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: HP MC / Service Guard / Oracle 8i

Re: HP MC / Service Guard / Oracle 8i

From: Jim Krol <b0432263_at_boeing.com>
Date: Thu, 27 Feb 2003 20:09:36 GMT
Message-ID: <3E5E7080.975AE4F@boeing.com>


Christian Hartmann wrote:
>
> On Thu, 27 Feb 2003 15:47:31 GMT, Jim Krol <b0432263_at_boeing.com>
> wrote:
>
> >> Did I understood it correctly, that when I use Oracle along with HP's
> >> MC / Service Guard I have one running instance on one machine and one
> >> not running instance (same init-file) on a scond-machine connecting to
> >> the same disk-devices? And if the primary instance goes down the
> >> ServiceGuard then starts the second instance on the second machine?
> >>
> >
> >You seem to be confusing Serviceguard/OPS and regular serviceguard. Without OPS
> >you must create an oracle package, that contains all that is necessary to run
> >that package. The package needs to be on shareable disk, and should be wholly
> >contained on that disk. When smoke starts to billow from the machine that is
> >running the package, the package will be switched to another machine in the
> >cluster. This is disruptive. The cluster will restart the oracle package on the
> >new machine, and the applications will have to reconnect to the package. The key
> >element here is that the disk is shared by the cluster systems, but only ONE
> >system has write access to the disk.
>
> Well Jim, first of all, thank you for the fast and detailled reply.
>
> I am going to use Oracle without OPS/RAC.
>
> One further question: Am I correct that the second server is running
> but hasn't start the oracle-instance until ServiceGuard tells him to
> do that?
>

The second server is not involved until failure. We currently have three oracle servers in an regular Serviceguard cluster. We have a group of instances running as a "package" on each server. Two of the servers are production one is a test server. The package has its own IP address and DNS entry. Your tnsnames.ora points to the package name and the application uses that to connect to the instance. During a failure serviceguard discovers that the oracle server is no longer there. The backup server then activates the oracle vg's on the backup server, brings up the oracle instances, and uses arp to point the IP address of the package to the backup server. The applications will reconnect to the package name which now points to the backup machine. The package name acts just like a server name. You use tns to point to DBpackage rather than DBmachine.

> Can you give an example for that "package-thing"? Do you mean that I
> after I connected to an oracle-instance I always have to call the
> package within each statement? Can't believe that!

Its transparent to the user. The package emulates a physical machine. A package is just a definition of all that is necessary to run an oracle instance(s). The package runs a script to activate all the disks that oracle needs. It then runs scripts similar to dbstart.
>
> Regards,
>
> Christian
Received on Thu Feb 27 2003 - 14:09:36 CST

Original text of this message

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