Operating System - HP-UX
1833159 Members
3076 Online
110051 Solutions
New Discussion

Load balancing with MC/service Guard

 
SOLVED
Go to solution
Lacrosse
Regular Advisor

Load balancing with MC/service Guard

In a windows cluster it is possible to load balance your applications accross multiple nodes in the cluster. If this feature availble with MC/Service Guard or would you just have to start up each app on a desired node.
Also during a failover of an app is it possible to make it so there is no interruption to the user
4 REPLIES 4
Kent Ostby
Honored Contributor
Solution

Re: Load balancing with MC/service Guard

Hookem --

1) There is no built in load balancing feature in MC/Serviceguard per se.

2) The only application where I think you can guarentee that customer won't go off line would be with SG/OPS where you are running Oracle Parallel Server on multiple nodes.

With "normal" failover scenarios, SG should boot up its portion of the cluster within 1 to 2 minutes. How long the overall failover takes is dependent on the application you are running.

Best regards,

Kent M. Ostby
"Well, actually, she is a rocket scientist" -- Steve Martin in "Roxanne"
Rita C Workman
Honored Contributor

Re: Load balancing with MC/service Guard

Sizing your applications and your hardware accordingly is how load balancing is done.

Remember, Windows likes to have control and make the decisions....BUT UNIX/HPUX believes that the Administrator sitting behind the keyboard is best person to make the decision. Considering the stability of HPUX over Windows......it would seem that HPUX Administrators do the job much better.

Regards,
Rita
Lacrosse
Regular Advisor

Re: Load balancing with MC/service Guard

So if Iwere running in app in OPS in a MC service guard environment I could have users accross the nodes and if one node went down the users would be transferred over without interruption....is that a correct statement?
Rita C Workman
Honored Contributor

Re: Load balancing with MC/service Guard

No...it would not be totally invisible to the end user.

The application (if properly configured) should automate and failover to the other node. Obviously processes would cease on the failed node and have to be restarted (i.e. your application) on the new node.
Hence, your end users would need to log back in and then continue on the new node.

Based on your application's end user configuration (ex: tnsnames.ora) would determine if the end user had to change anything to continue work on the other node.

Example: If your Oracle pkgA (pkg ip=100.100.100.2) were on nodeA (node ip=100.100.100.1) and your end user had a tnsnames.ora file that pointed to an IP or hostname (i.e. nodeA or 100.100.100.1), THEN each end user would have to reconfigure their tnsnames.ora file on their PC to the new nodeB information. But if their tnsnames.ora simply pointed to pkg name (i.e. pkgA), then it would be a simple matter of them re-logging in. No additional changes required.

So the end user will feel the interuption. And how you set up your SG Design will determine how much extra work you may or may not have to do.

Rgrds,
Rita