RE: dataguard and operators
Date: Tue, 29 Sep 2015 14:31:00 +0000
Message-ID: <CY1PR0201MB1049E64852D8F51EFE3BDEA9D44E0_at_CY1PR0201MB1049.namprd02.prod.outlook.com>
I find it interesting how many people are working on similar projects. In our case we have a project that involves 11 different RAC dbs that support applications which are integrated, which means if one RAC is switched over (or failed over) all other 10 need to do the same. One of the dbs is a destination of GG replication, adding a bit more complexity. 3 different EM repositories are involved monitoring these systems and it was decided there wasn't enough time to have primary and standby's configured on the same EM (plus EM doesn't have the ability to create an EM Job off of built-in functionality, as in using EM's switchover in a job and assigning all 11 targets). As a last *special* feature, the broker and Oracle Wallet under 11g don't play nice so 10 dbs have switchover procedures using the broker, one using sqlplus.
For the above management wants a switchover done ASAP for releases (app, hw, oracle,...), ideally one command, one button, as simple as possible, similar to what you described. I ended up creating scripts for various steps that could be submitted as EM Jobs for all targets, then documented the steps to/from the standby along with various errors they may run into (and also details of the scripts and EM Jobs too). Through this process I found that it's very, VERY important to make sure all dbs involved are consistently configured. Meaning if possible make sure that you come up with and follow standards for items as basic as upper/lower case for broker names, that init.ora settings for primary are the same on the standby (since the standby could serve as the primary at any given moment), to name a few. I found out the hard way that sometimes DBAs were creating RACs with the db name, instance name, unique name in upper case and sometimes lower case (our standard is lower case). They didn't think it mattered but when involving the broker it does. Within the broker they used upper case, sometimes lower case, sometimes not even matching the case of the unique name/instance name. When in dgmgrl if the db name or instance name is defined as upper case you'll have to quote it. Not a huge deal unless you're scripting this process and suddenly have to dynamically match case. This can also have an impact with the global_name set in listener.ora as it needs to match EXACTLY (even with case) the static identifier for the broker (this is all 11g, btw).
Sorry for dumping so much but I hope some small nugget can be of help.
Regards,
Dave
-----Original Message-----
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Howard Latham
Sent: Tuesday, September 22, 2015 5:11 AM
To: ORACLE-L
Subject: dataguard and operators
Oracle 11 Redhat 4E
My boss wants scripts /instructions so 'any fool ' can do a dataguard switchover. I think that's hard if not impossible as you can't predict what is going to happen How many of you have successfully handed this over to an operator? And if you have can I see example scripts please?
--
Howard A. Latham
--
http://www.freelists.org/webpage/oracle-l
†Ûiÿü0ÁúÞzX¬¶Ê+ƒü n– {ú+iÉ^ Received on Tue Sep 29 2015 - 16:31:00 CEST