[plug] Ethernet Bonding

[Adrian Woodley]

Have thought about that, but the convergence time is a bit more that 
we'd like. Also it requires explicit support on the switch (not such a 
problem really I guess).

We also considered running quagga on the servers and dynamic routing 
(OSPF?) all the why to the server. But again, the convergence time is a 
little long.

Bonding works really well, provided you make sure nothing leaves the 
standby interface(s).

Having separate mac's for all interface involved would largely mitigate 
this.

Adrian Chadd wrote:
> Then run STP and bridging, then just treat one interface as down
> and one as up.
> 
> I remember there being a tweak in the stuff you're probably using
> under Linux to dictate which interface(s) traffic goes out.
> Ask me after exams. :)
> 
> 
> 
> Adrian
> 
> On Thu, Oct 18, 2007, Adrian Woodley wrote:
>> Only if you're doing link aggregation (which I'm not), not 
>> active-standby (which I am).
>>
>> With active-standby, the switch(es) should only see traffic from one 
>> interface (and therefore one mac-address). If the primary link fails, 
>> traffic will start emanating from the standby interface. Regardless of 
>> whether the mac-address is the same (no arp needed, but switch(es) have 
>> to re-learn mac/switchport relationship) or different (switch should 
>> already know mac-address of standby, but hosts will need to re-arp for 
>> mac to IP relationship) there is no special config needed on the switch(es).
>>
>> The problem I'm looking to avoid the problems caused by traffic leaving 
>> a standby interface with a source mac-address of the bond interface. 
>> This will cause the switch to (incorrectly) associate that mac-address 
>> with the standby interface / switchport.
>>
>> Adrian
>>
>> Senectus . wrote:
>>> On 18/10/2007, *Adrian Woodley* <Adrian at screamingroot.org 
>>> <mailto:Adrian at screamingroot.org>> wrote:
>>>
>>>    By default, all interfaces in a bond group will be assigned the MAC
>>>    address of the first slave interface.
>>>
>>>    Does anyone know a way to have each interface retain its own MAC,
>>>    including the bond interface itself? 
>>>
>>>
>>> Uh, I might be thinking of the wrong thing here, but doesn't that depend 
>>> on the capability of your switch?
>>>
>>>
>>>
>>> -- 
>>> Ubuntu Dapper 7.04
>>> The less you know, the more you believe. - Bono
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> PLUG discussion list: plug at plug.org.au
>>> http://www.plug.org.au/mailman/listinfo/plug
>>> Committee e-mail: committee at plug.linux.org.au
>> _______________________________________________
>> PLUG discussion list: plug at plug.org.au
>> http://www.plug.org.au/mailman/listinfo/plug
>> Committee e-mail: committee at plug.linux.org.au
>