RE: Disk level cloning and ASM
Date: Mon, 2 Mar 2009 16:53:31 -0500
Message-ID: <6AFC12B9BFCDEA45B7274C534738067F0D9722C3_at_AAPQMAILBX02V.proque.st>
Just a quick follow-up to summarize what I did:
First, I could have just left it, and ignored it. This would have been a bit sloppy though, as I would have ended up w/ a separate set of directories under both 'pqprd' and 'pqpep'. This would be annoying, and just the sort of thing that comes back to bite at 4:30am in the middle of a crisis a year from now. So, at a minimum, I want it to be consistent.
So, it turns out that the parameter that governs what database name will be used in the ASM paths is 'db_unique_name'.
Hrishy pointed out an interesting posting on Alejandro Vargas' blog(http://blogs.oracle.com/AlejandroVargas/2008/01/asm_database_clones_naming_iss.html), that talks about this. He suggests that you either set up a database in advance, w/ db_unique_name set. In my case, I have pqprd(production) and pqpep(preprod). So, I could create pqprd from scratch and set db_unique_name to, for example, 'pq'. Then, all my paths would be of the form:
+PQ_DATA_DG/pq/.....
And, when I cloned pqprd, I could change the db_name to pqpep, and w/ db_unique_name set to 'pq', all my files would still go to the same place. That's great if you're setting up brand new databases with with type of cloning in mind. But, I've already got a production database that I can't just go re-creating.
The other thing he suggests on his blog is to use RMAN to move the files from one directory structure to another. This is also a good suggestion, but, it entails using RMAN to copy all datafiles. The problem here is, this could be quite time consuming.
So, what I did, in the end, is compromise. I left all the files under 'pqprd', and I set the preprod db_unique_name=pqprd. So, all new files in preprod will be created under '+PQ_DATA_DG/pqprd/'.
So, ultimately, it turned out to be a simple fix. It's a little bit sloppy, that pqpep database files are under pqprd folder in ASM, but, it was easy to implement, and should not be too much of a hassle.
Thanks to Hrishy for the pointer to Alejandro's blog, as well as to Alan, Mark, and Matt, for their comments.
-Mark
-----Original Message-----
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Bobak, Mark
Sent: Sunday, March 01, 2009 8:57 PM
To: oracle-l_at_freelists.org
Subject: Disk level cloning and ASM
Hi All,
I've got two 10gR2 databases running on ASM. One is pqprd (production), and one is pqpep (preprod).
I've recently cloned pqpep from pqprd. This was done at the storage level, by copying all of the raw devices and presenting them to the preprod hosts. (Should have mentioned, both prod and preprod are 3-node RAC.)
My problem is, now the filename paths are of the form:
+PQ_DATA_DG/pqprd/*
because they were copied from prod.
They should be:
+PQ_DATA_DG/pqpep/*
So, is there a way to deal with this?
Do I use 'nid'? Will that take care of it? Some other method?
Any help is appreciated....
Thanks!
-Mark
-- http://www.freelists.org/webpage/oracle-l -- http://www.freelists.org/webpage/oracle-lReceived on Mon Mar 02 2009 - 15:53:31 CST