Date: Mon, 26 Jun 2000 11:08:29 +0200 From: Thomas Zenker <thz@Lennartz-electronic.de> To: Don Lewis <Don.Lewis@tsc.tdk.com> Cc: scsi@FreeBSD.org Subject: Re: Invalidating pack messages Message-ID: <20000626110829.A563@mezcal.tue.le> In-Reply-To: <200006220729.AAA07327@salsa.gv.tsc.tdk.com>; from Don.Lewis@tsc.tdk.com on Thu, Jun 22, 2000 at 12:29:59AM -0700 References: <20000620172810.A84355@albury.net.au> <200006200754.AAA28201@salsa.gv.tsc.tdk.com> <20000621143609.A3012@albury.net.au> <200006220729.AAA07327@salsa.gv.tsc.tdk.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Jun 22, 2000 at 12:29:59AM -0700, Don Lewis wrote: > On Jun 21, 2:36pm, Nick Slager wrote: > } Subject: Re: Invalidating pack messages > > } I pulled out the power supply this morning, and replaced it with a brand new > } unit. The system has just crashed again with the 'Invalidating pack' error > } messages. > > [ snip ] > > } If this is the case (and I'm not doubting what you say), what else could cause > } this problem? > > If your seeing funny blinking lights on the drive, and you are not the only > person having problems with this particular drive model, I would be very > suspicious that a drive firmware bug is being tickled. The best solution > in this case would be to obtain a better version of the firmware from the > vendor, but lacking that you might try turning off tagged command queueing > or just reducing the number of tagged openings. I've noticed interactions > between tagged command queueing and write caching on Seagate drives, so you > might try turning off write caching and leaving the number of tagged > openings alone. You can do all this with camcontrol. > > I'm not a fan of write caching since it violates the assumptions behind > softupdates, so I turn it off on all my drives. I haven't seen any > real performance penalty in doing so, though I think I've heard reports > that it makes newfs run slower. > > > You may be seeing this in FreeBSD and not NT because I think the CAM SCSI > system can push the drives a lot harder than the SCSI drivers in NT. > I am seeing the same problem here with an aic7892 onboard controller on a new SuperMicro PMIIIDM3 motherboard and Seagate drive (ST39236LW) here. If tagged queueing enabled it falls in timeout with heavy writes (iozone 10 suffices). I have disabled write caching: no change. Disabling tagged queuing helps (does a buildworld and izone 6000 flawless), but write performance drops to a third. I have tried to change the drive with a ST39175LW, which I took from an other computer with aic7890 onboard and works there very well: it shows the same problems on the aic7892 as the original drive. -- Thomas Zenker c/o Lennartz electronic GmbH Phone: +49-(0)7071-93550 Email: thz@lennartz-electronic.de To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20000626110829.A563>