One of the interesting aspects of working at my current site is that the DBAs don’t install Oracle. Yes, I know, and I’m not going to elaborate on that statement other than to say it’s a little frustrating and software management has become a much more interesting experience than I ever imagined it would be!
I wanted to mention a particularly interesting moment today, though. When I was sitting beside ‘the installer’ last week, I was perturbed when we were trying to install 9.2.0.1 and the Universal Installer prompted us to select additional nodes that were part of this cluster. Given that we don’t use RAC at all, I was worried. Still (and more fool me) I let it go and we carried on, leaving just the current node selected. I should have known.
When someone attempted to STARTUP MIGRATE the first database to be patched today, this is what happened.
SQL> startup migrate
ORACLE instance started.
Total System Global Area 2520746368 bytes
Fixed Size 743808 bytes
Variable Size 2483027968 bytes
Database Buffers 33554432 bytes
Redo Buffers 3420160 bytes
ORA-32700: error occurred in DIAG Group Service
At which point they called me over. Mmmm, looks like RAC! Sure enough ….
SQL> quit
Disconnected from Oracle9i Enterprise Edition Release 9.2.0.8.0 - 64bit Production
With the Partitioning, Real Application Clusters, OLAP and Oracle Data Mining options
JServer Release 9.2.0.8.0 - Production
A little more investigation on Metalink revealed this (from several Metalink notes, but this was from 368916.1) …
“The AIX server is not part of a cluster but has the high availability HACMP software installed.
The Oracle Universal Installer (OUI) detects this and assumes RAC is being installed, even if it is not.”
Time to relink the server without RAC, but that wasn’t successful either. Next blog, maybe …
It must be detecting whether 3rd party clusterware is present based on certain files present just like in 10g.
Couple places to check:
– installer log file can reveal what file it looks for.
– take 10g RAC install and check init.cssd script (I think it’s the same for all platforms so on Linux – /etc/init.d/init.cssd); there are cases for different platforms and CRS starts differently with 3rd party clusterware. I think that detection mechanism could be the same.
Removing those files temporarily should make installer do non-RAC install.
Hth!
Thanks for the suggestions.
AIX is a little different, at least here. All of the startup is in /etc/inittab, which has a series of script calls in there and there’s no /etc/init.d The same principles apply though. e.g.
$ grep -i ha inittab
brc::sysinit:/sbin/rc.boot 3 >/dev/console 2>&1 # Phase 3 of system boot
rcec:2:wait:/etc/rc.ether start > /dev/console 2>&1 # configure etherchannel
ha_star:h2:once:/etc/rc.ha_star >/dev/console 2>&1
hats:2:once:/usr/bin/startsrc -g hats > /dev/console 2>&1
hags:2:once:/usr/bin/startsrc -g hags > /dev/console 2>&1
haem:2:once:/usr/bin/startsrc -g haem > /dev/console 2>&1
In retrospect I’m wasn’t sure that the HACMP software was installed. We checked and the sys admin confirmed it wasn’t so I suspect it’s failures in previous Oracle installs. There’s a lot of uncertainty about how software has been installed in the past, and I’m being polite 😉
Anyway, another blog to come on this …