lag we saw coming in from an unnamed IP was *very* noticeable, and looked to the person trying to view the page like a problem on our side. Turning off lookups was like night and day... > > I found that after 4 weeks of running IP logging and using the results from > 3 servers that 5 hours to run reports vs 5 minutes was just a bit too much. We nice it down in the script. If a customer wants "instant" hit counts, we set it to look at one week's worth of logs with DNS resolving turned off, and the results come up within seconds. We've found most marketing types are happy enough with numbers for short term trends, and they look at the weekly totals with DNS for the "big picture". > There is another problem with this as well. If a site has a very large > logfile, 10 Mb, 100 Mb, or more even, Analog will start using a lot of > memory. Once the server starts swapping it takes a bit of a performance > hit. Not to mention that it was thrashing our name servers. Hmmm... Haven't seen this here, even on 200M files as long as the server is not overloaded. We set resolv.conf on the webservers to look at one of our little-used secondaries, so the deluge of DNS requests will not affect our two main nameservers. I haven't really seen this touch the load on the nameserver, though. Also, analog sets up a cache, so you'll see over a period of time the logs take less time to crunch. The current analog is very efficient, and other than a mind-spinning array of command line switches, it's fairly easy to tweak to your liking... Charles > Personally I find that FBSD and Apache work well together and with tweaking > and ample memory you can handle a lot of traffic. The "savings" of not > logging IP was non-exisistant. > > > ------------------------------------------- > Jeff Mountin - System/Network Administrator > jeff@mixcom.net > > MIX Communications > Serving the Internet since 1990 >