RE: Server Architecture

From: Mark W. Farnham <>
Date: Thu, 3 Jan 2008 15:51:08 -0500
Message-ID: <00bb01c84e4a$5ec73880$>

Without addressing the separate owner part of the thread, if you are indeed providing for sparse individualization of ORACLE_HOMEs to accommodate, for example, varying uptime service level requirements, and especially to avoid the timeliness of oratab edits and memo-passing for multi-DBA, sysadmin, and operations teams, it can be most useful to have a symbolic link as the listed ORACLE_HOME with one link per database. Then the atomic action of updating the link maintains coherency, unless you have multiple DBAs simultaneously trying to work on the same database without real time coordination, and that is a different problem entirely introducing unknown orders of magnitude greater chaos into the equation.  

Part of the thread started using the word "instance" in place of "database" - at least that is what I got from a speed read of the thread - let's please avoid that. Rare indeed is the patching of an instance.  



From: [] On Behalf Of Matthew Zito
Sent: Thursday, January 03, 2008 12:38 PM To:;; Oracle L Subject: RE: Server Architecture  

Well, that can get messy with a) remembering what databases are in what homes, and b) naming conventions and the like. It's useful to be able to say, "I know what the home is purely by virtue of the instance name" or something similar. Or if someone moves a database over and another DBA doesn't get the memo, and it gets started out of the wrong home or what have you. Badness can result.  



From: [] On Behalf Of Dan Norris
Sent: Thursday, January 03, 2008 11:26 AM To:; Oracle L
Subject: Re: Server Architecture  

If they're all on the same patch level today, why introduce all the ORACLE_HOMEs today? Personally, I'd keep them all on the same ORACLE_HOME now and when one wants to patch and the others aren't ready, spin up a new ORACLE_HOME, patch it, then migrate the DB in question to use the new ORACLE_HOME at that point. That way, if you don't run into patching conflicts, you'll never need the extra ORACLE_HOMEs. Is there something wrong with my logic?


Received on Thu Jan 03 2008 - 14:51:08 CST

Original text of this message