MAC Flapping Between Port-Channel Interfaces in Voice Network

[ErnestConcepcion]

Our network topology has four switches that are configured to use PAgP trunks between them. 

I have recently replaced a UCS C-Series server as part of a Refresh Upgrade of our CUCM systems.  It came with two quad NICs which have been physically configured to have four cables connected from one NIC into a switch and four cables connected from the other NIC into a different switch.

The four cables from each NIC have been bundled into a Static ‘ON’ EtherChannel connection from the Cisco switches to the server with the default load balancing of src-mac.  On the UCS server side, via vSphere, we have configured NIC Teaming with Route Based IP Hash load balancing.

However, since doing this I keep getting flooded with syslog messages that state that the Mac addresses of both our Call Manager and Unity server VMs are flapping between port channels (i.e., the 'static Ether-Channel, Po30 and the PAgP Ether-Channels, Po1, 2, 3 & 10–see example messages below): 

38229: Dec 4 07:25:50.900: %SW_MATM-4-MACFLAP_NOTIF: Host 000c.2989.d591 in vlan 60 is flapping between port Po2 and port Po1

54819: Dec 4 2017 02:25:49 UTC: %SW_MATM-4-MACFLAP_NOTIF:

Host 000c.2989.d591 in vlan 60 is flapping between port Po30 and port Po10

43534: Dec 2 2017 11:16:45 UTC: %SW_MATM-4-MACFLAP_NOTIF:

Host 000c.29d2.15b3 in vlan 60 is flapping between port Po30 and port Po10

My questions are the following:

1. Is this normal behavior or do I have to change the load balancing on the switch side from src-mac to src-dst-ip to match the UCS server of NIC teaming–Route Based IP Hash?

2. Because NIC teaming–Route Based IP Hash can only be used with static ‘ON’ EtherChannel, will changing the EtherChannel trunks between the switches from PAgP to Static ‘ON’ resolve the MACFLAP error?

Any assistance would be greatly appreciated.

Regards,

Ernest, CCNA

[Georgios Fotiadis]

Hello,

to better understand your topology, can you mark on the network map the port channels numbers?

Georgios.

[ErnestConcepcion]

[Georgios Fotiadis]

If all 8 ports of the UCS server belong to the same port channel, and S1, S2 are not stacked (which are not), then this could be causing the L2 loop that is depicted in your logs.

Can't you create 2 port channels, one per NIC?

Georgios.

[ErnestConcepcion]

I can.  I just have RSPAN setup to monitor voice traffic and configured Po30 as source.  I guess I can setup two separate sources for the RSPAN monitorization.  I will try and post my results.  Thanks.

[Jonathan Schulenberg]

Wait, the two switches are not stacked, or otherwise in some type of virtual chassis configuration?!
If that’s true this won’t work. A Standard vSwitch doesn’t support multiple port channels; it assumes all active adapters are in the same EtherChannel connected to the same logical switch. If the switches are two separate entities then you have two seperate EtherChannels regardless of the port-channel interface number you define.
https://kb.vmware.com/s/article/1004048

Also, yes you should be using src-dst-up.

[nmdservices]

Hello

did you get a fix for this issue/topology?

Regards

nmds