MTU not correctly applied for Service Engine data interfaces in GCP cloud environment

Products

• VMware Avi Load Balancer

Issue/Introduction

When se_mtu is configured in SE-Group properties, the MTU is not correctly applied on SE data interfaces in GCP cloud deployments if RSS or multiple dispatchers is enabled. 

The symptoms for this issue could be

1.  Application traffic delay due to packet fragmentation.
2. Connection timeouts on latency sensitive application. 

 

To verify if you're running into the issue:

• Step 1: Confirm if RSS or multiple dispatchers are enabled in the SE-Group by referring to the following documentation
  https://techdocs.broadcom.com/us/en/vmware-security-load-balancing/avi-load-balancer/avi-load-balancer/22-1/vmware-avi-load-balancer-configuration-guide/se-advanced-networking/configuring-tso-gro-rss-.html
• Step 2: Confirm if se_mtu is set in the SE-Group config
  

  [admin:admin]: > show serviceenginegroup <SE-Group-name> | grep se_mtu

• Step 3: Check if the MTU is not correctly applied on the SE interfaces
  

  [admin:admin]: > show serviceengine <SE-name> interface | grep mtu

  
  
  

We will only hit this issue if we see a mismatch in the values in Step 2 and Step 3. 

Environment

GCP cloud 

Cause

se_mtu configuration may not take effect if either RSS (max_queues_per_vnic) or multiple dispatcher configurations are enabled in the SE-Group

Resolution

This issue is fixed in 30.2.3-2p2, 30.2.4, 31.1.2, 31.2.1