From: "Ian C. Blenke" <ian@blenke.com>
Date: Fri, 7 Sep 2001 12:37:28 -0400
To: Simon Matter <simon.matter@ch.sauter-bc.com>
Cc: Bret Hughes <bhughes@elevating.com>, linux-xfs <linux-xfs@oss.sgi.com>
Subject: Re: compaq smart 2 raid and XFS 1.0.1 redhat install problems
Message-ID: <20010907123728.A10983@blenke.com>
References: <3B988116.E31FAF1C@elevating.com> <3B9893DB.F59296E@ch.sauter-bc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <3B9893DB.F59296E@ch.sauter-bc.com>
User-Agent: Mutt/1.3.18i
Sender: owner-linux-xfs@oss.sgi.com
Precedence: bulk

On Fri, Sep 07, 2001 at 11:31:07AM +0200, Simon Matter wrote:
> > First I tried an upgrade that appeared to be fine but lilo and I got
> > confused ( at least I was ) I told the installer to use the MBR and that
> > is where compaq utilities stuff gets started from.  It was still booting
> > the 2.2.3-16 kernel.  Very weird.  I putzed with it for too long trying
> > different iterations of upgrades (both XFS and regular redhat 7.1)
> > trying to figure out what was happening.  My partner found the deal
> > about compaq needing the mbr so I did the old dos diskette and fdisk
> > /mbr.  This let me boot to the right kernel but the system hung right
> > after the compaq smart 2 initialization and gave the message checking
> > partitions:
> > and a reference to using ida/c0d0

I'm using two 1850Rs with SmartArray2 controllers with no problems. Both
are running Debian 2.2 (Potato) with handbuilt 2.4.6 XFS kernels, the CVS 
XFS cmd tools, and lvs 0.9.1_beta6 (all built from source).

When building the servers, I used the SmartStart 4.70 (from memory) wipe,
configured the RAID array, and installed the SmartStart partition. From there
it took a little hand holding to get Potato on the system.

I'm using grub (the older build from potato no less - what fun), not lilo.

The biggest pain was getting lvm's vgscan to deal well with the 
/dev/ida/c0t0d2 style device naming (ended up with a symlink to /dev/sda 
devices to get it to act consistently). If you're using a I2O device,
good luck (/dev/i2o/hda is even more obtuse).

In the end, I ended up building everything as a module and hacked up
/sbin/lvmcreate_initrd to include xfs/xfs_support/pagebuf, cpqarray, scsi_mod,
aic7xxx, and sd_mod (the latter three for the onboard controller).
Now I can build a new kernel, run lvmcreate_initrd, and avoid hand-building
an initrd everytime.

> > I figured I had screwed up the packages so I did an install (not an
> > upgrade ) to an unused partition and get the same error.  I have worn
> > out my mouse and me searching the mailing list archives for references
> > to what might be happening.  I tried various things like linux
> > devfs=nomount at the lilo prompt. No joy.

I'll not get into religious debates about RedHat's RPM hell on this list ;)
Been there. Done that. Burned the t-shirt.

While I've built the CVS xfs cmd tools as DEBs, it's just not proper to
pass them around ;) "Use the source". Keep it handy.

> > Are ther issues with the Compaq smart-2 raid controllers and XFS?  I see
> > that there is some patching that went on but in reference to it but I
> > could not tell exactly what the issues are.  do I need to do the CVS
> > deal and compile my own kernel?  How do I get it onto a machine that I
> > cannot boot? Please help.  I am fried.

None. I have two systems working wonderfully now: one with ~27GB (raid5),
and the other with 54GB+44GB (raid5). No worries.

You may *want* to build from CVS (I do, habitually now).

I'm running two VA Linux boxen with XFS/LVM, and two Compaq 1850R boxen
on SmartStart2 controllers with XFS/LVM, as well as a couple of happy XFS 
laptops. I've had incredibly few problems with XFS so far.. and 
lvextend/xfs_growfs really is a lifesaver.

Good luck.

- Ian C. Blenke <ian@blenke.com>