Arp question

Unanswered Question
Aug 6th, 2008

We have two servers connected to the same 4500 Multilayer switch, the two servers are working in active/stanby mode using virtual IP address between the two servers. On the 4500 switch, the MAC address of the servers virtual IP is the MAC address of the active servers. When the active server fails, the arp table on the 4500 does not update the VIP address with the backup server, so no one is able to access the server, till we clear arp. I thought of changing the ARP time out from 4 hours (default) to something like 10 sec, but my concern is that could lead to many broadcast requests. Any idea?


I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
francisco_1 Wed, 08/06/2008 - 02:06

i think it down to the nodes to advertise their active NIC mac address to the switch. say you are using windows clustering for example, the servers only uses the same virtual IP not MAC address. so when the standby node takes over, the node should announce it mac address to the switch as the active node instead of you flusing the arp manaully.

Wantser1981_2 Wed, 08/06/2008 - 02:12

I would imagine is more to do with the fact that the backup server isnt advertising that it now has the address to the switch until it sees a request initiated to it, which it wont get whilst the old ARP entry still exists. Flushing the ARP will clear the location of the server and thus flood the ports with a request that the backup will see and reply to hence adding the new location in the CAM table.

The CAM default age time should be 300 seconds, not 4 hours. As far as I am aware, so this might have been changed.

francisco_1 Wed, 08/06/2008 - 05:37

The default ARP table aging time is 4 hours while the CAM holds the entries for only 5 minutes


This Discussion