05-07-2008
4,
0
Join Date: May 2008
Last Activity: 7 May 2008, 7:50 AM EDT
Posts: 4
Thanks Given: 0
Thanked 0 Times in 0 Posts
Lots of stuff to think about!
You need to think about the reasons for running RAC. Is it to overcome downtime due to scheduled server outages? Or for high availability (ie 24x7 or as close as possible)? Are you trying for a 2-site cluster (much harder to administer and set up)?
Anyway you need a way to get to the data which is shared between the cluster nodes, raw disk or a clustered filesystem or NFS- you can use ASM (raw disk shared across cluster ie all backups via Oracle) or GPFS (filesystem shared across cluster, backups via Oracle interacting with backup s/w such as TSM). Both supported with AIX. Also you can look at HACMP shared raw LVs (not v popular these days) or a NAS (though then access to DB is via NFS).
You will need to consider what happens in the different failure scenarios. Especially for 2-site scenario, as you will likely find you will need 3 sites (3rd site is a voting site).
There are documents available with examples for setting up RAC and GPFS from the IBM Oracle Competency Centre. Contact oraclibm fr ibm com for the "Oracle 10g RAC cookbook on IBM pseries running AIX".
Then you can start thinking about tuning as well!
Don't forget about having enough capacity available in case one server is unavailable for a long time.
For availability you will need to consider lots of other stuff to reduce outages (e.g. AIX multibos)
Luck!
Last edited by indulis; 05-07-2008 at 08:44 AM..
Reason: correction