Disparity in L4 Connections and L7 Traffic with Least Connection Load Balancing

Hello everyone,

Greetings!


We've configured load-balancing the request based on least connection(member), and the two nodes (NODE-***-C and NODE-***-D) hosts the application running on two service port 8080 and 8081. The virtual server is configured with cookie-based session persistence.

The backend hosted JBoss service seemed to down due to the huge volume request on NODE-****-D:8180 service port, but even though there is a huge L4 connection - 1.1K, a huge disproportion among all the other active pool members, the received requests, bits and packets are comparatively very low.

 

 

Based on the analysis, there are few health monitor status going down and up among the pool members, during those few nodes going down, all the new inbound request might have landed on any other available, and due to cookie persistence, the landed user may have persisted the connection of L4 till the cookie timeout didn't occurred (180 seconds).

But still how the NODE-BPM-D:8180 ended up with more than 1K+ request and with such a huge disparity among all the pool members.

 

Does anyone have insights into what could be causing this imbalance and why NODE-***-D:8180 is handling significantly more L4 connections despite lower L7 traffic? Any suggestions on how to address or further investigate this issue would be greatly appreciated.


Also regarding least connection(member) only considers the least L4 connection among the pool members in a pool or any additional factors also comes into play regarding the least connection(member) method?