ICH5R, ICH7R or Promise PDC20378 RAID?

[Ted Mittelstaedt]

----- Original Message ----- 
From: "Philippe Lang" <philippe.lang at attiksystem.ch>
To: "Ted Mittelstaedt" <tedm at toybox.placo.com>; "Doug Poland"
<doug at polands.org>
Cc: "freebsd-questions" <freebsd-questions at freebsd.org>
Sent: Wednesday, June 21, 2006 2:32 AM
Subject: RE: ICH5R, ICH7R or Promise PDC20378 RAID?


> Ted Mittelstaedt wrote:
>
> >>> On Mon, Jun 19, 2006 at 10:18:05AM +0200, Philippe Lang wrote:
> >>>> Hi,
> >>>>
> >>>> I've got two boxes that would be perfect for a small server. The
> >>>> motherboards are:
> >>>>
> >>>> - ASUS P4C800-E Deluxe, with ICH5R and Promise PDC20378
> >>>> - ASUS P5WD2, with ICH7R
> >>>>
> >>>> I intend to install FreeBSD 6.1, and I wonder which chipset you
> >>>> would recommend me for RAID 1? I need something rock-solid, of
> >>>> course...
> >>>>
> >>> I'm running -STABLE on a P4C800-E.  I've never had a problem with
> >>> the PDC20378.  Recently, I added two drives on the ICH5R with no
> >>> issues. A little benchmarking shows the ICH5R to be a little faster.
> >>
> >> Hi,
> >>
> >> I wasn't as lucky as Doug with my ASUS P4C800-E Deluxe board, both
> >> with the ICH5R and Promise RAID controller.
> >>
> >> 1) With the Promise controller, after simulating a disk problem
> >> (shutdown - disk disconnection - reboot - shutdown - disk
> >> reconnection - reboot), I wasn't able to rebuild the array at all,
> >> either with "atacontrol rebuild ar0" or from the bios. I suspect a
> >> spare disk is necessary for a complete rebuild.
> >>
> >> 2) With the ICH5R controller, installation went fine too. If I
> >> disconnect a disk, the system still boots, but when I plug the disk
> >> back (computer shut down), no boot is possible anymore. At boot, a
> >> kernel panic says: "softdep_setup_inomapdep". No array rebuild is
> >> possible from the bios, so I'm stuck.
> >>
> >
> > You need to take the disk you unplugged to some other machine
> > and wipe it, then make sure the one disk with the system left
> > on it is in the 0 position, put the wiped disk in the 1
> > position, then you should be able to boot.
> >
> > Your simulating a failure when you pull the one disk, but
> > your not simulating a recovery when you insert that disk back in.
> >
> > Ted
>
> Hi,
>
> I'm sure there are solutions to make things work, but personnaly, I
> don't feel confortable with a RAID system that acts that way, and that
> is supposed to make the whole system more robust and problem-free.

Your missing the point.

If a disk crashes it wipes the metafile off the disk.

If you want to simulate a disk crash, issue a raid detach command
first, then pull the disk.

The ata raid in the ata driver isn't written to handle the condition of
a disk being pulled that has nothing wrong with it then reinserted.
It is written to handle the condition of a disk crashing, then being
pulled, then a new one replaced.

> I used to play with hardware SCSI RAID controllers, and was never able to
> trash the system.

I have SCSI raid controllers too that I can do that with.  But keep in
mind that these controllers firmware is written to if a disk is pulled then
replaced, it immediately assumes that the disk is toast, and rebuilds it
from scratch.

> Arrays are being rebuilt in the background, that's
> just great. I hope this is the kind of ease and quality I will get with
> the 3WARE raid controller I have ordered... Apparently, everyone says
> very good things about 3WARE.
>

The only problem we had with ours is that the very latest firmware did not
work with the FreeBSD driver.  This was about 3-4 months ago though so
they might have fixed that problem.  We always automatically update
all cards we buy to the latest firmware before we install them for the first
time, that is why we hit that problem.

Ted