aacraid, 2.4.26, adaptec 2810sa

aacraid, 2.4.26, adaptec 2810sa

[Will Lowe]

I've got several Adaptec 2810SA SATA raid cards in use running Debian
Woody with a 2.4.26 kernel using the aacraid driver.  The boxes are
MySQL (4.1.1) servers with dual Xeon 2.8G cpus (hyperthreading is on)
and the filesystem is reiserfs.

At irregular intervals we get this:

Sep 24 13:59:45 cd-grapherdb02b kernel: aacraid: Host adapter reset request. SCSI hang ?
Sep 24 13:59:55 cd-grapherdb02b kernel: scsi: device set offline - command error recover failed: host 1 channel 0 id 0 lun 0
Sep 24 13:59:55 cd-grapherdb02b kernel: SCSI disk error : host 1 channel 0 id 0 lun 0 return code = 6000000
Sep 24 13:59:55 cd-grapherdb02b kernel:  I/O error: dev 08:11, sector 128
Sep 24 13:59:55 cd-grapherdb02b kernel:  I/O error: dev 08:11, sector 532385968
Sep 24 13:59:55 cd-grapherdb02b kernel: :11, sector 113398656
Sep 24 13:59:55 cd-grapherdb02b kernel:  I/O error: dev 08:11, sector 113400000

... followed by lots and lots of lines of IO errors.  The box doesn't
crash -- it's booting from another drive -- but the aacraid partitions
are unusable.  So far we can't reproduce the bug on demand.

Here's what the cli control app says about the controller:

Component Revisions
-------------------
                CLI: 4.1-0 (Build #6127)
                API: 4.1-0 (Build #6127)
    Miniport Driver: 1.1-0 Beta (Build #9999)
Controller Software: 4.1-0 (Build #7211)
    Controller BIOS: 4.1-0 (Build #7211)
Controller Firmware: (Build #7211)
Controller Hardware: 2.64

It looks like there's a newer firmware available at Adaptec.com, but
it "requires" a version of the driver that seems to be Windows-only.

Googling shows a lot of people with the "Host adapter reset request"
message going back at least a year with lots of different hardware --
and no known fixes.

Any ideas? 

-- 
        				thanks,
		
					Will

Re: aacraid, 2.4.26, adaptec 2810sa

[Mark Haverkamp]

On Fri, 2004-09-24 at 15:14, Will Lowe wrote:
> I've got several Adaptec 2810SA SATA raid cards in use running Debian
> Woody with a 2.4.26 kernel using the aacraid driver.  The boxes are
> MySQL (4.1.1) servers with dual Xeon 2.8G cpus (hyperthreading is on)
> and the filesystem is reiserfs.
> 
> At irregular intervals we get this:
> 
> Sep 24 13:59:45 cd-grapherdb02b kernel: aacraid: Host adapter reset request. SCSI hang ?
> Sep 24 13:59:55 cd-grapherdb02b kernel: scsi: device set offline - command error recover failed: host 1 channel 0 id 0 lun 0
> Sep 24 13:59:55 cd-grapherdb02b kernel: SCSI disk error : host 1 channel 0 id 0 lun 0 return code = 6000000
> Sep 24 13:59:55 cd-grapherdb02b kernel:  I/O error: dev 08:11, sector 128
> Sep 24 13:59:55 cd-grapherdb02b kernel:  I/O error: dev 08:11, sector 532385968
> Sep 24 13:59:55 cd-grapherdb02b kernel: :11, sector 113398656
> Sep 24 13:59:55 cd-grapherdb02b kernel:  I/O error: dev 08:11, sector 113400000
> 

I don't see that the 2.4 aacraid driver has the timeout recovery code
like the 2.6 version does.  Mark Salyzyn at Adaptec may be able to point
you at an aacraid driver for 2.4 that does have that code in it though.

Mark.
-- 
Mark Haverkamp <markh@osdl.org>

Re: aacraid, 2.4.26, adaptec 2810sa

[Will Lowe]

Mark Salyzyn suggested we upgrade the firmware -- that the linux
drivers should work with any version of the firmware -- and we've gone
that route for now.  If the firmware doesn't fix it we'll try the
newer driver.

Since we can't trigger the issue on demand, we're in wait-and-see
mode.

Thanks to both Marks for their input on this.

On Fri, Sep 24, 2004 at 04:02:14PM -0700, Mark Haverkamp wrote:
> On Fri, 2004-09-24 at 15:14, Will Lowe wrote:
> > I've got several Adaptec 2810SA SATA raid cards in use running Debian
> > Woody with a 2.4.26 kernel using the aacraid driver.  The boxes are
> > MySQL (4.1.1) servers with dual Xeon 2.8G cpus (hyperthreading is on)
> > and the filesystem is reiserfs.
> > 
> > At irregular intervals we get this:
> > 
> > Sep 24 13:59:45 cd-grapherdb02b kernel: aacraid: Host adapter reset request. SCSI hang ?
> > Sep 24 13:59:55 cd-grapherdb02b kernel: scsi: device set offline - command error recover failed: host 1 channel 0 id 0 lun 0
> > Sep 24 13:59:55 cd-grapherdb02b kernel: SCSI disk error : host 1 channel 0 id 0 lun 0 return code = 6000000
> > Sep 24 13:59:55 cd-grapherdb02b kernel:  I/O error: dev 08:11, sector 128
> > Sep 24 13:59:55 cd-grapherdb02b kernel:  I/O error: dev 08:11, sector 532385968
> > Sep 24 13:59:55 cd-grapherdb02b kernel: :11, sector 113398656
> > Sep 24 13:59:55 cd-grapherdb02b kernel:  I/O error: dev 08:11, sector 113400000
> > 
> 
> I don't see that the 2.4 aacraid driver has the timeout recovery code
> like the 2.6 version does.  Mark Salyzyn at Adaptec may be able to point
> you at an aacraid driver for 2.4 that does have that code in it though.
> 
> Mark.
> -- 
> Mark Haverkamp <markh@osdl.org>
> 

-- 
					thanks,
		
					Will