On Dec 29, 2007 2:59 PM, Aaron <[EMAIL PROTECTED]> wrote:
> Otto Moerbeek wrote:
> > I think your problem will be solved if you assign an alias in the
> > 192.168.3.0 net to fxp0 and an alias in the 192.168.247.0 net to fxp3.
> > Just like Henning already suggested.
> Henning wrote:
>
> >that depends wether you "external" carp interface has numbered or
> > unnumbered parents.
> > if the parents ("carpdev") are unnumbered (no ipassigned),it is quite
> > normal. otherwise you have sth wrong.
> >
>
> I guess I'm missing something or I didn't understand what he means by (no
> ipassigned).
> All carp parents are numbered by the inverse of the definition he gave for
> unnumbered,
> because there are ip's assigned to all of the carpdev interfaces, just not
> with the same
> network as the carp interfaces. Is it required for the carp parents' ip
> addresses
> to be in the same network as the carp interfaces? I didn't see that anywhere
> as a requirement.
The typical configuration is that the CARP interfaces will be assigned
addresses on the same IP subnet as the parent interfaces. I don't
believe that this is a requirement, per se, but it is hinted at in
ifconfig(8):
carpdev iface
If the driver is a carp(4) pseudo-device, attach it to iface. If
not specified, the kernel will attempt to select an interface
with a subnet matching that of the carp interface.
This configuration is the only way that makes sense to me; you don't
have to overlap subnets on the same Ethernet segment, you don't have
to fiddle with interface aliases, and if you need to reach the
"natural" IP addresses for the real (parent) interfaces, they're
routed and reachable the same as the CARP addresses.
Again, not knowing if this impacts your problem, but may be worth testing.
DS
