new CARP implementation

[Andre Oppermann]

On 14.10.2011 21:12, Gleb Smirnoff wrote:
> On Fri, Oct 14, 2011 at 09:06:57PM +0200, Andre Oppermann wrote:
> A>  >  Won't ifconfig it down work?
> A>
> A>  Not if I want to continue to use the primary or other non-carp address.
>
> Well, the best idea for that case would be switch them to backup mode.
> In presense of active master, they won't show up any way either on wire,
> neither in the stack.

Though you run the risk of getting a lot of traffic when the master fails
while working on the backup.  At the moment one can simply down the carp
interface and be in the clear.  Anyway, this enable/disable can be added
later again if reality shows it to be necessary.

> A>  >  A>   >   btw, you can configure a VHID and don't attach any addresses
> A>  >  A>   >   to it.
> A>  >  A>
> A>  >  A>   How does that work? What purpose does it serve?
> A>  >
> A>  >  This is actually how it is configured: SIOCSVH, then SIOCAIFADDR.
> A>  >  There is no practical use in addressless vhid now.
> A>
> A>  If it can cause problems or non-standard behavior it should not be
> A>  allowed and the carp status should be suspended until fully configured.
>
> Status of such addressless vhid is always INIT, so it causes no activity
> either on wire or in stack :)
>
> A>  One more question: Can one interface with multiple addresses become a
> A>  member of different carp groups (with a different address each)?
>
> Sure. All imaginable options are configurable: arbitrary number of
> addresses on interface, arbitrary number of vhids on interface,
> arbitrary number of addresses configured to a particular vhid.

Excellent.  And much simpler to use as well.

-- 
Andre