Help Needed for RAC with TAF

[mail.madhankumar]

Hi Tim,
when we use taf service to connect the rac database . it is connecting to preferred node based on TAF.. so server load will be in main node only , then why are they saying Load balancing feature..

ex:
we have 2 node rac. 1st node is rac1 and second node is rac2 . here rac1 is preferred node and rac2 is availability on taf.
i connect to rac2 node and connect to schema using taf service . but that connects to rac1 instance.. where will be the server load?. rac1 server or rac2 server?. or will it connect to rac2 instance after some connections established.? using load balancing..? please explain me...

Regards,
Madhan

[Tim...]

Hi.

If doesn't make sense to define a preferred node, then try to use load balancing. They are trying to achieve totally different things. You need to either:

- Have a service with a preferred node and not attempt to load balance. This way the connections will be made to the preferred server, but will go to the other server if the preferred server is down.

or

- Not have a preferred server and have a load balanced connection. This way, load will be spread evenly across both nodes.

The two goals are mutually exclusive.

Cheers

Tim...

[mail.madhankumar]

Hi,
if we have preferred server...and also mentioned load balance true what will happen..

your point
- Not have a preferred server and have a load balanced connection. This way, load will be spread evenly across both nodes.

load shared on both node using load balancing.. one node got failure means.. wat will happen.. will they loss the connection or failover to another node..

if both feature works together means it's gud.. rite..

Regards,
Madhan

[Tim...]

Hi.

I'm not really sure what would happen if you combine a preferred node and load balancing. I've never done it because it's stupid. So I've never tested the consequence of that configuration. If you want to know, then test it.

If you have load balancing and no preferred server your load is spread between nodes. If a node goes down, then only the working node is used. It would be pretty stupid of Oracle to design a HA solution that wasn't HA right?

These are things that are easy for you to test.

Cheers

Tim...