Pavel Bykov Mon, 08/20/2007 - 13:36
User Badges:
  • Silver, 250 points or more

There can be two possible answers.

1. If you are talking about NLB like HP Teaming for network card on a server, then it depends on what you have setup on the server. If it's load balancing or failover redundancy. You can either connect both interfaces into same VLAN, or you can create etherchannel for those ports.

2. If you are talking about Cluster solutions, that use same MAC on both interfaces, than the only way I know of to cleanly solve the problem is to create a separate VLAN for that cluster. The rule is that no two same MAC addresses can exist on two ports (for unicast) therefore, switch creates a "virtual" MAC for the ports, that does not exist in reality. That means that all packets that go to the cluster are broadcast to all hosts in the VLAN. Since VLAN separates broadcast domains, one VLAN per server cluster is the cleanest sollution. Maybe a bit harder from administration point of view.

We use HP Teaming for network card on a server for failover redundancy with two cisco 3020, we connect both interfaces in the same VLAN, but when we use NLB unicast mode it isn't work properly, we haven't answer if we ping to other host or default gateway from inside the server but dedicated IP or virtual IP its respond ok, what can we do?


This Discussion