mps/LSI SAS2008 controller crashes when smartctl is run with upped disk tags

[Jason Wolfe]

On Thu, Nov 3, 2011 at 3:31 AM, Peter Maloney <
peter.maloney at brockmann-consult.de> wrote:

> Dear Jason,
>
> Does this mean it didn't hang? or it ran your smartctl -a test for 10
> minutes before a hang?
>
> I am also trying the mpslsi driver now, but I couldn't reproduce the
> problem using "smartctl -a" (also tried -A, -h and -i) with the mps
> driver. Tags was set to 255 on all disks. I only tried it on the backup
> server, which didn't crash randomly on its own either. So I will just
> have to assume it works if it doesn't do the same thing in a month or two.
>
> However, with the mpslsi driver, during a scrub on the backup server
> (probably during smartctl -a), I got these messages (including what
> looks like a controller reset), and no disks were lost, with no read
> errors reported in zpool status. But I can't get it to happen a second
>  time. So I hope that means our problems are over.
>
>   Peter
>

Peter,

No, the server remains stable in the wake of the errors, the mpslsi driver
just appears to handle the time outs gracefully and is more verbose in its
messages.  I see it was mentioned prior in the 'AOC-USAS2-L8i zfs panics
and SCSI errors in messages' thread, but we are basically waiting on LSI to
release the code to us which is held up in legal currently.  Seems the
standard response will be 'use the LSI driver' for now as folks
understandably don't want to duplicate the effort.

I'll continue to run the mps driver with a smartctl wrapper until we manage
to get the code from LSI.  Either way I have to cycle my machines, so
better it be to the open one when the time comes.

Jason