Fail over

[Rushank]

ENV: Service guard two node cluster, L-class, Oracle, HPUX11.
one package run on each box. (oracle application package and Oracle DB package)

We found that there are lots of traffice in between these two servers,(about 60 % of the network traffic) The package ip on main subnet. Can I connect back-to-back connection between these two servers to avoid network traffice on main subnet. If yes How it will effect during fialover??

[A. Clay Stephenson]

I suspect that your better option is to configute APA (Auto-Port Aggregation) which will increase bandwidth at all times and provide failover. In a MC/SG environment, you would really need two switches and APA conmnections to each switch.

By the way, you just might find that your optimum setup is to run both packages on the same server so that there is no DB network traffic between the DB and App. I know that this would make one server sit there and do nothing but I would ceratinly test this because I suspect this configuration will outperform your current configuration.

[Rushank]

Thanks Clay,

Yes we did plan to run both package on one box but than other box has to wait until failover occurs to perform.

With current configuration the traffic on main subnet is too much. I don't think my Company will purchase APA now.
My pln here is to have a back-to-back connection and see if it does make any difference but I don't know the consequences of failover.

[Rushank]

Well here is the test result.

I configured lanport and connected it as back-to-back. Now both servers can ping, telnet to each other with this new private ip (100.100.100.1 and 100.100.100.2) Then I modified host entry and give package-ip name to both and removed original package-ip entry from hosts file. Now packages are running on private ip. I infact moved a package to next node still cluster and packages are up. Is it OK If I live with this setup??

[Rushank]

Back on top..

If I connect these two nodes via a switch making a separate VLAN for package IP how service guard would act during failover??