From owner-freebsd-isp Tue Sep 16 12:42:48 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id MAA15262 for isp-outgoing; Tue, 16 Sep 1997 12:42:48 -0700 (PDT) Received: from sabre.goldsword.com (sabre.goldsword.com [199.170.202.32]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id MAA15256 for ; Tue, 16 Sep 1997 12:42:45 -0700 (PDT) Received: (from jfarmer@localhost) by sabre.goldsword.com (8.7.5/8.7.3) id PAA17639; Tue, 16 Sep 1997 15:46:02 -0400 (EDT) Date: Tue, 16 Sep 1997 15:46:02 -0400 (EDT) From: "John T. Farmer" Message-Id: <199709161946.PAA17639@sabre.goldsword.com> To: freebsd-isp@FreeBSD.ORG, steve@ns1.cioe.com Subject: Re: best news server config Cc: jfarmer@goldsword.com Sender: owner-freebsd-isp@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk On Tue, 16 Sep 1997 09:37:27 -0500 (EST) Steve Ames said: >I'm having some serious problems with our newsfeed. The bottom line seems >to be that our server just isn't keeping up with the feed. The server >is a Pentium 150 with 96M of memory. I have two 4G SCSI drives bound >together with ccd (interleave 65536 as recommended in the ccd man page). > >I have any the actual news spool mounted with async and noatime. The partition >with the history and active files is mounted async. > >I am using inn 1.5.1 compiled using MMAP. More spindles. Smaller drives. Multiple controllers. Stripe it across spindles & controllers. Seperate drives for the active and history files. Seperate drives for swap & system. More memory. More memory. etc. What type of network interface are you using? Is streaming enabled? At both ends? Browse the freebsd archives for any one of the several series of articles by Joe Greco. Also, check out the following URL: http://www.nntp.sol.net/ Joe has built several serious news servers & knows what he is talking about. Most of what I know is based on listening to him and thinking about it, and thinking about it, and thinking about it... >Everything seems quite sprightly. However when I check the news stats from >one of my feeds (sprintlink) it shows that they are throwing about 65% of >the articles away before trying to send them to me. I also have a constant >backlog of about 9700 articles (cap being 10,000 before they start throwing >away articles instead of spooling). > >Comparing the articles that show up on my server with articles that show up >on another ISPs news server is quite depressing. > >I can't seem to peg the bottle neck here. 'top' shows that the CPU never >goes below 60% idle. So not CPU. Very little swapping is occuring so >memory shouldn't be a real big issue. That only leaves drive and network >speed. Only 60% idle?!?!? That sounds like something is eating cycles somewhere. A news server should be spending most of its time in I/O wait. >The drives are normal scsi attached by an adaptec 2940 controller. Is this >the problem? Would putting in an external raid array with a wider controller >fix my problem or am I looking in the wrong place? Depends on how the raid box is configured. Usually, the reliablity goes up without any increase in transaction rate or sequential I/O. Take a look at the above references & see if they can help you. John ------------------------------------------------------------------------- John T. Farmer Proprietor, GoldSword Systems jfarmer@goldsword.com Public Internet Access in East Tennessee dial-in (423)470-9953 for info, e-mail to info@goldsword.com Network Design, Internet Services & Servers, Consulting