Has anyone else been as screwed by ARP suppression and windows clusters as hard as I am right now?

Windows cluster won’t failover because arp suppression replies to arps causing windows to detect duplicate IP and not failover. Even if you just shutdown one of the machines it won’t failover until arp times out in the switch which stops the vxlan network from being able to do arp suppression. This clears windows thinking there is a duplicate IP and it will finally failover.

This can also be reproduced by simply trying to move an IP from behind one vtep to a secondary IP on a windows box behind some other vtep. Windows will make it as duplicates be NEVER install it even after arp clears out.

PS Arista doesn’t have a way to turn it off yet.

Edit: Bit more details on topology.

Arista EVPN fabric MLAG shared VTEPs Symmetric IRB (with the routing VNI)

To temporarily resolve until Arista has disable arp suppression. I’ve removed these gateways back out of EVPN IRB and bridge the traffic over the fabric to a pair of switches that have the GW for these few VLANs.