At 21:37 Uhr +0200 4.4.2005, Manuel Bouyer wrote:
>On Sat, Apr 02, 2005 at 11:02:06PM +0200, Hauke Fath wrote:
>> At 18:50 Uhr +0200 2.4.2005, Manuel Bouyer wrote:
>> >I had a look at the driver.
>> >These:
>> >mpt0: Unknown async event: 0xb
>> >mpt0: Unknown async event: 0xb
>> >
>> >are defined in the mpt_mpilib.h as "MPI_EVENT_INTEGRATED_RAID".
>> >Does these board have RAID functionality ?
>>
>> Yes. They can mirror (RAID level 1) two drives - a poor man's RAID
>> controller, so-to-say. We are running all four machines in that mode.
>
>And does the problem happen when not in RAID mode ?

I never tested this. And since the machines are in production, it's
difficult now to switch on the fly.

>I'm not sure the driver supports this.

The controller presents the RAID 1 (two Fujitsu 36GB disks) as one disk to
the OS. No further software support needed. Works fine under light and
medium I/O load.

If you intended to say: "not fully supported" - that's something else.  ;)

>At last in this mode the firmware emits some events that
>are not properly handled by the driver.
>Does the firmware do all the work, or do you also have a raidframe on the
>2 devices ?

See above.

>> The thing is that I only get the above log message from high traffic
>> (untaring NetBSD binary tarballs, or nightly find(1) runs).
>
>But the sense key and ASC/ASQ reported are quite clear, it's a hardware
>error,

I'm not sure how far I should trust the driver there.

>not even a protocol error on the SCSI bus (or these disks are
>braindead).

>It could be a power issue (power supply not strong enouth,
>for example).

The machines are decent 2HE rack-mountable boxes with 4 drive bays (2
used). One has a redundant PS, the other only one (450 W, IIRC). From the
router box I've gotten > 130d uptime.

As I said, the sister machines are running flawless under linux 2.6.x.

	hauke

--
/~\  The ASCII Ribbon Campaign
\ /    No HTML/RTF in email
 X     No Word docs in email
/ \  Respect for open standards