We've seen this in the past when 'per packet' or 'round robin' load balancing (sometimes called 'operating mode') was turned on on the host. The host would send alternate packets down each NIC in their team thus causing the flapping. This round robin mode is the default for at least some versions of some operating systems.
I think generally the risk of out of order packets on per packet load balancing for most application would lead one down the path of setting up 'per session' load balancing... or in your case since since the hosts are connected to different switch stacks then perhaps also consider changing those hosts to an 'active/failover' mode where the hosts use just one NIC unless they detect a failure. cr -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Bill Blackford Sent: Sunday, June 27, 2010 7:55 AM To: [email protected] Subject: [c-nsp] MACFLAP Message Had an issue the other day that may or may not be related to the following message. Jun 24 11:30:37.354: %SW_MATM-4-MACFLAP_NOTIF: Host 0015.1761.8140 in vlan 311 is flapping between port Po4 and port Po5 Jun 24 11:30:51.665: %SW_MATM-4-MACFLAP_NOTIF: Host 0015.176f.a6e4 in vlan 311 is flapping between port Po4 and port Po5 Jun 24 11:30:52.672: %SW_MATM-4-MACFLAP_NOTIF: Host 0015.1761.8140 in vlan 311 is flapping between port Po5 and port Po4 Jun 24 11:32:18.924: %SW_MATM-4-MACFLAP_NOTIF: Host 0015.176f.a742 in vlan 311 is flapping between port Po5 and port Po4 Jun 24 11:32:24.460: %SW_MATM-4-MACFLAP_NOTIF: Host 0015.176f.adae in vlan 311 is flapping between port Po5 and port Po4 Jun 24 11:34:12.237: %SW_MATM-4-MACFLAP_NOTIF: Host 0015.176f.a6e4 in vlan 311 is flapping between port Po4 and port Po5 Jun 24 11:34:12.900: %SW_MATM-4-MACFLAP_NOTIF: Host 0015.176f.adae in vlan 311 is flapping between port Po5 and port Po4 Jun 24 11:34:13.278: %SW_MATM-4-MACFLAP_NOTIF: Host 0015.176f.a742 in vlan 311 is flapping between port Po5 and port Po4 VLAN311 is an Oracle heartbeat L2 vlan that spans across the data center. This message was logged on the 3750 VC stack (core). Each node is connected to access switches that each connect into the core via LACP bundles. Each of these MAC's are part of a Linux BOND group on various hosts. IOW, each bond interface member connects to each of the (in this case) two access switches. The topology is loop free from the perspective of the network switches as the LACP bundles eliminate the need for spanning tree. Now, this may be more of a question for how Linux bonding works across multiple access switches but I need to start here. I'm not finding a lot of information about this message. Does anyone on the group have any insight? Thank you in advance, -b -- Bill Blackford Senior Network Engineer Technology Systems Group Northwest Regional ESD Logged into reality and abusing my sudo priviledges _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/ _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
