RE: 10g RAC Install with Raw Devices

  • From: "Peter McLarty" <p.mclarty@xxxxxxxxxx>
  • To: <tcarlson@xxxxxxxxxx>, <oracle-l@xxxxxxxxxxxxx>
  • Date: Thu, 29 Mar 2007 10:04:02 +1000

Hi Todd 
Our experience with this sort of problem has entailed us rebooting all
the nodes in the cluster usually a round robin event, 1 after the other
as when we do ASM seems to be resolving it. 


We are thinking the perm_disk section of oracleasm in /etc/init.d does
the fixing.

Are you using ASM or GFS for this installation.

We have Oracle looking at this at present also HP as we have EVA SANS
here the EVA3000 is providing our storage



Cheers

Peter 

-----Original Message-----
From: Carlson, Todd [mailto:tcarlson@xxxxxxxxxx] 
Sent: Thursday, 29 March 2007 12:15 AM
To: oracle-l@xxxxxxxxxxxxx
Subject: 10g RAC Install with Raw Devices

Hey Guys,

I am doing a RAC install on an x86_64 RH4 box. We are using 10.2.0.1 and
raw devices. The SAN is a MSA 1500. The issue that I am having is during
the Cluster install. The installer is complaining that all nodes can't
see the raw device location that I point it to for the OCR. I can dd
from all nodes to the device. Has anyone seen this before? Does anyone
know what test Oracle is using to determine connectivity? Are there any
other options that I should try before creating an SR?

hostName = grid0.hprsn.com
INFO: Query Returned: grid0
INFO: Calling Query fileQueries10.1.0.2.0  exists

filename = /dev/sda2/
INFO: Query Returned: true
INFO: Calling Query ClusterQueries10.2.0.1.0  IsItDir

PathName = /dev/sda2/
INFO: Query Returned: false
INFO: Calling Query cfsprereqQueries10.2.0.1.0  isOHonCFS

oraclehome = /dev/sda2/

nodeList = grid0,grid1,grid2,grid3,

localNode = grid0.hprsn.com
INFO: 
Query Exception: RuntimeException
Query Exception Class: class oracle.sysman.oii.oiil.OiilQueryException
INFO: false
INFO: Calling Query ClusterQueries10.2.0.1.0  IsItRaw

PathName = /dev/sda2/
INFO: Query Returned: OPENFAILONDEV
SEVERE: The location /dev/sda2/, entered for the Oracle Cluster Registry
(OCR) is not shared across all the nodes in the cluster. Specify a
shared raw partition or cluster file system file that is visible by the
same name on all nodes of the cluster.

Thanks!

Todd

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


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


Other related posts: