[plug] Ethernet Bonding

[Craig Foster]

> -----Original Message-----
> From: plug-bounces at plug.org.au [mailto:plug-bounces at plug.org.au] On
> Behalf Of Adrian Chadd
> Sent: Friday, 19 October 2007 12:18 PM
> To: plug at plug.org.au
> Subject: Re: [plug] Ethernet Bonding
> 
> On Fri, Oct 19, 2007, Adrian Woodley wrote:
> > 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.
> 
> Eww. :)
> 
> > Bonding works really well, provided you make sure nothing leaves the
> > standby interface(s).
> 
> Only because you're not running it in actual bonded mode on the
switch!
> Read the http://linux-net.osdl.org/index.php/Bonding stuff; there's
> options
> to control how traffic is sourced.
> 
> > Having separate mac's for all interface involved would largely
> mitigate
> > this.
> 
> It doesn't work as well as you expect. Why not just use etherchannel
> or 802.3ad?
> 
> 
> 
> 
> adrian
> 
> --

+2

Even buying a cheap semi-managed switch supporting link aggregation /
etherchannel / trunking is worth the extra time you're spending on
active-standby which I've found bodgey at best (got it working once,
using separate NICS with a bonded interface with a faked MAC). 

If you've got at least one Intel NIC, you could try using iANS services'
adaptive load-balancing? I know that works with non-trunking switches,
in Windows and Linux. :)

CraigF.