Re: ASM disk path in exadata

From: Frits Hoogland <frits.hoogland_at_gmail.com>
Date: Tue, 21 Feb 2012 13:07:24 +0100
Message-Id: <0EC154E4-E199-423B-95EA-3B04B7359B57_at_gmail.com>

Content-Transfer-Encoding: quoted-printable Content-Type: text/plain;

        charset=us-ascii

There is no udev/path translation on the database/computing layer in = exadata.

During startup, ASM first needs to find the disks. It does that by = picking up the cellserver information (addresses) in = /etc/oracle/cell/network-config/cellip.ora Because the discovery string is set to "o/*/*", it picks up all = griddisks available on the cellservers defined in cellip.ora

Because of the entirely different path, ASM knows it needs to = communicate over the network to get to the griddisks, instead of using a = locally attached disk. This combination is RDS (infiniband) on Exadata, = but can be UDP (ethernet).

In fact, the discovery can manually be done by using kfod: kfod asm_diskstring=3D"o/*/*", disks=3Dall this is what the asmca uses behind the GUI to discover disks, also on = normal platforms.
When specifying the o/*/* diskstring, kfod too reads = /etc/oracle/cell/network-config/cellip.ora to know where to look.

Frits Hoogland

http://fritshoogland.wordpress.com
mailto:frits.hoogland_at_gmail.com
cell: +31 6 53569942

On Feb 21, 2012, at 10:32 AM, gaurav mehta wrote:

> I have been trying to figure out how exadata griddisks are presented =
as candidate asm disks. We know that the disk discovery path for asm = instances using exadata is "o/<cell_ip_address>*/<griddisk_name>*". I = was wondering what actual path / udev in the operating system on the db = machine does this translate to ?. =20
> --
> http://www.freelists.org/webpage/oracle-l
>=20
>=20

--
http://www.freelists.org/webpage/oracle-l

Received on Tue Feb 21 2012 - 06:07:24 CST