Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 30 Nov 1999 11:42:45 -0500
From:      "J. Maynard Gelinas" <mgelinas@bbn.com>
To:        freebsd-questions@freebsd.org
Cc:        mgelinas@bbn.com
Subject:   Problems writing a disklabel/filesystems to Chaparral RAID
Message-ID:  <199911301642.LAA03784@bbn.com>

next in thread | raw e-mail | index | archive | help
   Hi folks,
   I've been having a problem with repeat SCSI timeouts while attempting to 
write a partition table and filesystems to a Chaparral based RAID array. The 
hardware specifics are as follows:

   HOST:
   PIII/500/256MB RAM with PCI Adaptec 2940-U2W controller, 789x based.
     boot disk attached to one channel, the RAID attached to the external
     channel.

Nov 30 11:52:19 ocean /kernel: ahc0: <Adaptec 2940 Ultra2 SCSI adapter> rev 
0x00 int a irq 10 on pci0.18.0
Nov 30 11:52:19 ocean /kernel: ahc0: aic7890/91 Wide Channel A, SCSI Id=7, 
16/255 SCBs

   RAID:
   External SCSI RAID manufactured by SAG using a Chaparral RAID controller.
   Eight IBM DRHS36D 36GB drives organized into a RAID 5 array.

   Here's what the kernel detects:

Nov 30 11:52:20 ocean /kernel: Waiting 15 seconds for SCSI devices to settle
Nov 30 11:52:20 ocean /kernel: pass2 at ahc0 bus 0 target 1 lun 7
Nov 30 11:52:20 ocean /kernel: pass2: <SAG ELEC G5312 G2.2> Fixed Processor 
SCSI-2 device
Nov 30 11:52:20 ocean /kernel: pass2: 80.000MB/s transfers (40.000MHz, offset 
127, 16bit), Tagged Queueing Enabled
Nov 30 11:52:20 ocean /kernel: pass3 at ahc0 bus 0 target 5 lun 0
Nov 30 11:52:20 ocean /kernel: pass3: <JMR ELEC FORTRA SERIES. 1.00> Fixed 
Processor SCSI-2 device
Nov 30 11:52:20 ocean /kernel: pass3: 3.300MB/s transfers
Nov 30 11:52:20 ocean /kernel: da1 at ahc0 bus 0 target 1 lun 0
Nov 30 11:52:20 ocean /kernel: da1: <SAG ELEC G5312 G2.2> Fixed Direct Access 
SCSI-2 device
Nov 30 11:52:20 ocean /kernel: da1: 80.000MB/s transfers (40.000MHz, offset 
127, 16bit), Tagged Queueing Enabled
Nov 30 11:52:20 ocean /kernel: da1: 209323MB (428693760 512 byte sectors: 255H 
63S/T 26684C)
Nov 30 11:52:20 ocean /kernel: changing root device to da0s1a
Nov 30 11:52:20 ocean /kernel: da0 at ahc0 bus 0 target 0 lun 0
Nov 30 11:52:20 ocean /kernel: da0: <IBM DNES-309170Y SA30> Fixed Direct 
Access SCSI-3 device
Nov 30 11:52:20 ocean /kernel: da0: 80.000MB/s transfers (40.000MHz, offset 
31, 16bit), Tagged Queueing Enabled
Nov 30 11:52:20 ocean /kernel: da0: 8748MB (17916240 512 byte sectors: 255H 
63S/T 1115C)

   da0 is the boot disk, da1 is the RAID. And here's what happens when I 
attempt to read the disklabel:

ocean# disklabel -r da1
disklabel: /dev/rda1c: Input/output error
ocean#

   At this point the machine hangs. Though:

ocean# fdisk /dev/da1
******* Working on device /dev/da1 *******
parameters extracted from in-core disklabel are:
cylinders=26684 heads=255 sectors/track=63 (16065 blks/cyl)

Figures below won't work with BIOS for partitions not in cyl 1
parameters to be used for BIOS calculations are:
cylinders=26684 heads=255 sectors/track=63 (16065 blks/cyl)

Media sector size is 512
Warning: BIOS sector numbering starts with sector 1
Information from DOS bootblock is:
The data for partition 1 is:
sysid 165,(FreeBSD/NetBSD/386BSD)
    start 63, size 428678397 (209315 Meg), flag 80 (active)
	beg: cyl 0/ sector 1/ head 1;
	end: cyl 1023/ sector 63/ head 254
The data for partition 2 is:
<UNUSED>
The data for partition 3 is:
<UNUSED>
The data for partition 4 is:
<UNUSED>
ocean# 

   Works just fine.  Also, I was able to use the /stand/sysinstall utility to 
write a label (though the machine crashed in the attempt). I was then able to 
build filesystems, which worked, though I don't feel terribly confidant that 
this will be stable over the long haul.

   My vendor suggested the crazy idea that we needed to set the RAID to LUN 1
at it's current ID because "UNIX doesn't like RAID's to be at LUN 0." I've
never seen any commercial UNIX complain about SCSI devices living at LUN 0
(other than tape devices which may need multiple LUN support)... so I think
these guys are just plain wrong. 

   I've tried this with Solaris/x86 and while I can write a disklabel I get
many SCSI timeout errors when reading/writing to the array. The vendor is
claiming that they tested the unit with NT and that it works just fine (I've
already sent it back once), and they suggest that I run NT if I want to
continue getting support for the hardware. This just seems crazy... it's a SCSI
disk as far the the OS is concerned, it just shouldn't matter what OS I use.
The only possible issue I can see is if the disk array has more cyls, heads,
sects than is supported by FreeBSD or Solaris/x86, but I can't imagine that
FreeBSD would have a problem writing disklabels while Solaris/x86 would have a
problem reading/writing to the array and that *this* is caused by a disk array
being too large. 

   When I sent the unit back I explicitly asked my vendor to double check all 
firmware revisions on each disk and the RAID controller, which they claim are 
set properly. I'm at a loss.

   Has anyone else seen these kinds of problems with similar hardware? And can 
you recommend a good RAID vendor who will support UNIX/BSD based solutions 
instead of telling me to run NT?

Thanks,
--Maynard
   





To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199911301642.LAA03784>