[Linux-HA] Autofailback problem

[Horms]

On Sat, Jun 12, 2004 at 08:40:53PM +0100, Colin Bruce wrote:
> Dear Alan,
> 
> Thanks for the e-mail.
> 
> I should have said this is a test setup at the moment and is actually two old
> PCs connected to a CISCO 1900 switch. There are no firewalls anywhere near
> them I'm afraid. In fact both machines are grabbing both the addresses and
> neither seems to think there is a problem with that. Also as I said if I
> type
> 
>     ./IPaddr 192.168.255.7 stop
> 
> then not only does the eth0:0 which is 192.168.255.7 stop but so does
> 192.168.255.22 (the other address). It seems to work fine if I just create
> a single service address and have a master/slave arrangement.
> 
> I am sure there is a configuration problem but I have never seen a
> configuration file that works. The example I was advised to use earlier
> doesn't work with version 1.2.2 of heartbeat as far as I can see.

When you say stops, what do you mean?

The command above should only effect 192.168.255.7. You can verify this
in a couple of ways.

1. Run bash -x ./IPaddr 192.168.255.7 stop and sift through the output

2. Take a look at the output of ifconfig before and after running this 
   command.

3. Run tcpdump, ethereal or something similar while the above
   command (with or without bash -x) runs and look for gratuitous
   ARP packets. You should see they are only for 192.168.255.7.

4. Inspect the ARP tables on machines/switches/routers/whatever
   that are communicating with 192.168.255.7. The Cisco switch
   you mention above might be a good start.

   On Linux (and probably other similar OSes) you can inspect the
   arp table by running arp -n (the -n tells it not to resolve hostnames)

   On Cisco IOS I believe the command is show arp

-- 
Horms