From owner-freebsd-isp Sun Sep 15 15:53:00 1996 Return-Path: owner-isp Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id PAA20286 for isp-outgoing; Sun, 15 Sep 1996 15:53:00 -0700 (PDT) Received: from orion.webspan.net (root@orion.webspan.net [206.154.70.41]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id PAA20280 for ; Sun, 15 Sep 1996 15:52:58 -0700 (PDT) Received: from localhost (gpalmer@localhost [127.0.0.1]) by orion.webspan.net (8.7.5/8.6.12) with SMTP id SAA21893; Sun, 15 Sep 1996 18:52:30 -0400 (EDT) X-Authentication-Warning: orion.webspan.net: Host gpalmer@localhost [127.0.0.1] didn't use HELO protocol To: petzi@zit.th-darmstadt.de (Michael Beckmann) cc: isp@freebsd.org From: "Gary Palmer" Subject: Re: INN history file and disk I/O In-reply-to: Your message of "Sun, 15 Sep 1996 21:38:40 +0200." Date: Sun, 15 Sep 1996 18:52:30 -0400 Message-ID: <21889.842827950@orion.webspan.net> Sender: owner-isp@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Michael Beckmann wrote in message ID : > Hello ISPs, > > I have run into a problem with one of my newsservers recently. It is a > P120, 64 MB, Adaptec 2940, 3 SCSI drives running FreeBSD 980801-SNAP. INN > version 1.4unoff4. It receives and sends several full feeds, but is hardly > used by readers. > In general, the performance of this system is good; I see about 5 - 10 % > cpu load and almost no swap usage. I get immediate response from the > system, except when I try to telnet to the nntp port. nntp readers simply > time out; it takes forever until the INN prompt appears, if it appears at > all. This situation changes when I throttle the INN. Then I get the INN > prompt immediately. > Using systat -vm, I have found out that it has 80 - 100 disk seeks per > second on the disk that carries the history file (which is about 90 MB > large). When I move the history file, it is the other disk that gets hit, > so it definitely is the history file that causes this I/O. Could this be > the reason for my problem ? You could probably use more memory, since innd will try to keep the history file in memory. Also, if the machine(s) are receiving full feeds, there will be quite a lot of data being written to the history file. It could be thrashing however (since 90Mb > 64Mb) I only see (on average) < 20 seeks/sec on my history partition, except when update(8) does its job. > Sometimes I see messages like this in the log: > Sep 15 21:07:00 news innd: ME cant sendto CCreader bytes 4 No such file or > directory > I don't know if this is relevant here. Dunno about this. Gary -- Gary Palmer FreeBSD Core Team Member FreeBSD: Turning PC's into workstations. See http://www.FreeBSD.ORG/ for info