Clustered logical standby - Best practices?
Date: Thu, 31 Jan 2008 13:06:15 -0800 (PST)
Oracle 10.2.0.3 64 bit on SuSE 10 (2.6.16)
We are looking into implementing up a two node RAC logical standby database for a three node RAC primary database. We have this set up already as a proof of concept. We have standby redo logs and are using immediate apply.
My question is related to best practices. In a clustered physical standby environment Oracle advocates multiple instances receiving the redo, and one recovering it (http://download.oracle.com/docs/cd/ B19306_01/server.102/b14239/rac_support.htm#sthref2202). I didn't see anything for a clustered logical standby database, or at least the diagram in the link above didn't appear to reference it.
Has anyone seen (or developed themselves) a similar document for a clustered logical standby? When I looked at gv$logstdby, the instance that is performing SQL apply is performing all the processes (COORDINATE, READ, BUILD, PREPARE, etc.), and the instance that receives the redo from the primary cluster is doing just that.
Has anyone found this to be of any measurable benefit when using SQL
Apply? (given the standby redo logs are on shared disk in ASM, so the
I/O is the same regardless)
Are the CPU cycles saved on the reception of redo by the other node worth it?
Is there a best practices document to which someone can refer me?
Steve Received on Thu Jan 31 2008 - 15:06:15 CST