RE: How to setup RAC to avoid Inter-Instance Parallelism
Date: Wed, 10 Jun 2009 13:02:50 -0400
For a quick test you could set up connect identifier for instance 1 on your tnsnames.ora and point it to IP/name of that one node. Set the environment to that identifier and run your query. You may need to set up a separate listener service with the name of the connect identifier on the server node.
If your query goes to just one node and the behavior persists, there may be some node interconnect issues.
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of GBA-DBA
Sent: Wednesday, June 10, 2009 12:11 PM
To: Oracle Discussion List
Subject: How to setup RAC to avoid Inter-Instance Parallelism
I have a two node 10.1.0.5 RAC setup
serv1 - instance 1
serv2 - instance 2
We are using parallel query for batch processing. The time to complete batch is ok when we run in a non clustered environment (that's what we have in dev and integrated testing) but when we move the batch to a clustered testing environment similar to prod everything takes 2 or 3 times longer.
I have noticed that some of the slave processes go to instance 1 and some of them go to instance 2. I would like to confine the parallel slaves to one instance but I don't fully understand how to setup the instance_groups and parallel_instance_groups to achieve this.
Could anyone help me please?
Thanks in advance.
http://www.freelists.org/webpage/oracle-l Received on Wed Jun 10 2009 - 12:02:50 CDT