Date: Wed, 11 Oct 2000 08:58:13 -0400 From: Tom Coughlan <coughlan@missioncriticallinux.com> To: dmarkh@cfl.rr.com Cc: Eric.Ayers@compgen.com, aic7xxx@FreeBSD.ORG Subject: Re: AHA 29160 adapter and clustering Message-ID: <39E463E5.62D1265E@missioncriticallinux.com> References: <14819.35403.149833.289712@gargle.gargle.HOWL> <39E39DA5.B98B81F1@cfl.rr.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Re. Automatic termination. My reading of the documentation indicates that auto termination enables the termination on a bus segment when there is one cable plugged in to that segment on the module. Auto termination disables the termination when there are two cables plugged in to a segment on the module. When you want to attach the module to an external SCSI bus, and you want to use external termination, then you can not connect anything to the internal connector for that bus segment. I believe that in this situation auto termination will enable the on-board termination, which is not what you want. Eric: I have found that I get CRC errors whenever I use the Adaptec 29160 on a multi-initiator bus with external termination. I have tried pass-through terminators and Y-cables, with several terminator types, and gotten the same results. I would love to hear from someone at Adaptec on this. I have found that all my troubles go away if I use the 29160's on-board termination. This is not an optimal cluster configuration, because a host can not be removed and leave the SCSI bus operational, but it does appear to be a work-around, for now. There is one important note regarding the on-board termination: the termination becomes disabled when the power to the module is removed, no matter how it was set in the BIOS set-up utility. This is not good for a multi-initiator configuration. You can solve this problem by installing a jumper on the "TE" posts on the module. This permanently enables the termination, and overrides the BIOS setting. (The jumpers are conveniently described in the 39160 documentation, but not mentioned in the 29160 documentation.) -- Tom Coughlan Mission Critical Linux tel: 978-606-0262 coughlan@missioncriticallinux.com Mark Hounschell wrote: > > "Eric Z. Ayers" wrote: > > > > Hello all, > > > > I'm working on a clustering setup for high availability using shared > > storage. The point of sending this email is to: > > > > 1) Discover if there is some "simple" solution to my problem using > > AHA7892 cards. (see description below) > > 2) If not, then why does this card have a problem? What's going on? > > 3) Validate the fact that single ended devices are due to be phased > > out and that LVD is the way to go for the future. > > 4) Seek suggestions on other LVD SCSI hostadapters that have been shown > > to work well in a clustering setup. Someone from the kimberlite mailing > > list has suggested that the Tekram (ncr 5380 based cards) do not > > allow termination to be disabled. > > > > -------------- > > > > The cluster will use 2 x86 boxes running Linux and shared SCSI to two > > external hard disks. Here is my dilemma. My first setup was with > > single ended signaling and two AHA2940 cards. It seemed to work > > pretty well. > > > > The fellow spec'ing hardware is convinced that single ended SCSI > > drives will soon be scarce as hen's teeth, so we bought two new > > AHA29160 cards and 2 LVD Seagate hard drives. The firmware rev on > > these cards is 2.57.2. > > > > Oct 10 12:16:15 dru1b kernel: (scsi0) <Adaptec AIC-7892 Ultra 160/m SCSI host ad > > apter> found at PCI 2/13/0 > > > > My problem is that these two cards do not seem to like to be plugged > > together on the same SCSI chain in a cluster. Before we even started, > > we called Adapted and they saidHere's what my scsi > > chain looks like: > > > > T > > | > > +-disk > > | > > +-disk > > | > > +-machine A > > | > > +-machine B > > | > > T > > > > Termination on the cards is turned off. REset bus on IC initialization > > is also turned off. I have 2 external terminators. > > > > What I'm seeing is that rebooting machine A makes machine B hang, and > > vice versa. If I disconnect one of the machines from the SCSI bus, > > that machine will not boot, unless I put a terminated SCSI bus on it, > > or set termination back to auto. > > > > The OS doesn't even have to be booted yet for this to happen. It's > > been locked up hard in the POST procedure or while trying to load the > > Linux kernel from an IDE hard disk. The terminators I'm using are > > active terminators. > > > > I'm sending this request to some other mailing lists, but not CC'ing > > between them. Any advice or suggestions would be greatly appreciated! > > > > Thanks, > > -Eric. > > Have you tried setting the auto term on the cards to always be auto. > I was always under the impression that auto meant just that, in that > the card somehow knows if there is external termination or not and > will automatically turns it's termination on if there isn't external > termination found and turn it off if there is external termination > found. If that doesn't work then either my impression is incorrect > or your cards don't work as spec'd. > > Good luck.. > > -- > Mark Hounschell > dmarkh@cfl.rr.com > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe aic7xxx" in the body of the message To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe aic7xxx" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?39E463E5.62D1265E>