Symptom
Multiple Symptoms may be observed.
- ARP Suppression cache on a Nexus 9k is stuck with Stale entries.
e.g., After an endpoint moves from VTEP A to B, a remote VTEP C may still point to VTEP A for the end point in the ARP suppression-cache.
- In the case of an IP Mobility, post-move of the endpoint, VTEP C does not update the new MAC Address in the ARP Suppression-cache table (for remote entries) and will point to VTEP A (when it should have updated to VTEP B)
- post-move of endpoint from VTEP A to B, VTEP A may not show any entry for the endpoint that just moved in the ARP Suppression-cache (ideally, the VTEP-A's suppression-cache should point to VTEP B for the endpoint).
Once this defect is hit, East/West Traffic may be impacted within a fabric as the VTEPs start responding with incorrect information when south side hosts initiate ARP Requests
Note that when impacted by this defect, the MAC table, URIB, BRIB on remote VTEPs will show correct entries post-move of an end point; Only ARP Suppression-cache will have the stale entry
Conditions
- Impacts Nexus OS version lower than 10.3(5) in 10.3 train
- Impacts Nexus OS version 10.4(1)
- Versions 9.x and 10.2(x) are NOT Impacted
- Manifested for the VNIDs that are enabled with ARP Suppression if ARP Suppression is disabled on one or few other VNIDs
- A reload may not fix the issue as if the last VNID that becomes active is the one with ARP Suppression disabled, this may still cause problems for the VNIDs that have ARP Suppression enabled (the command- 'show ip arp internal event-history control' will show which VNID came up last that has ARP Suppression disabled/enabled).
Workaround
Enable ARP Suppression for All VNIDs or remove the same from all VNIDs.
Further Problem Description
The output of the command `show ip arp internal event-history control` will show below message if the ARP Suppression was disabled
[arp] E_DEBUG ARP process HMM suppression mode Event=processNotification, suppressKey[0]=, arpLcacheMode=Enabled(L2/L3 ARP Suppression), prevMode=Enabled(L2/L3 ARP Suppression), currentMode=Disabled(ARP Suppression Disabled)
