2950 Not Caching MAC Address of HSRP Gateway

[clay.white]

A 2950 switch (A) is trunked from two copper interfaces (etherchanneled) on another 2950 (B) which is in turn trunked via gig fiber to redundant 6506's. The 6506's share an HSRP address and this address is used as the gateway on all 2950's. Switch A could ping the physical interfaces of both the 6506's, but not the HSRP address. I did a 'show arp' on switch A and the hardware address for the gateway showed as Incomplete. After I manually inserted the IP and MAC address of the HSRP address in the arp table, I could then ping switch A from everywhere on my network. Why did I have to enter the HSRP IP and MAC addresses into the arp table on switch A when I have not had to do that on switch B or any of the other 2950's that trunk directly to the 6506's?

[ankurbhasin]

Hi Clay,

Many a times 2950 does not lean a complete arp entry and show incomplte arp entries. Now when you have already inserted a static arp entry and its working try to remove that static arp entry and then try ping I hope iot will ping.

I have practically seen this problem and this workaround worked for me.

HTH

Ankur

[Kevin Dorrell]

If the ARP entry shows "incomplete", this mans that the switch has put out an ARP broadcast for the HSRP IP address, but has not received a reply. The ARP request will be broadcast on the management VLAN of the 2950. So the question is: is the management VLAN of the switch the same as the VLAN of the HSRP address it is trying to ping, and on the same subnet? Also, check the masks correspond on both sides.

Perhaps it would be useful to see the relevant config lines from the 6506's and the 2950.

Kevin Dorrell

Luxembourg

[clay.white]

I have discovered the problem. Due to the odd circumstances that we have to work around in this business at times, the two trunk ports on switch B connect directly into copper/fiber media adapters. The link on the copper side of the media adapters was up, but one of the fiber links to switch A was disconnected. Switch B still was maintaining it's channel group because of the "link up" condition on both its trunk ports. Switch A had only one of the interfaces up in its channel group. I am not certain as to why switch A was not updating its arp cache with the HSRP address. However, due to the fact that there are media adapters in the mix, I removed any ether-channeling between the switches and allowed spanning-tree to block one link and forward on another. Ether-channeling will never work properly with the adapters should a fiber link ever fail.