From owner-freebsd-questions Tue Oct 27 19:04:55 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id TAA09871 for freebsd-questions-outgoing; Tue, 27 Oct 1998 19:04:55 -0800 (PST) (envelope-from owner-freebsd-questions@FreeBSD.ORG) Received: from pixel.zycor.lgc.com (pixel.zycor.lgc.com [134.132.112.130]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id TAA09852; Tue, 27 Oct 1998 19:04:52 -0800 (PST) (envelope-from rsnow@lgc.com) Received: from diablo ([134.132.78.117]) by pixel.zycor.lgc.com (8.8.8/8.8.8) with SMTP id VAA04111; Tue, 27 Oct 1998 21:03:35 -0600 (CST) (envelope-from rsnow@lgc.com) Message-ID: <000c01be021f$896fcd80$754e8486@diablo> From: "Rob Snow" To: "Doug White" Cc: , Subject: Re: CAM question 3.0-RELEASE Date: Tue, 27 Oct 1998 21:03:26 -0600 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.00.0518.4 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.0518.4 Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Yeppers, matter of fact FreeBSD showed a problem with my hardware, again. (You'd think I'd have learned since 1.1.5.1) It hit me when I woke up this morning, PCI problem. Turned out that PCI write combining was it, turned it off and everything works like a charm. I'm guessing that having the 2940UW and the fxp running at the levels that I can put them under with a 2xP6-233 FreeBSD and 1x300a@450 Linux box on full duplex 100 and a UW 'cuda showed the flaw. Now I can write to the network at over 10MB/sec. (at home :-) This is what I preach at the office all the time, the network doesn't have to be the bottleneck. Thanks for the info, this is my first run with CAM. I've been running a non-CAM -current for a while. Upgraded to 3.0-RELEASE as part of the server upgrade I did this weekend. (Accidently, /kernel didn't know my processor type and panicked on boot, kernel.GENERIC was left over from 2.2.5 and wouldn't mount shit. Woops) -Rob -----Original Message----- From: Doug White To: Rob Snow Cc: ; Date: Tuesday, October 27, 1998 6:39 PM Subject: Re: CAM question 3.0-RELEASE >On Tue, 27 Oct 1998, Rob Snow wrote: > >> I'm still debugging my problems with lockups during heavy network writing. >> I've installed two 2940's and can make it fail on either controller, >> eliminating my thought about some SCSI bus trash. >> >> Now, both of my drives seem to reset the tag queues when they go under load, >> is this normal? >> >> My Micrapolis 3243-19 (On 2940): >> tagged openings now 35 >> >> My Segate 39173W (On 2940UW): >> tagged openings now 63 >> tagged openings now 62 >> ....... >> tagged openings now 49 > >This is normal. The system tries to figure out how many tags each unit >can support by experimentation and observation. Some disks are >broken here and have to be quirk'd to turn off or reduce tags. > >> Is that supposed to happen? I'm wondering if that is an indication of a >> problem. The Seagate will drop them down slowly and then all of the sudden >> they plumet before it locks the system cold. > >Sounds like your Seagate doesn't handle tags correctly. I'd suggest >checking with Seagate for a firmware upgrade, and in the meantime adding a >quirk entry to /sys/cam/cam_xpt.c Search for 'quirks' and you'll find it. > >> Now, this happens with either local or network writes to the drives, >> however, local writes do not lock the machine. It's the heavy network >> writes from the wire that kill the machine. > >It's simply high outstanding disk transactions (which heavy writes would >cause). Nothing wrong the network code, in fact it's probably a good >thing that our network code can do better than the disk code :) > >Doug White >Internet: dwhite@resnet.uoregon.edu | FreeBSD: The Power to Serve >http://gladstone.uoregon.edu/~dwhite | www.freebsd.org > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message