RE: How to setup RAC to avoid Inter-Instance Parallelism

From: Bobak, Mark <>
Date: Wed, 10 Jun 2009 12:31:42 -0400
Message-ID: <>

See the first result from above search.

But, basically, if you want to restrict all parallel query activity to a specific node, then, on each node, set instance_groups to the instance name: inst1.instance_groups=inst1

Then to run on instance inst1, do:
Alter sesstion set parallel_instance_groups=inst1;

Should be pretty straightforward. This is a simple example. Value of instance_groups can be set to any number of arbitrary identifiers, so, a specific group could span a certain subset of instances. See the paper for more detailed examples.


From: [] 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

Hello list,

I have a two node 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.


-- Received on Wed Jun 10 2009 - 11:31:42 CDT

Original text of this message