Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 8 Aug 2008 22:57:49 -0700
From:      Andrew Thompson <thompsa@FreeBSD.org>
To:        Jacob Owens <jacoblowens@gmail.com>
Cc:        freebsd-net@freebsd.org
Subject:   Re: lagg failover not automatic
Message-ID:  <20080809055749.GA95107@citylink.fud.org.nz>
In-Reply-To: <c292b6660808082149m421d1a96q53d8a6934e11ed70@mail.gmail.com>
References:  <c292b6660808082149m421d1a96q53d8a6934e11ed70@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Aug 08, 2008 at 11:49:27PM -0500, Jacob Owens wrote:
> Hello.
> 
> I've got a old sun V100 which features two gigabit ports (using dc driver).
> I've been trying to get lagg failover working. on SPARC64 7.0 RELEASE
> 
> On the box I put the following in /etc/rc.conf:
> ifconfig_dc0="UP"
> ifconfig_dc1="UP"
> cloned_interfaces="lagg0"
> ifconfig_lagg0="laggproto failover laggport dc0 laggport dc1 50.40.0.3netmask
> 255.255.0.0
> 
> I even added this to my loader.conf (per the man page)
> if_lagg_load="YES"
> 
> I'm not using a smart switch, so no STP.
> 
> What happens when I unplug the "master" interface, is that the network will
> stop working. the second I type in 'ifconfig -v' to see what the situation
> is, there is a small pause, and then the network fails over to the second
> nic and starts working. So it seems that typing 'ifconfig' somehow wakes the
> config up... Dmesg does not show the appropriate update (dc0: link state
> changed to DOWN/dc0: link state changed to UP) until after i type ifconfig
> either. Here is the before and after outfut of 'ifconfig -v'

lagg (failover/loadbalance) relies on the hardware/driver reporting the
link state changes to do failover, its the only way it can tell if the
link has gone down. Running ifconfig will cause the hardware media to be
polled as it needs the current state for the 'media: ...' line, this
will cause the needed linkstate event to happen since it is different to
the stored value. This is meant to happen automatically, this is a bug
in the dc driver.

One thing to note is that LACP mode does not have this problem as
heartbeat frames are exchanged with the peer, apart from being slower as
its a timeout vs link event.


cheers,
Andrew



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20080809055749.GA95107>