Resonate Software Load Balancer and NSX-T

[grimsrue]

Has anyone worked with or had any issues with Resonate Software Load Balancer sitting outside of NSX-T (Underlay) that is trying to load balance to VMs sitting on an NSX-T Overlay?

I have an issue with an app team that is having problems with very slow response times from a Web page sitting on VMs sitting on an NSX-T Overlay. The Clients are connecting to the Web page using a VIP that is running on a Resonate Software Load Balancer that is sitting on the physical network (Underlay).

I have found that Resonate uses an agent that is installed on the VMs that are part of a Resonate LB group. The agent on the web server VMs does a handshake with the Resonate scheduler and Encapsulates the VMs IP addresses before sending the required info back to the Client.

I am wondering if the Resonate LB Encapsulation is having a problem with the NSX-T Overlay Encapsulation. Resonates IP encapsulations is proprietary to Resonate.   

[animesh41]

just to isolate further, if you haven't yet . You can try to setup the same configuration on NSX-T VLAN backed segments with the LB still outside NSX and check if it makes any difference. This will help you to narrow down if its an overlay or non-overlay problem

[grimsrue]

That will be the next thing I am hoping to do. I have to work with the application team to change IP addresses of the VMs, as well as DNS. It is a process.

My next steps are to create a couple of Test VMs that will run in the Overlay and another set on VLAN backed segments and see what happens.

I was just hoping someone out there has run across this Resonate Networking Load Balancer in their network and if so were you experiencing the same issues with slow connectivity or timeouts to the load balanced VMs sitting on the NSX-T overlay.

[grimsrue · Accepted solution]

I managed to figured out the issue, for anyone that is wondering. uRPF - Unicast Reverse Path Forwarding

Resonate Load Balancers that are running a separate server agent forward request directly to the client and bypass the LB. The LB makes the initial handshake between the client and the server, but after that the server responds directly back to the client using the VIP and the server MAC.

All Overlay Segments have a "URPF Mode" setting. By default it is set to "Strict". To fix my issue I had to set this URPF to "None"

VLAN Back Segments and Bridged Segments do not have a "URPF Mode" option.