It is possible to encounter multiple problems when adjusting a fabric to support VMotion, here we will analyze one possible scenario encountered

Main symptoms:

• VM machines we are losing connectivity during vmotion. The only action which solves the problem is to run a "clear etherswitching-table " on the switch/VTEP where the machine was present before the move. Otherwise it might take minutes or hours for the machine to become reachable again
• L2/T2 entries duplicated between leafs with connection to clusters

Possible causes include:

The default settings of MAC MOBILITY parameters are quite aggressive. However, in some cases we need to have optimum settings to allow this MAC move in valid scenarios like VRRP convergence, STP, VMotion, Server Migration etc.

During VMotion, we can see up to 6 times the MAC being moved from one hypervisor to another. The following setting should help in overcoming that:

set protocols evpn duplicate-mac-detection detection-threshold 20
set protocols evpn duplicate-mac-detection detection-window 5
set protocols evpn duplicate-mac-detection auto-recovery-time 5

Another useful article:

How to identify and correct duplicate MAC address

In addition to this, it is worth mentioning that it is convenient to check the following settings

• route distinguisher
• vrf/route target
• IPs of the router IDs

Having a duplication of these could generate these same symptoms since the overlay will have problems to differentiate correctly the routing.

In case any of the possible causes are observed, please open a JTAC case to check the device and version functionality

Possible logs to enable:

traceoptions for EVPN and L2ALD.

set protocols evpn traceoptions file evpn.log
set protocols evpn traceoptions flag all
set protocols l2-learning traceoptions file l2ald.log
set protocols l2-learning traceoptions file size 50m
set protocols l2-learning traceoptions file files 10
set protocols l2-learning traceoptions level all
set protocols l2-learning traceoptions flag all