observed LB behavior on 10.2.0

 

We were hoping to reintroduce some active load balancing into our pools, and deployed config using Observed, however found the exact behaviour described here:

 

http://support.f5.com/kb/en-us/solu...11332.html

 

Despite this only apparently affecting 10.1.

 

 

We were seeing this on large scale volume tests in our clone environment, and I recreated a labs scenario with 3 pool members, one responding after 0.5s, another after 1.5s and a third after 2.5s. running 500 “random” requests into this over about 10 minutes, we saw over 80% or requests hit the 2.5 service, with only 2% hitting the 0.5s node. These endpoints were actually a single wildcard vip within a iRule on it, so clearly totally identical outside of the login in the iRule.

 

 

We’re also intending to be on 10.2.4 HF6 by August, if this will have any impact on what we’re seeing.

 

 

Any comments on this would be appreciated, as there appears to be NO consistent description on what Observed really is when you try and understand it.

 

 

Thanks

 

 

Chris