Re: 9iRAC/AIX5.2 Installation Questions

"Daniel Morgan" <damorgan_at_x.washington.edu> wrote in message news:1079292263.284985_at_yasure...
> Howard J. Rogers wrote:
>
> > "Daniel Morgan" <damorgan_at_x.washington.edu> wrote in message
> > news:1079267739.527863_at_yasure...
> >
> >>A. Fuentes wrote:
> >>
> >>
> >>>Fellow Oracle Users:
> >>>
> >>>I will be installing 9iRAC on two AIX nodes.
> >>>
> >>>
> >>>1)Any pointers to web resources and guides?.
> >>>
> >>>2)Any pitfalls/advice?
> >>>
> >>>Thanks in advance,
> >>>
> >>>
> >>>A. Fuentes
> >>
> >>Keep the IBM consulting team, especially the DB2 ones, as far away
> >>from your checkbook and your hardware as you can.
> >>
> >>Be very sure to test that you have a clustered file system before
> >>proceeding to install anything from Oracle.
> >
> >
> > What if he wants to RAC with raw??
>
> Then there is not cluster file system to test.

My point exactly.

> > I think you mean 'test you have a working cluster before worrying about
the
> > Oracle stuff'.
>
> You are correct.
>
> > In which case, fair enough, though it is equally plausible to install
> > non-RAC Oracle executables onto two separate machines, cluster the
machines,
> > and then add in the extra RAC components (otherwise how would you ever
be
> > able to upgrade from single instance Oracle to multi-instance?).
>
> You can but if problems exist they are far harder to diagnose.

Problems can exist at any time with a RAC install. Bolting on the extra RAC bits on top of an already established and tested RAC-less install has proved (for me) to be far less of a hassle than going straight for the RAC-and-push method, numerous times.

> > Because of
> > the myriad problems you can get with the bit of the install where Oracle
> > replicates the executables over the interconnect to the second node, it
is
> > actually (IMHO) preferable to minimise the amount of stuff to be
transferred
> > in such a way precisely by doing a RAC-less install onto each machine
first.
> > Only the RAC-specific executables then get pushed across the
interconnect.
> >
> > Regards
> > HJR
>
> I would agree with you except that, at least with RedHat EL AS 3.0
> we have so far been unable to make it work. I'll grant it is a 3.0
> issue ... but an issue non-the-less.

What doesn't work? The upgrade from two RAC-less installs to a RAC one using the 'push across cluster' technique? It's worked fine for me in the past. What errors do you get?

Regards
HJR Received on Sun Mar 14 2004 - 14:40:46 CST