WLBS (W2K Load Balancing) and 'single' IP Affinity

[John]

When 'single' affinity is used in Windows Load Balancing, what is the best
way to stop clustering without affecting users currently connected.

I previously used 'wlbs stop' when 'none' affinity was used. However, when
'wlbs stop' is used with 'single' affinity, it causes the next http
connection to time out. Following the timed out connection, future http
connections are then redirected to other server. I need to avoid any time
outs for users.

I have also attempted to use 'wlbs drainstop' but I can't seem to tell how
long it will take before load balancing is stopped on the server being taken
offline.

Thanks,

John

[Steve Riley]

You're encountering the differences in how the NLB algorithm works. With no
affinity, each incoming request is *newly* analyzed by the algorithm and
sent to any of the available servers. So dropping a server from the cluster
doesn't have any effect on the clients. With single affinity, once a client
makes its *initial* connection to a cluster member, the algorithm drops out
of the picture and just directs all future client requests to the same
member. When that member disappears, the algorithm won't automatically remap
clients to different memebers until the client tries, and fails, to connect
again. This triggers the algorithm to drop that client from its table;
future connections from the client cause the algorithm to pick a new member
from those remaining in the cluster.

--
Steve Riley
Microsoft Telecommunications Consulting in Denver, CO
ster...@microsoft.com

"John" <jo...@photoavenue.com> wrote in message
news:u7jewimrAHA.1444@tkmsftngp03...