MAC Flapping error on Cisco Switch 3650, Host HP UX Server

[Jimesh_Makawana]

Hi admins,

Recently we have deploy hp ux on intigrity servers.

but i receive continue mac flapping error on my cisco l3 switch (model  3650)

ERROR IN CISCO L3 SWITCH

%SW_MATM-4-MACFLAP_NOTIF: Host 6431.5000.8f30 in vlan 20 is flapping between port Gi0/15 and port Gi0/8

%SW_MATM-4-MACFLAP_NOTIF: Host 6431.5000.4dd0 in vlan 20 is flapping between port Gi0/15 and port Gi0/8

%SW_MATM-4-MACFLAP_NOTIF: Host 6431.5000.7f64 in vlan 20 is flapping between port Gi0/8 and port Gi0/15

HP Switch 1

Port 1 & 9 continues up & down

HP Switch 2

Port 1 & 9 continues up & down

NODE 1

# lanscan

Hardware Station        Crd Hdw   Net-Interface  NM  MAC       HP-DLPI DLPI

Path     Address        In# State NamePPA        ID  Type      Support Mjr#

LinkAgg0 0x643150004DD0 900 UP    lan900 snap900 6   ETHER     Yes     119

LinkAgg1 0x643150004DD8 901 UP    lan901 snap901 7   ETHER     Yes     119

LinkAgg2 0x000000000000 902 DOWN  lan902 snap902 8   ETHER     Yes     119

LinkAgg3 0x000000000000 903 DOWN  lan903 snap903 9   ETHER     Yes     119

LinkAgg4 0x000000000000 904 DOWN  lan904 snap904 10  ETHER     Yes     119

NODE 2

# lanscan

Hardware Station        Crd Hdw   Net-Interface  NM  MAC       HP-DLPI DLPI

Path     Address        In# State NamePPA        ID  Type      Support Mjr#

LinkAgg0 0x643150008F30 900 UP    lan900 snap900 6   ETHER     Yes     119

LinkAgg1 0x643150008F38 901 UP    lan901 snap901 7   ETHER     Yes     119

LinkAgg2 0x000000000000 902 DOWN  lan902 snap902 8   ETHER     Yes     119

LinkAgg3 0x000000000000 903 DOWN  lan903 snap903 9   ETHER     Yes     119

LinkAgg4 0x000000000000 904 DOWN  lan904 snap904 10  ETHER     Yes     119

[Matti_Kurkela]

You have some link aggregates configured: if they are using a load-balancing mode, the messages about MAC address flapping would be a necessary consequence of that in your network topology.

For example:

Node 2 sends packets out of lan900 with load balancing, so about half of the outgoing packets go to switch1 and the other half to switch2. The HP switches don't see anything wrong with that, but the Cisco L3 backbone switch sees the node2 lan900 MAC address jumping back and forth between the two switches and does not like it too much.

Too bad you did not identify the models of the HP switches nor your HP-UX version.

I'm not really a switch configuration guru, but... Maybe it's possible to configure them as interconnected (stacked) switches, so they could accept IEEE 802.3ad link aggregates with one leg in each switch? Then you could configure each host link aggregate and the uplinks of the HP switches to use 802.3ad: that would allow the Cisco switch to understand the situation properly.

If that is not possible, you may have to configure the link aggregates to failover mode (active/passive) instead of load balancing (active/active) on the HP-UX servers. That should cause the MAC address to jump only when there is a link failure, and that should keep the Cisco switch happy.

The port up & down behavior seen by both HP switches on ports 1 & 9 suggests a different configuration problem on node 1 with both link aggregates (lan900 and lan901). Are they configured exactly the same as on node 2? Are the switch ports configured the same way on the HP switches (port1 vs. port2 and port9 vs. port10)?

If the servers are running HP-UX 11.31, run "nwmgr -S apa" to see the link aggregate configuration, or "nwmgr -S apa -v | more" to see full details (the output of the second command may be very long; if you post it, please copy & paste only the parts that refer to active link aggregates and their member NICs).