9.2.0.4 upgrade - new Oracle Home?

[gopi]

I will be upgrading my production database from 9.2.0.1 to 9.2.0.4 Saturday night. I will have plenty of B/Us.

Do you recommend creating a new Oracle home for the 9.2.0.4 binaries or just upgrade the current Oracle home ? I can always reinstall 9.2.0.1 and go back from a cold backup.

Here is my experience - I did not create a new Oracle home for any of these:
1. Development upgraded very easily
2. QA was a nightmare due to a corrupt JVM..
3. Eventually I cloned production into QA and then the upgrade went very smoothly (this gives me big confidence that the prod upgrade will work).

Thanks

[stmontgo]

I've never created a new home for a patchset which is really not an upgrade, nontheless I always backups the binaries and registry

[marist89]

I ALWAYS install into a new $OH and apply the patches to the new $OH.

[stmontgo]

Originally posted by marist89
I ALWAYS install into a new $OH and apply the patches to the new $OH.

why so? do you mean you could end up with an oracle home for 92020, 92030, 92040,92050 etc

[marist89]

Originally posted by stmontgo
why so? do you mean you could end up with an oracle home for 92020, 92030, 92040,92050 etc

In theory, sure. However, once I've patched 9.2.0.4 to 9.2.0.5, I un-install 9.2.0.4. While I'm upgrading, I have two $OH, but while everything is on the same version, I have one $OH.

I run a lot of machines with multiple databases. (I know, I know, and TKyte agrees with you. It works for my environment.) If these databases are down for more than about 2 hours on the weekend, all heck breaks loose. I use multiple $OH for a couple reasons:
1. I can install and patch the software during the week. This will cut about 30 minutes off my upgrade time. Also, since I have to install with a GUI, I can't do it remotely over dial-up.
2. When I'm upgrading, I don't have to upgrade every database on the host at the same time. While this may spread my upgrade over several weekends, it still keeps the database down for less than 2 hours.
3. My rollback plan is simple, shutdown the existing database, switch the $OH, startup, rerun the catalog scripts. If something goes wrong in the upgrade, I can still have the DB up in 2 hours.
4. Since all the GUI work is done, I can run the upgrade over dial-up. While this has nothing to do with my 2-hour window, it saves me the 2.5 hours of commuting back and forth to the office for an hour's worth of work.

[gopi]

Originally posted by marist89

My rollback plan is simple, shutdown the existing database, switch the $OH, startup, rerun the catalog scripts. If something goes wrong in the upgrade, I can still have the DB up in 2 hours.

Interesting...rather than restore from a cold backup