Start a Conversation

Unsolved

This post is more than 5 years old

1284

August 24th, 2011 03:00

RAC protection with MV/S and DataGuard

Current environment, two Linux servers running with RAC on a CX4 480, it uses ASM

Expected protection,

Setup storage local mirror with a CX4 240 and MV/S to quickly start RAC on CX4 240 when primary storage CX4 480 failed (I believe this part is ok)

Using spare Linux server (hereinafter referred to as Server C) with interal disks as DataGuard Target of RAC system. (I believe this part is ok)

Using Server C with RAC data replica on CX4 240 to start DB service when the whole of RAC and CX4 480 failed. ( I have couple of questions about this)

Is a standalone server able to mount a replica of RAC system?

Is a physical server able to hold two instances, one for DataGuard Target, the other for mounting a replica of RAC system?

Thanks,

Eddy

98 Posts

August 24th, 2011 06:00

Hey, Eddy!

The fine folks over at DBA Oracle have instructions to set up a replica of a RAC cluster on a stand alone server.  Zip on over to http://www.dba-oracle.com/real_application_clusters_rac_grid/single_instance.htm.

After reading through it I bet that most of the activity can be scripted within Replication Manager.

Rob

225 Posts

August 24th, 2011 20:00

Thanks for your reply, this article is good material talking about setup standby single node instance to a RAC with DataGuard.

But I like to know if RAC data replica on CX4 240 after fracturing MV session is able to be mounted on a single node instance, and which is able to co-exist with DataGuard target instance on a single physical server (note, customer like to run MV and DataGuard at same time to protect their RAC system.)

Eddy

98 Posts

August 25th, 2011 06:00

Okay.  In theory it would be possible but it'd be quite a bit of trouble to set up.  Here's the issue:  The dataguard and MV/S replicas are going to have the same db name and the same ORACLE_HOME based upon best practice.  Alas, you can't have two database instances on the same server using the same ORACLE_HOME directory.  (This is per the Oracle documentation and specifically mentioned in the Oracle Forums at http://forums.oracle.com/forums/thread.jspa?threadID=1133868 ).  With that in mind one could argue that if you could reconfigure ALL of the references to the ORACLE_HOME variable in one of the replicas you'd be fine.  I imagine that'd be quite a bit of work to script.

Here's my question for you:  Why?  If server C can handle two instances then why not virtualize it, put on two guest OSs and then run the instances independently.  Off the top of my head I can't see any reason to have them both running concurrently.  Thoughts?

Rob

225 Posts

August 25th, 2011 19:00

Thanks for your reply about two instance co-existing on a physical server.

About another my question, I read some material. For now here is my understanding, I like you guys to look at it, any comment is appreciated.

Every single instance in a RAC has its own SGA, background process, Redo log and Undo Table space, therefor the number of RAC instance on Source / Target sites has to be same, if RAC on Target site is able to take over the Storage replica when fail-over occurs. On other word, Target should have enough instance to take the Redo log, undo space replicated to target site.

Am I right?

To your question, why not a virtualization solution for this scenario?

First, it is not a design from fresh, It is about how to make it run under a fixed configuration.

From my respective, I full agreed with you about virtualization solution, even I would consider VPLEX Local ( in this case, all requirement in a same DC)

My design would be,

Three servers are running ESX servers.

RAC is running VMs on ESX servers and VMware HA to Server C.

VPLEX Local w/ two CX4 array are setting up local mirror device to achieve a high-availability storage environment.

No Events found!

Top