Question regarding load balance method

You really have two options here depending on how precise you need to be:

1. Ratio-based load balancing - setting ratios on pool members sounds exactly like what it is. You define a set of ratio values to each pool member (ex. 2:1, 10:2, etc.) and the pool members are load balanced according to these ratios (ex. 10 connections to member1 for every 2 connection to member2).

    

   https://support.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/ltm-concepts-11-5-1/5.htmlunique_452626201

    

2. Connection limits - if you need more control than that, you can simply specify a connection limit for each pool member. Once that limit is reached the system does not send any new connections to that member.

    

So to answer your second question, if you cap server A at 10 connections, and server B goes down, server A will still only be able to serve 10 connections. This is usually a really good reason to use ratio balancing first and then apply a reasonable connection limit only if needed.

thanks for the response Kevin,

The URLs to the application look something like…

xyz.com/AAAAAAA/ xyz.com/BBBBBBBB/ xyz.com/CCCCCCCC/

Where the AAAAAAA, BBBBBBBB, CCCCCCCC represent different distinct customers using the application

More what I had in mind was redirecting a small number of specific customers to the new server, SERVER_B using a rule something like… IF URI contains “/AAAAAAA/” then redirect SERVER_B ELSE IF URI contains “/BBBBBBBB/” then redirect SERVER_B ELSE redirect SERVER_A

I would be concerned that if we have a specific customer’s users (e.g. customer AAAAAA with 20 users) being redirected to different servers, we would not have sufficient measure on performance improvements if one user is still on SERVER_A and another is hitting SERVER_B.

Well, context-based load balancing is pretty straight forward with either an iRule or LTM policy - sending requests to different pools based on the request URI. So there's likely some combination of that and pool ratio/connection limiting that you'd want to employ. Are you wanting to sending customer A users to server B if server A reaches some threshold?