Cisco Switch MAC flapping to VC Flex-10 1Gbit SFP uplinks

[chuckk281]

Tom was having an issue with a Cisco switch and Virtual Connect:

**********************

Customer has a C7000 with 2 VC Flex-10 with an active / active 1Gbit connection to a VC network. One side of this config is getting MAC flapping warnings on the stacked Cisco 3750 switch they are connected to. Trying to understand what is causing this flapping on this one VC module in Bay 1.

OA is V 3.32, VC is at 3.18. Picture of their VC network uplinks attached. They have 2 uplinks but have not enabled LACP on Cisco switch (trying to get them to do this) but I do not see how that would cause flapping as one uplink shows and should stay in standby.

Here is what their Cisco switch logs show as the flapping:

> Mar  9 15:46:39 EST: %SW_MATM-4-MACFLAP_NOTIF: Host 101f.7478.2bc9 in

> vlan

> 75 is flapping between port Gi3/0/41 and port Gi2/0/42 Mar  9 15:47:09

> EST: %SW_MATM-4-MACFLAP_NOTIF: Host 101f.7478.2bc9 in vlan

> 75 is flapping between port Gi3/0/41 and port Gi2/0/42 Mar  9 15:47:39

> EST: %SW_MATM-4-MACFLAP_NOTIF: Host 101f.7478.2bc9 in vlan

> 75 is flapping between port Gi3/0/41 and port Gi2/0/42

Blades are running ESX 4i. They also have a SUS with 10Gbit uplinks to Nexus switch with no flapping issues – only this bay 1 module exhibiting this behavior.

Any ideas on what this could be – bad SFP or cable perhaps? This same customer had an issue with same enclosure where the 1Gbit SFPs and support swapped their VC modules for reworked ones per EA c03208179.

************************

The discussion:

From Mark:

Cisco is updating the MAC/CAM table when LLDP is disabled on the Cisco switch, certain models behave like this and “is expected behavior”. The LLDP frames originating from the VC module have a common Layer 2 address (MAC address), not a MAC address per port. There is no hardware problem here, nothing should be replaced. As a workaround, LLDP can be enabled on the Cisco switch and it will process those frames on the Management CPU instead and not update CAM table.

From Edwin:

We as ISS GCC tower has opened a case towards L3 EPM to get this behavior documented in an advisory. Within HP we all know this is an Cisco related issue but we don’t have official document that will describe this behavior. Due to that it is sometimes hard to convince customers that this is a Cisco bug since they claim they see the issue after doing a VC firmware update.

As a workaround the MAC flapping can be stopped by adding a filter to the Cisco ports connected to the Virtual Connect Uplink ports. For example:

Cat3750Switch(config)# mac access-list ext filtermac
Cat3750Switch(config-ext-macl)# deny host <VC internal MAC> any
Cat3750Switch(config-ext-macl)# permit any any
Cat3750Switch(config-ext-macl)# exit
Cat3750Switch(config)# int g1/0/40
Cat3750Switch(config-if)# mac access-group filtermac in

******************

Comments or suggestions?

[chuckk281]

Herman also had some input:

****************

Remember… It is possible on Cisco switches to have LLDP transmit enabled (VC sees LLDP info from the Cisco switch) but to have LLDP receive disabled (LLDP not sent to Cisco mgmt CPU and instead is “learned”) as these are two separate configuration items in Cisco IOS.  Might be worth checking to ensure “lldp receive” is enabled on the interfaces connecting to VC.

****************