|
Here's the setup I've got
(63.86.125.41) (63.86.125.43)
D1 ----------- D2
(192.168.0.3) | (192.168.0.4)
|
|
|
|----------------|
| |
R1 R2
(192.168.0.100) (192.168.0.101)
All four of these machines are connected to the same hub. The default
route for R1 and R2 is 192.168.0.3 so all packets must go back out through
one of the directors, whichever has the default gateway IP. The only
router involved here is the default gateway for D1 and D2 which should be
totally outside the loop as far as the real servers are concerned, right?
When D1 fails, D2 takes over 63.86.125.41 and 192.168.0.3 as well as all
VIP's currently on D1. The VIP's can be changing dynamically so this is
accomplished by rdist'ing a config file between D1 and D2. This config
file is then executed on D2 when D1 fails. The config file simply calls
/etc/ha.d/rc.d/IPaddr for each IP to be taken over.
Any outside client can ping all of the IP's taken over by D2. All the
real servers can ping all of the IP's taken over by D2, both private and
public addresses, but none of the real servers can ping out to the public
net.
Brad Benson
On Mon, 9 Oct 2000, Steve Gonczi wrote:
> Date: Mon, 9 Oct 2000 14:43:29 -0400
> From: Steve Gonczi <Steve.Gonczi@xxxxxxxxxxxxxxxxxx>
> To: 'Brad Benson' <brad@xxxxxxxxxxx>
> Subject: RE: routing issues
>
> I was more concerned about your router. Does it know about the address
> change?
> I assume that the address you are taking over is a VIP ( an advertised
> externally
> visible address), so it would be important that your router's ARP
> table is correct.
>
> /sG
>
|
