Wouldn't you want to load balance it so it's fully redundant rather than having 1 machine doing nothing?

---

indeed, there is a K2 knowledge base article on how to set up Windows Load Balancing for scalability & resilience

we have this setup in our production environment and it works fine. The only thing i can think to note that we have discovered is ensuring that you restart both K2 servers when configuration settings/string table values change as there seems to be some form of caching done by the K2 server

---

I followed the documentation for load balaning K2, worked with K2 support, and K2 consulatant.

Install K2 on two separate windows 2003 servers. Created cluster name using NLB.

The _server table has both severs

All connection are done using the cluster name. I also tested connection to K2 using server 1 and server 2 individually. All works fine.

K2 trace log on server 1 has almost 100% activity and server 2 has almost nothing.

How does the trace log looks like in your environment for each of the server in your farm (NLB)?

To me it does not look like K2 is using the second server in our NLB cluster.

Any ideas, comments is greatly appreciated.

---

lchoing:

I followed the documentation for load balaning K2, worked with K2 support, and K2 consulatant.

Install K2 on two separate windows 2003 servers. Created cluster name using NLB.

The _server table has both severs

All connection are done using the cluster name. I also tested connection to K2 using server 1 and server 2 individually. All works fine.

K2 trace log on server 1 has almost 100% activity and server 2 has almost nothing.

How does the trace log looks like in your environment for each of the server in your farm (NLB)?

To me it does not look like K2 is using the second server in our NLB cluster.

Any ideas, comments is greatly appreciated.

from what i understand, Windows NLB supposedly averages the load over a longer period of time than you might expect. It isn't just a case of 'round robin' requests to each server, i.e. take it in turns, it justs has it's own algorithm to spread the load.

Providing you have tested that each individually works & you can also test the NLB side of things by draining each of the servers off one at a time to check that the other starts recieving requests - your work is done

---