Pool round-robin load balancing not balanced

Would anything in here apply? http://support.f5.com/kb/en-us/solutions/public/10000/400/sol10430.html

I've checked that documentation. The only iRule being applied is selectiveSNAT, which maps SNAT IPs. Straight Round robin for the pool.

The only thing that I see that could come into play is the selection of Type of Performance under the Virtual Server Properties. But the documentation doesn't indicate this will result in any preference being given to members based on response time.

OK. I assume you are getting enough connections to make the statistics valid. What statistics are you actually looking at? Are these borne out by similar stats on the servers?

Could you post the VIP and Pool configuration from the bigip.conf file?

Ok, see below, pretty straightforward. I also tried changing the profile in the Virtual definition to Standard, made no difference..

pool pool_live.sample.com_40001 {
   monitor all slow_icmp
   members {
      10.48.2.37:40001 {}
      10.48.2.38:40001 {}
      10.48.2.39:40001 {}
      10.48.2.42:40001 {}
   }
}
pool pool_live.sample.com_http {
   monitor all slow_icmp
   members {
      10.48.2.37:http {}
      10.48.2.38:http {}
      10.48.2.39:http {}
      10.48.2.42:http {}
   }
}

virtual live.sample.com_http {
   pool pool_live.sample.com_http
   destination 10.48.254.79:http
   ip protocol tcp
   rules SelectiveSNAT
   profiles fastL4 {}
}
virtual live.sample.com_https {
   pool pool_live.sample.com_40001
   destination 10.48.254.79:https
   ip protocol tcp
   rules SelectiveSNAT
   profiles fastL4 {}
}

What platform is this running on? Also, when you say 6 to 8 times more, is it 6 to 1 or 600 to 100 or 6000 to 1000? I ask because if the stats are low and you are on a platform with 8 cores or so (like an 8900) then CMP could easily explain the difference. But if the statistics count is high, then I am not really sure.

Depending on your LTM CMP may be responsible. see SOL7751 Scenario 2: Virtual server with CMP enabled on a BIG-IP 8800

Four connections are made to the virtual server, unlike the first scenario where CMP was disabled, the BIG-IP distributes the connections across the multiple TMM processes. The BIG-IP 8800 with CMP enabled can use four TMM processes. Since each TMM handles load balancing independently of the other TMM processes, it is possible that all four connections are directed to the same pool member.

--Connection 1--> | | --Connection 1--> TMM0 --> 10.0.0.1:80 --Connection 2--> |-> BIG-IP Virtual Server ->| --Connection 2--> TMM1 --> 10.0.0.1:80 --Connection 3--> | | --Connection 3--> TMM2 --> 10.0.0.1:80 --Connection 4--> | | --Connection 4--> TMM3 --> 10.0.0.1:80

The CMP feature is designed to speed up connection handling by distributing connections across multiple TMM processes. While initially this behavior may appear to favor one or several servers, over time the load will be distributed equally across all servers.

Well as it says in the SOL7751 http://support.f5.com/kb/en-us/solutions/public/7000/700/sol7751.html Your running version 10 so you are using CMP Clustered Multi-Processing

Clustered Multi-Processing (CMP) is a feature that was added in BIG-IP 9.4.0. CMP allows specific platforms with multiple processing cores to use multiple Traffic Management Microkernel (TMM) instances to increase traffic management capacity.

Clustered Multip-Processing (CMP) can be a significant factor in how the BIG-IP system distributes connections across active TMM instances, which in turn can affect distribution to pool members.

Not sure of the exact model number, but I have a system with 2 CPUs, running 10.2.1...

'tmsh show sys hardware' should tell you all you/we need to know.