We have upgraded some centrally managed spark clusters from R81.10.00 to R81.10.07 to activate SDwan and run into some serious stability issues. The problem seems to only be related to Vmac in ClusterXL.
It looks like internet lines are flapping, but by doing some troubleshooting, we see the reply traffic arriving in the standby firewall.For some reason, the standby firewall is sending gratuitous ARP messages for the VIP IP addresses, sent out with the VMAC as source. This broadcast message will trigger a change in the mac table of the switch, and traffic for the VIP is put on the wire of the standby firewall.
TAC case is created and confirmed to be an issue, other customers have also seen this behavior. Disabling vmac solves the issue, shutting down the standby firewall also !
Be aware when upgrading to this version when you have VMAC enabled ! R81.10.00 does not have the issue.