Quote:
Originally Posted by Ammad
Dear all,
I want to install oracle on Linux OS cluster instaed of Oracle RAC. I need help on this i haven't configured cluster on linux and want to know setps i am using OEL 5.3. i am trying to install this on HP servers and HP Storage. i am using HP-ILO for fencing.
do i first configure Cluster and then install all binaries on node 1 and then db on Storage. what is your recommendation and process to perform this.
Thanks.
|
This is old but, here goes:
For Active-Passive:
It doesn't matter when you do the cluster install, Oracle won't know about it. You will install the Oracle binaries on each server as stand-alone (non RAC) instances. The database storage will need to be shared and mapped the same on both servers. Once you have the database running on node 1, shut it down and transfer any needed files to node 2 (namely the pfile if it's not on shared storage). You'll want to make sure you can get the instance up and running on node 2 without any issues. This shouldn't be difficult. The setup will be the same as it is on node 1. You CANNOT start both instances at the same time with this setup.
As far as the OS cluster part is concerned, I don't know how to set that up. There will probably be a private interconnect and what not going on there, but that's a 3rd party thing from my point of view (I'm an Oracle DBA, not a system admin). The OS Cluster software will detect when node 1 goes down and automatically startup the Oracle instance on node 2 and handle all failover of virtual IPs and what not for you. From a DBA's perspective, it will just look like 2 single-instance databases with the same name on different servers. Only one instance can run at a time.
If you're using ASM, you'll have to do some extra installation steps and what not to get ASM going (in 11gR2, it's part of the grid infrastructure). You won't be able to mount the diskgroup used by the database in both locations (I've tested and got an error, no harm done). You can run an ASM instance on both nodes, but before mounting the diskgroup on node 2, it much be dismounted on node 1. Basically it's the same, 2 single-instance databsases that use the same data files with the starting and stopping during a failure being controlled by the OS cluster software rather than Oracle clusterware.
This isn't true high availablility though. I've had to work with a setup like this where the cluster software vendor was Veritas.
Active-Active: Just use RAC. I've never seen it attempted any other way yet. You can setup a 2-node RAC with the standard edition license so there is no additional cost unless you are trying to avoid paying for an Oracle DBA that knows how to set it up and administer it.