nomenclature standardization: where best to put oraInventory?

  • From: "Mark W. Farnham" <mwf@xxxxxxxx>
  • To: "Oracle-L" <oracle-l@xxxxxxxxxxxxx>
  • Date: Tue, 10 Aug 2004 12:35:42 -0400

Level setting:

I'm writing this in (L)unix context, provide your own translation to
convicted monopolist based systems.

FTFM: One oraInventory per node (aka system, aka cluster iff you've got
cluster shared software), pointed at by /etc/oraInst.loc

FTFM: You can have multiple ORACLE_BASE installations on a node. (For
example to be managed by different owners and/or groups).

FTFM: "Oracle recommends that you choose the following path:

/oracle_base/oraInventory

So, clearly this does not bind very well, and is subject to disaster, if,
for example, a project based ORACLE_BASE is relocated to a different
machine. Even if the obvious opportunity for disaster does not occur, you
are left with keeping an obsolete ORACLE_BASE just to hold oraInventory.

Further, if you have multiple oracle_bases in the first place to separate
ownership, the additional owners of oracle_bases will have to be able to
write there. (I don't have a solution to that problem, btw, even if
oraInventory lived in a place not bound to any oracle_base.)

Now I suppose changing the contents of /etc/oraInst.loc could be used to
relocate oraInventory, although it is not clear to me that is a supported
workaround.

So I'm thinking that locating oraInventory subsidiary to any individual
ORACLE_BASE is a crock, and I'm toying with alternatives:

/opt/oraInventory
/node/oraInventory

What all y'all think?

I would also be thinking along these lines for the CRS Home directory,
except that FTFM (B10766-02, p 5-20),

"The directory that you specify must be a subdirectory of the Oracle Base
directory." [In reference to CRS Home]

Now in the case of the free form grid architecture, this really gives me
pause. So let's say I have node X, that for purposes of grid services may
deploy instances supporting database A and database B which are managed by
different DBA groups. There is only one CRS Home per node, and it must be in
"THE" Oracle Base directory. (emphasis on "THE" is mine.) Now how the duel
is arranged between group A DBAs and group B DBAs to allocate node X to them
and bring up their instance(s) is not the current topic. Presumably this can
be worked out peaceably.

However, will it even work to have CRS in a different ORACLE_BASE? Obviously
the overhead to test this is moderately high, so if anyone has already
undertaken the complexities of nodes belonging to multiple grids (aka RACs,
but really I think this only becomes an issue in 10g when you can actually
grid up (RAC) nodes that are not members of the same cluster.)

Has anyone cooked up anything good for locating CRS Home?

Thanks,

mwf


----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to:  oracle-l-request@xxxxxxxxxxxxx
put 'unsubscribe' in the subject line.
--
Archives are at //www.freelists.org/archives/oracle-l/
FAQ is at //www.freelists.org/help/fom-serve/cache/1.html
-----------------------------------------------------------------

Other related posts: