Redeploy / upgrade / resize operations of NSX-V Edge appliance (ESG or DLR) configured for High Availability to NSX-V 6.4.9 may lead to data path impact

Products

• VMware NSX

Issue/Introduction

Symptoms:

• Upon upgrade of NSX-V 6.4.9 HA enabled edge appliance, if edge-0 is the active edge node in the topology, there may be impact to the data flowing through the edge
• Upon redeploy or appliance size change of NSX-V 6.4.9 HA enabled edge Appliance, if edge-0 is the active edge node in the topology, there may be impact to data flowing through the edge

Cause

If edge node-0 is the Active node in the HA pair before upgrade or redeploy, during upgrade or redeploy new active edge (node-1) takes over and marks new edge node-0 as standby. During this process new edge node-0 (standby) sends out GARPs to all directly connected networks. Since new edge node-0 is standby, the edge is not participating in traffic forwarding, thus resulting in all the traffic directed to node-0 standby  edge and getting dropped.

Resolution

This issue is resolved in VMware NSX Data Center for vSphere 6.4.10, available at VMware Downloads .

Workaround:

Recovery after edge is impacted:

Option A:

1. On the impacted edge HA pair, set admin state to down on the active edge node from NSX UI.
2. Refresh the UI.
3. Once the standby node changes HA state to active, set admin state back to UP from NSX UI.
4. Refresh the UI and ensure both the edge nodes show admin state UP in UI and HA state has converged(Active/Standby)

Option B:

1. Reboot the active edge appliance via vCenter server.

In case the edge upgrade or the redeploy operation has not been performed use one of the options below as a potential workaround:
 

Option 1: Disable HA prior to the upgrade or redeploy and keep edge running without HA (stand alone)

Option 2: Prior to upgrade or redeploy of HA edge, please make sure edge 'node-0' is on standby and node-1 is active state, otherwise follow the Option A as mentioned above.

 

Additional Information

Impact/Risks:
Data path on the edge, under redeploy or upgrade is impacted.