Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Two Oracle owners on one box .....

Re: Two Oracle owners on one box .....

From: Pete's <empete2000_at_yahoo.com>
Date: 17 May 2002 07:21:30 -0700
Message-ID: <6724a51f.0205170621.1c69b7a2@posting.google.com>


I don't see the logic either, my only thought about this is that you do not have enough disk on your failover system. Also, you are already maintaining two installations of Oracle. I have some thoughts for you as I too am in the middle of installing a 2-node IBM HACMP Cluster. Here's a relativly short and rough guide to what I am going to do:

  1. Install Oracle on each server on Non-Shared disk. The reasoning, in order to do any patching, you have to shutdown all oracle instances anyways(at least that's what Oracle wants you to do and is a good idea). In your situation, that means you have to totally shutdown the cluster on both sides or just shutdown the databases on the node that currently has the resources. If you shutdown the cluster instead, you then have to varyon the shared volume group(s) on one of nodes to do any kind of Oracle patching, kinda clunky if you ask me. No matter what you do, the secondary side of the cluster will need to be shutdown as you certainly don't want it to take over in the middle of a patch or an install.
  2. Create a shared area for your Admin directories on the shared disk in order to ensure that the all pfiles are up to date. Also, I want to add the listener file, but not the tnsnames file(you'd still want this available in order to connect to db's on the other node in the cluster), to this shared area as well.

The benefit to this during a patch installation is that I could shutdown the secondary side of the cluster, install any patches necessary, bring the node back into the cluster and have it do a graceful takeover, then patch the primary side of the cluster. Then bring it back into the cluster without a takeover. Then later one, you could plan i.e. set a time with your users for when the primary side will take over again so that you are running back on the primary side again. This way, you minimize the impact for how long the database will be down.

HTH and good luck
Pete's

Daniel Morgan <dmorgan_at_exesolutions.com> wrote in message news:<3CE3F5DD.8AD20BC7_at_exesolutions.com>...
> Patrick wrote:
>
> > I am doing the installations. What I have no control over is the fact
> > that they bought a massive IBM server and told us to move as many
> > databases as possible to this server. I fought this rather heavily,
> > but I ended up battered and bruised and still having to go through
> > with this. In addition to moving multiple databases to the new
> > server, they introduced a new mission critical application.
> >
> > We decided to implement HACMP as a high availability solution. A
> > send, much smaller, server was purchased for the HACMP fail-over.
> > Only the mission critical database will fail-over to the second
> > server. Now, to prevent any miscues with patch-installation, we
> > installed the Oracle binaries for the mission critical database on
> > shared disk. This disk will fail-over with the HAMP services. So, if
> > a failure happens, this disk will be unavailable on the primary
> > server.
> >
> > If the failure is fixed and the primary server is brought back up, we
> > don't want to have to failback the mission critical database during
> > production hours, but we also want to be able to re-start the other
> > databases on the primary server that aren't part of the HACMP
> > configuration.
> >
> > In any case, to make all of our monitoring scripts run correctly, we
> > decided to have two Oracle owners on the primary server. The
> > processes for one owner fail-over with the HACMP services, and the
> > processes for the the second owner always stay on the primary server.
> >
> > I just wanted to be sure that editing the /etc/oraInst.loc file was a
> > good idea.
> > I want to keep the installations totally seperate and to insure that
> > the two owners can never stomp on the others owners stuff. It osunds
> > like tjhis will work.
> >
> > Thanks for the input everyone.
> >
> > later ....
> > Patrick
> >
> > norwoodthree_at_my-deja.com (NorwoodThree) wrote in message news:<ba03e2c.0205152104.596693fa_at_posting.google.com>...
> > > BTW: Why are the users doing the installs? You are the DBA correct?
> > > YOU, as the DBA, should be doing the installs, the setup, and the
> > > maintenance. Why would you want to even maintain something you have
> > > not control over.
>
> What does moving multiple databases to a server have to do with multiple installations of the Oracle software?
>
> Daniel Morgan
Received on Fri May 17 2002 - 09:21:30 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US