ARP Issue

[John Apricena]

Hello All,

We are currently using two load balancers for handle failover for email. The load balancers use a "floating IP" along with the mac address of the "active" node. When we fail the active node over to passive, I can no longer ping the floating IP from the gateway router. After doing some troubleshooting, I found that the ARP table keeps the same mac address of the active node even after I fail it over. When I perform a clear arp on the router it fixes the issue, and the new active node gets rediscovered on the network. Does anyone know why this may be and how I could resolve, without having to make the ARP timeout like 5 seconds instead of the default of four hours.

Thanks in advance!

[Bilal Nawaz]

Hi John, the load-balancer's normally proxy arp for the VIP address - primary or secondary, whichever is active should take this role of proxying for it. Is this an option that we can check for?

It may be that we have to have a timeout for 5 seconds, but I do not think that this is normal.

hth

Please rate useful posts and remember to mark any solved questions as answered. Thank you.

[Richard Burts]

John

Am I understanding correctly that when the load balancer fails over that the floating IP should then become associated with the different MAC address of the backup server? This suggests that the server that has just become active should send out an ARP update associating the floating IP with its MAC address.

If the router is not learning the new MAC address then it might be that the advertisement of the MAC change is not getting to the router, or that the advertisement gets there but that the router does not process it.

As a first step in investigating this I would suggest that you enable terminal monitor on the router (and be sure that the level of monitor is set to debug). Then turn on debug arp on the router. Then cause a failover of the load balancer. The output of debug arp should show whether an advertisement of the MAC change is received by the router.

If the advertisement is not received then your investigation needs to look into why it was not received. If it was received but not implemented then your investigation needs to focus on why it was not implemented.

HTH

Rick

[John Apricena]

Hey Guys, thanks for the prompt responses! The vendor of the load balancer's is saying that cisco equipment has an ARP/cache that could be prventing the passive node from regsiterering it's new mac address. Could this be possible? I will follow the instructions you both laid out for me in our next outage window. Thank Again!