[rescue] E5000 disk issue

[Francois Dion]

On 11/25/05, velociraptor <velociraptor at gmail.com> wrote:
> On 11/24/05, Dan Williams <williams.dan at gmail.com> wrote:
> > It might not be the disk, have you changed the termination. I have had
> > similar errors (although without the write error) which has been
> > termination.
>
> We have also seen that type of behavior on SCSI with a "marginal"
> SCSI controller (replaced), and on FC with a bent-too-sharply
> FC cable.

I got some SBUS SCSI controller I can try, but I might as well update
my SBUS IO board (SOC, not SOC+) and see if that makes any difference.

> If the errors are  isolated to a single disk, though, I'd try removing it
> first before fiddling with other bits.

I removed it and got good performance for a while. I added another
36gb disk and for a while I had decent performance, but again, it is
slow.

>  I hope you have some kind of
> LVM RAID at least?

Bah, no raid. I dont have enough disks to mirror anything but the boot
disk. I can restore from UFSDUMPs pretty quickly tough. If it was a
production box, I would have raid 10, but this is one of the servers
in my basement. I'm just playing with this right now. Eventually I'll
add an A5200 or something like that.

Of course, it complicates things that this is mounted thru a zone. I
might shutdown the zone and mount it on the global zone and see what
happens there.

Is there a script or app somewhere that has a standard throuput test?
I know about iozone but I have nothing to compare against.

Ciao,
Francois