Re: Moving from single instance to RAC

  • From: "Syed Jaffar Hussain" <sjaffarhussain@xxxxxxxxx>
  • To: "Jeremy Paul Schneider" <jeremy.schneider@xxxxxxxxxxxxxx>
  • Date: Thu, 31 May 2007 20:09:17 +0300


This afternoon we have moved our databsae from the single instance to two
node RAC. Initially we planned for extended RAC, which will  have a single
node at head office premises and the other one would be in the DRC which is
less than a km away from the head office. When we were on testing we found
out that frequently cluster break downs on the DRC as we use satellite
connection between head office and DRC due the crane(uses for building
construction) movement, a building is being constructing next of our office.
We changed the IDEA and created two nodes at the head office using different
Today the activity was successful with couple of minor problems. We need to
wait until Saturday to know the real issues as today and tomorrow is
week-end for us.
Currently we plan not distribute the load equally between two nodes. We are
allowing 20% of connection from one node and 80% from another node to
monitor the RAC behavior.

Thank you all for your valid recommendations.

On 5/31/07, Jeremy Paul Schneider <jeremy.schneider@xxxxxxxxxxxxxx> wrote:

ASM does require RMAN, but if you do RAC with OCFS then it's still
possible to do user-managed backups.  But yeah, RMAN has some great features
and there aren't many reasons left to not use it...

The big question would be how well does your OLTP app scale on SMP...
generally speaking apps that run well there will run well on RAC.  (There
are of course exceptions.)  If you have small concurency problems now that
have escaped notice then they could become much bigger on RAC once you
introduce cache fusion and global locks.

You also mention "stretched cluster" - do you mean an extended distance
cluster, where you will have cluster nodes at different physical sites that
are miles apart from each other?  There are some special issues to be aware
of for those deployments...  interconnect speeds, data access paths, and
quorum/voting disk layout being among them...  I'd be cautious about doing
an OLTP workload on an extended-distance cluster, but I'm sure that you guys
are planning it carefully and testing like crazy.  :)


On 5/31/07, Matthew Zito <mzito@xxxxxxxxxxx> wrote:
>  There's no requirement to use RMAN if you're using RAC, though Oracle
> is definitely encouraging you to use RMAN universally (not just for RAC).
> Matt
>  ------------------------------
> *From:* oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
> *On Behalf Of *Peter McLarty
> *Sent:* Wednesday, May 30, 2007 3:53 AM
> *To:* sjaffarhussain@xxxxxxxxx; oracle-l
> *Subject:* RE: Moving from single instance to RAC
> You have to use RMAN
> We are planning to move one of our highly OLTP Oracle 10gR2 database on
> AIX from single instance to 2 node RAC.
> I just would like hear from any of one here about the issues or
> potential problems when moved from a single instance to RAC.
> I would be gald if you guys share you experince or throw some limelight
> or tips.

Jeremy Schneider
Chicago, IL

Best Regards,
Syed Jaffar Hussain
Oracle ACE
8i,9i & 10g OCP DBA
"Winners don't do different things. They do things differently."

Other related posts: