On Apr 13, 2008, at 5:08 PM, Chris Cantwell wrote:
> I finally solved this one.
>
> I changed to a spare identical motherboard (good thing I had one),  
> with
> identical results, the computer stalled after identifying the 320GB  
> hard
> drives.  I was beginng to wonder if I needed to try another  
> motherboard with
> a different SATA controller, but everything pointed to the drives  
> beign
> corrupted somehow by the zpool command.  As a last gasp effort, I  
> destroyed
> that zpool and created another one, using slightly different device  
> names,
> added "p0" to the end of each device:
>
> # zpool create extpool raidz c2d0p0 c3d0p0 c4d0p0 c5d0p0
>
> This ends up with the same zpool as the previous effort (se below  
> message),
> and the computer will not stall during the POST, allowing me to  
> reboot it.
>
> Whew!  I don't know what was going on there, but it's fixed.

What is going on there (the difference between using the p0 and not)  
is the difference between using a disk partition vs. the entire  
disk.  When you created the ZFS pool the first time you used the  
entire disk.  By adding p0 you used the first (DOS-style) partition  
on the disk.  There is nothing wrong in particular doing it either  
way.  However I have a suspicion that the RAID controller was reading  
the disk looking for its unique RAID signature and that whatever it  
saw from ZFS confused it.  By simply using a partition you caused the  
disk to appear more "normal" and then the RAID controller would no  
longer get confused.

I hope that's clear.  Glad you were able to figure it out.

/BAK/
-- 
Ben Klang
Alkaloid Networks LLC
[EMAIL PROTECTED]
404.475.4850
http://projects.alkaloid.net

>
> Chris
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Sun, 13 Apr 2008 02:17:29 -0400
> From: "Chris Cantwell" <[EMAIL PROTECTED]>
> Subject: ZFS raidz causes BIOS POST to fail
> To: <[email protected]>
> Message-ID: <[EMAIL PROTECTED]>
> Content-Type: text/plain; charset="us-ascii"
>
> Hello, I have a bizarre situation that I've never encountered before.
>
>
>
> I installed Nextenta 1.0.1 test3 on a AMD 939 motherboard.  Two  
> primary
> drives for the OS on a SiI3132 SATA (PCI Express) controller, and four
> additional 320GB SATA drives on the onboard nVidia nForce SATA/RAID
> controller, the onboard RAID was disabled.  Nextenta recognizes all  
> the
> drives as follows:
>
> c1t0d0              80GB                SiI3132 controller
> c1t1d0              80GB
> c2d0                 320GB
> c3d0                 320GB
> c4d0                 320GB
> c5d0                 320GB
>
> The OS installs nicely on the two 80GB drives, reboots, and when  
> you get to
> the console.   zpool status shows the drives are mirrored, and  
> working OK.
>
> I then create another zpool using the four 320GB drives:
> # zpool create extpool raidz c2d0 c3d0 c4d0 c5d0
> # zfs create extpool/data
> # ln -s /extpool/data /data
> Again zpool status shows everything OK.
>
> To test the setup I copy the entire /usr directory to /data,  
> everything
> looks good, no errors.  I reboot the machine and the computer  
> stalls during
> POST, right after the external drives are recognized by the BIOS.
>
> I try some different things with no success, but I eventually fixed  
> it by
> attaching the drives to another SATA RAID controller (Highpoint  
> RocketRAID
> 1740), and "initializing" the drives.  I reattached the drives to the
> onboard nVidia SATA RAID controller, and the machine booted.
>
> I replicated the situation by booting back into Nexenta, recreating  
> the
> zpool, and rebooting again.  The machine stalled once more after  
> recognizing
> the drives on the POST.
>
> Apparently ZFS has changed something on the 320GB drives the does  
> not agree
> with the nVidia nForce SATA controller.  The BIOS is reading some
> identification info from the drives, and the ZFS metadata on the  
> drives is
> locking up the POST.
>
> Does anyone have a workaround for this situation, so I can boot the  
> machine
> with the 320GB drives configured as raidz?
>
> Thanks,
>
> Chris
>
> ----------------------------------------------------------------------
>
> Message: 2
> Date: Sun, 13 Apr 2008 07:58:28 -0700
> From: Erast Benson <[EMAIL PROTECTED]>
> Subject: Re: ZFS raidz causes BIOS POST to fail
> To: Chris Cantwell <[EMAIL PROTECTED]>
> Cc: [email protected]
> Message-ID: <[EMAIL PROTECTED]>
> Content-Type: text/plain; charset=utf-8
>
> Chris,
>
> it could be either that your mobo's BIOS doesn't like EFI labels  
> and need to
> be upgraded, or make sure that c1 disks are first in the BIOS boot  
> order.
>
>
> _______________________________________________
> gnusol-users mailing list
> [email protected]
> http://lists.sonic.net/mailman/listinfo/gnusol-users

_______________________________________________
gnusol-users mailing list
[email protected]
http://lists.sonic.net/mailman/listinfo/gnusol-users

Reply via email to