Upgrade to 22.1.3 and above enables the LRO knob resulting in VS traffic disruption

Products

• VMware Avi Load Balancer

Issue/Introduction

• Auto enable of LRO post upgrade may cause Performance/packet loss issues.
• This issue may occur at L2 segment or in environments with IP routing enabled.
• Bigger packets will be seen on SE capture if LRO knob is enabled on Service engine group.This will eventually cause ICMP Fragmentation Needed error. (as DF bit is set)

Cause

• Starting with NSX Advanced Load Balancer 22.1.3 version, LRO is enabled on serviceenginegroup by default for supported environments, such as, Vmware/ NSX-T Cloud.
• LRO is a NIC offload technique which coalesces incoming packets to deliver larger packets to the stack. LRO when enabled on SE interface usually coalesces incoming packets to send a large packet, if the outgoing stack doesn't support this large packet due to limitation on MTU, an ICMP fragmentation needed error occurs.

Resolution

• The following are the CLI commands to determine if LRO is enabled at the Service Engine and the Service Engine Group level:

[admin:<Ctrl IP>]: > show serviceengine <se-name> interface | grep lro
| lro_on                                  | True

[admin:<Ctrl IP>]: > show serviceenginegroup <seg-name> | grep lro
| se_lro                                  | True

• Disable LRO at the SE group level. The following is the CLI to disable LRO on the SE Group level: 

[admin:<Ctrl IP>]: > configure serviceenginegroup Default-Group
[admin:<Ctrl IP>]: serviceenginegroup> no se_lro
[admin:<Ctrl IP>]: serviceenginegroup> save

• The service engines have to be rebooted for this configuration to take effect.