Re: HACMP Configuration

The HA configuration, aside from the use of the DLM and a few required patches/extras, will be very similar between version 7 and 8 of Oracle ( disclaimer: I've no experience with 8i and HA clusters, so... ).

If you do not plan to run OPS, I would steer well clear of running HACMP in Concurrent mode. Stick to the vanilla setup where you have a pool of shared disks ( these are connected to all nodes ), a set of resource groups containing the resulting Volume Groups, and a primary and secondary node for your Applications.

So, a minimal setup might have a single Volume Group (VG) defined across your "common" pool of disks. This volume group might contain everything required for your Oracle database ( datafiles, redo, undo, archived redo, control files, etc ) in both cooked filesystem and raw formats. The Oracle software tree could be shared, I suppose, but I wouldn't recommend it. You'll certainly need an OFA setup for all this to work properly.

Assuming a two node system, this VG is created and configured on only one of the nodes ( node A let's say ). Once everything is up and running there on Node A, you can then shut everything down and get the VG in a varied-off state. From there, an import followed by a varyonvg on Node B will make all the files available to that node.

Only one node will have the VG or VG's "mounted" at any given time. If that node fails, one of the remaining nodes ( or maybe the only surviving node ) will vary the VG on ( "mount" it ) and start up the database(s), listeners, and any external, non-oracle applications required. The new owner of these resource will also assume the IP address of the failed node.

Once the failed node is ready to re-join the cluster, depending on how things are configured, it will either become the "standby" node, or it will reacquire all of its resources from the takeover node.

This is very simplistic and certainly doesn't cover everything, however it should give you a good overview of how things might look.

Hth

-Kevin

"Ethan Post" <Blah_at_Blah.com> wrote in message news:1F6s7.32583$6c5.1309999_at_news1.rdc1.sdca.home.com...
> Thanks Kevin, I have a few RedBooks from the site and they cite Oracle 7
in
> thier examples. Yes I will be going with a standby configuration. I
having
> some trouble grasping the shared volume groups. Does this mean both
> machines have access to the same volumes or does one machine mount the
> volumes when a take over is required, or can it be either one? Will keep
> looking for info and read all the docs. As usually they want this done
> yesterday. - E
>
> "Kevin Brand" <kevin.brandx_at_tel.gte.com> wrote in message
> news:9oqns0$bu3$1_at_news.gte.com...
> >
> > IBM has some "redbooks" on their site that deal specifically with
setting
> up
> > the HACMP cluster for ORACLE use.
> >
> > If you plan to run Concurrent Access disk subsystems in HACMP with
Oracle
> > Parallel Server, you're in for quite a struggle if you are not familiar
> with
> > both. However, if you plan to simply use HACMP to make a database
> available
> > after a node failure, the process isn't quite as complicated or as hard
to
> > maintain.
> >
> > The HACMP Administration and Installation Guides/References should
> > definitely be required reading before attempting to setup anything.
> >
> > Good Luck,
> >
> > -Kevin
> >
> > "Ethan Post" <Blah_at_Blah.com> wrote in message
> > news:tP4s7.32271$6c5.1299287_at_news1.rdc1.sdca.home.com...
> > > I have been tasked with configuring HACMP failover for some databases.
> I
> > am
> > > looking at a white paper on it right now that is really dated. They
> also
> > > assume the databases do not exist yet. If anyone knows of some good
> > > resources or would allow me to give you a call and get a better
overview
> > of
> > > HACMP I would greatly appreciate it.
> > >
> > > Thanks,
> > > Ethan Post
> > >
> > >
> >
> >
>
>
Received on Tue Sep 25 2001 - 18:22:53 CDT