Subject: Re: hard driver CRC error with RAID 5 under 2.0 release
To: Felix Deichmann , Rui-Xiang Guo <rxg@myrealbox.com>
From: Brian Buhrow <buhrow@lothlorien.nfbcal.org>
List: port-i386
Date: 12/21/2004 10:26:09
	ABout four years ago, I built a NetBSD box with 16 IDE disks raided
together using Promise EIDE controllers.  When the raid ran at full tilt, I
saturated the PCI bus, causing strange hangs and error messages of the type
you're seeing.  To fix the problem, I ran the disks in UDMA mode 1, rather
than UDMA mode 4 at the time.  It looks like all of your disks are running
in UDMA mode 5, which is pretty fast, considering how many disks you have
going at once.  I might suggest compiling a special kernel and turning down
the maximum speed to UDMA mode 4, or, even 3 or 2.  Experiment with
successively slower speeds until the bogus interrupts go away.  Although
the disks will run slower individually, you  probably won't notice any real
performance loss from the "daily use" perspective.  Plus, any performance
hit you take will be more than compensated by the increased reliability
you'll get from the box.  Your users will be much happier when you don't
have to keep rebooting the box and calibrating the parity on the raid.
-Brian
On Dec 21,  1:52am, Felix Deichmann wrote:
} Subject: Re: hard driver CRC error with RAID 5 under 2.0 release
} Rui-Xiang Guo wrote:
} > wd2a: error reading fsbn 4348992 of 4348992-4349023 (wd2 bn 4348992; cn 4314 tn 
} > 7 sn 39), retrying
} > wd2: (aborted command, interface CRC error)
} > wd2: soft error (corrected)
} 
} I had an IBM DTLA disk which showed this behaviour when it was about to die.
} 
} Regards
} 
} Felix
>-- End of excerpt from Felix Deichmann