Policy for Service Guard fail over for oracle

[Tingli]

We are going to go to cluster for hp-ux 11.23 with 5 databases on them. We need to decide the fail over policy for oracle binary, oracle admin and database files.

Of course the database files need to fail over when needed. But for the sake of upgrade or scaling up and down the database, and any other event, we can have binary or admin in both nodes and no fail over for them.

Are there any pros and cons for different policies, or a best way for doing it?

[Steven E. Protter · Accepted solution]

Shalom,

There are an unlimited number of possibilities even within the active-passive world.

Most common are that the data files are on shared storage and the SG scripts activate the volume group, mount the files systems and then open the oracle databases on failover.

Having the binaries on each node normally is a little better on the performance side. Thats where i keep them, though I sometimes use Ignite replication to make sure the binaries are exactly the same on both nodes before proceeding with SG configuration.

SEP

[Kenan Erdey]

Hi,

in our company we keep binaries in shared disks. So patch, conf. changes, upgrade etc. is no problem. But if you don't keep binaries in shared disks, if there happens a problem in a node you can roll back.

Kenan.

[Kenan Erdey]

Hi again,

i found a document in oracle's website about their recommendations.

http://www.oracle.com/technology/deploy/availability/htdocs/faq.html#A311

Kenan.