[Oracleasm-users] Oracle ASMLib and IBM SDD

[Joel Becker]

On Tue, Jun 28, 2005 at 10:22:13PM -0400, Sheridan, Matt wrote:
> We're trying something similar. We have IBM x335's connected to an ESS
> (Shark) 750. We're running Red Hat ES 3.0 Update 4 and IBM's SDD. We
> were getting the same "Device <blah> is not a partition" when trying to
> run oracleasm createdisk. We found a way past that. It's kludgy, but
> seems to work. We ran the createdisk command using the underlying sd
> (ie. /dev/sdd1) instead of the vpath (ie. /dev/vpatha1). That created
> the asm VOL# successfully. We then set the ORACLEASM_SCANORDER parameter
> in /etc/sysconfig/oracleasm to "vpath sd" so the vpath would be picked
> up and used upon reboot. After rebooting, an oracleasm querydisk of that
> disk shows it using the major/minor #'s for the vpath instead of the sd.

	Very well done.  

> Could we be hitting another GETGEO issue, just in a different section of
> the code? Any way to strace the sqlplus command to verify this?
> (Joel/Heath: did you ever find a fix for your error?)

	No idea what the I/O error is, I think IBM is looking at it.
GETGEO is never called outside of createdisk.  Older ASMLibs had a bug
allowing full-disk createdisk, so we allow them to continue.  We
only do the verification at creation of a new ASM disk.

Joel

PS. Oracle considers full-disk a bug because it makes it really easy to
fdisk over your data.

-- 

"Against stupidity the Gods themselves contend in vain."
	- Freidrich von Schiller

Joel Becker
Senior Member of Technical Staff
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127