active DB cluster node

hipikll · Post by **hipikll** » 01 Apr 2008 6:27

Hi.
The situation is:
We have 2 nodes in cluster. On the cluster the ORACLE DB server is running. There are 5 DB instances. Each has it's own floating ip adress. If one instance fails, the cluster do DB switch to other cluster node. It means:
DB1 is fails on node1, then it is switched to node2.
So DB1 is run on node2 but all others DBinstances are still on node1.

Can the agent handle such a DB cluster switching. I know how to bind agent for psychical cluster node on one virtual IP address.
But how to bind agent to run along with the active DBinstance node, not a psychical one.
So - whenever is the active database, the agent will work on this node and no matter of the facts, it has to manage also other node's databases.

That's the question.
(be or not to be)

mauriziog · Post by **mauriziog** » 24 Apr 2008 9:15

For a similar problem with oracle RAC, I have used the loadbalancing method with a script that add or delete resources on the correspondent DB node active.

The script check if the DB switch and if occour deletes resources from that node and add in the other.

The best was connecting the DB switch with resources, when I done this I have used Patrol.

gglau · Post by **gglau** » 29 Apr 2008 5:57

Agent switching has to be coordinated with DB siwtching. Agent is installed on a volume available to the active node only and configured to respond to the DB virtual hostname. When DB switching occurs, shutdown agent and unmount the volume in the outgoing node, then mount the volume and start the agent in the incoming node.

hipikll · Post by **hipikll** » 29 Apr 2008 1:15

Yes, but .. agent can not be switched on other node, because it has to manage oracle jobs on first node. As it was said to me: switch to other oracle database node is not the server node switch.
so agent has to run on first and second server node, but has to determine on which server node one of the DBinstances is running. Others instances continue running on initial node.

mauriziog · Post by **mauriziog** » 06 May 2008 5:35

Use the quantitative resources, nodegroup and a script that checks the available node.
I do so and it works fine.

philmalmaison · Post by **philmalmaison** » 07 May 2008 5:00

install an agent on each physical host, create a node group with that 2 physical host