[CentOS] CentOS Digest, Vol 91, Issue 7

Tue Aug 7 16:13:07 UTC 2012
Cal Sawyer <cal-s at blue-bolt.com>

On 07/08/12 17:00, centos-request at centos.org wrote:
> Message: 6
> Date: Mon, 06 Aug 2012 21:34:43 +0200
> From: Alexander Dalloz <ad+lists at uni-x.org>
> Subject: Re: [CentOS] Configure LAGG Interface?
> To: centos at centos.org
> Message-ID: <50201C53.2070300 at uni-x.org>
> Content-Type: text/plain; charset=ISO-8859-1
>
> Am 06.08.2012 19:44, schrieb Rainer Duffner:
>> > 
>> > Am 06.08.2012 um 19:22 schrieb Cal Sawyer <cal-s at blue-bolt.com>:
>> > 
>>> >> In my experience, LAG/LACP won't provide aggregatation, only failover
>>> >> and fault tolerance.  For link aggregation, you don't need to configure
>>> >> the switch ports - just set bonding to mode=6 for balanced
>>> >> transmit/receive and plug up the the NICs to a group of ports on the
>>> >> switch.  However, balance-alb doesn't help with single stream rsync/FTP
>>> >> sessions, etc, but helps a lot with concurrent transmits/receives as
>>> >> encountered in typical fileserver scenarios.
>>> >>
>> > 
>> > 
>> > On FreeBSD, you don't get 2*1 Gbit from A to B, but 1*1 Gbit from A to B and another 1*1 Gbit from C to B.
>> > "B" being the server with the LAGG interface.
Yup, as i said, which isn't too bad (imho).  "Real" LAG/LACP can't
provide 2x bandwidth for one stream either, can it?
>> > 
>> > How is that in CentOS?
> It is of course the same, as long as we speak about standardized 802.3ad
> protocol.
>
> Alexander
The nice thing about balance-alb is that you don't have to care (or
forget, in my case) about how the switch is configured.  Ever come
across a switch that "someone else" has set up LAGs on but nobody can
remember (or admit) why? )

Throw arpwatch on the bonded-NICs server and see how the bonded
interface flips between the slave MACs as intended

- cal