NSX Load Balancer forwards requests to a single pool member despite the algorithm is LEAST_CONNECTION and other members are UP

Products

• VMware NSX

Issue/Introduction

• You configure NSX Load Balancer with LEAST_CONNECTION algorithm.
• You see NSX LB forward requests to a single server, even though other pool members are up and the requests do not have session persistence.

Environment

VMware NSX

Cause

LB has an internal value of connection numbers for each pool member and accordingly decides which pool member to forward a new request to.

Sometimes the value of connection numbers might go unexpected.

Resolution

For L4 LB, the issue is fixed in NSX-T 3.0.2 and later.

For L7 LB, the issue is fixed in NSX 4.2.2 and later. NSX 9.0.0 does not have the fix yet and future 9.0.x release is planned for the fix.

To workaround the issue, follow the steps.
Note: Session persistence will be cleared after the workaround.

1. Navigate to [Networking] - [Load Balancing] - [Load Balancer]
2. Edit the load balancer.
3. Disconnect the load balancer from the Tier-1 gateway and click [SAVE].
4. Edit the load balancer again and connect it back to the Tier-1 gateway.

It is another workaround to change the algorithm to ROUND_ROBIN or IP-HASH.