Can lagg0 failback be prevented?

[Barney Cordoba]

--- On Thu, 9/24/09, Nikolay Denev <ndenev at gmail.com> wrote:

> From: Nikolay Denev <ndenev at gmail.com>
> Subject: Re: Can lagg0 failback be prevented?
> To: "Peter Steele" <psteele at maxiscale.com>
> Cc: freebsd-net at freebsd.org
> Date: Thursday, September 24, 2009, 2:32 AM
> How long exactly is that "momentary
> network outage"?
> If it's longer than a few (1-5) seconds it may be due to
> RSTP or STP enabled on the switch.
> You could try disabling STP on these ports and see if it
> will
> start forwarding traffic faster.
> 
> --
> Regards,
> Nikolay Denev
> 
> 
> 
> 
> On 16 Sep, 2009, at 03:23 , Peter Steele wrote:
> 
> > We're using the lag driver to provide automatic
> failover in case of a network outage. The default
> configuration looks like this:
> > 
> > lagg0:
> flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST>
> metric 0 mtu 1500
> >       
> options=19b<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM,TSO4>
> >        ether 00:a0:d1:e3:58:26
> >        inet 192.168.17.40 netmask
> 0xfffff000 broadcast 192.168.31.255
> >        inet 192.168.22.11 netmask
> 0xffffff00 broadcast 192.168.22.255
> >        media: Ethernet autoselect
> >        status: active
> >        laggproto failover
> >        laggport: nfe1
> flags=0<>
> >        laggport: nfe0
> flags=5<MASTER,ACTIVE>
> > 
> > If nfe0 was to fail, we get an (almost) automatic
> failover to nfe1:
> > 
> > lagg0:
> flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST>
> metric 0 mtu 1500
> >       
> options=19b<RXCSUM,TXCSUM,VLAN_MTU,VLAN_HWTAGGING,VLAN_HWCSUM,TSO4>
> >        ether 00:a0:d1:e3:58:26
> >        inet 192.168.17.40 netmask
> 0xfffff000 broadcast 192.168.31.255
> >        inet 192.168.22.11 netmask
> 0xffffff00 broadcast 192.168.22.255
> >        media: Ethernet autoselect
> >        status: active
> >        laggproto failover
> >        laggport: nfe1
> flags=4<ACTIVE>
> >        laggport: nfe0
> flags=1<MASTER>
> > 
> > The problem we're having is when nfe0 comes online
> again, a failback occurs making nfe0 active again. This
> causes a momentary network outage that we want to prevent.
> Is there a way to configure the lagg device to stay with the
> currently active interface, even if the MASTER interface
> comes back online?
> > 

The "outage" is because the failover isn't detected until you get a 
failure. When a link down is detected, the backup port is enabled. 
So there is going to be some traffic loss. There may be packets in
the transmit queue as well.

I asked earlier why you were using failover if you really don't care
which link it uses? If you load balance it will use both links
unless one of the links is down, in which case it would use one link.
You'd avoid the problem that seems to concern you.

Barney