Floating ip is not pingable after failover in a Clustering Setup

[Amit R]

We have 2 linux boxes which form part of Clustering solution.The active host has a floating ip assigned to it. However whenever there is failover the floating ip is brought up on the the formerly-standby and the now-to-be-active host.The floating ip is assigned to a logical interface (which has mac address of the physical interface, as a result when failover happens the floating ip on the new host has mac address corresponding to the physical interface of the new host)
However when we ping from outside , say from a separate subnet,the ping fails after the failover even when ip is brought up on the new host.
It seems for some reason the CISCO switch doesnt update its arp table or the mac address resulting in failure of the ping.

Any suggestions/solutions making the floating ip failover transparent such that the pings are not lost when the failover happens will be great.

[Steven E. Protter]

Shalom,

Check if its pingable from the node its supposedly configured on.

If not: Cluster configuration or cluster manager is not working correctly. See /var/log/messages

If yes:
Check externally. Check main network configuration is correct because you can not specify netmask and such in the cluster.conf file.

CISCO routers may need to be configured to have the same IP address appear on two ports. That would make this a CISCO problem, not a Linux problem.

Make sure the Linux system is patched correctly, a bad kernel release could contribute to this problem.

SEP

[Amit R]

Thanks for the reply,
I think I complicated the matters.
I am not using Clustering Solution of Linux but something else.To simplify on failover we do bring down the ip on the failing node and bring it up on the new node.
However if some one is continously pinging it from outside (diff subnet) , the pings hang when it is brought down on the old host and brought up on the new host.Note that the ip is up when u do ifconfig on the new node.

[Amit R]

The following command exactly does what I wanted.
It sends an arp broadcast to all .In the process the CISCO updates its arp table to the new mac address associated with the floating ip on the new host.

arping -U -c -I
However one needs to mention the interface else it takes the default interface say eth0 and is of no use.
Also one needs to specify count else its infinite like ping.



arping [ -AbDfhqUV] [ -c count] [ -w deadline] [ -s source] -I interface destination

DESCRIPTION

Ping destination on device interface by ARP packets, using source address source.
OPTIONS

-c count
Stop after sending count ARP REQUEST packets. With deadline option, arping waits for count ARP REPLY packets, until the timeout expires.

-I interface
Name of network device where to send ARP REQUEST packets. This option is required.
-U
Unsolicited ARP mode to update neighbours' ARP caches. No replies are expected.