Date: Fri, 18 Feb 2005 11:41:38 +0300 From: Gleb Smirnoff <glebius@freebsd.org> To: Marko Lerota <marko.lerota@optima-telekom.hr> Cc: freebsd-net@freebsd.org Subject: Re: redundant ethernet interfaces Message-ID: <20050218084138.GB4027@cell.sick.ru> In-Reply-To: <86ll9mcm4y.fsf@redcloud.local> References: <86ll9mcm4y.fsf@redcloud.local>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Feb 18, 2005 at 09:23:09AM +0100, Marko Lerota wrote:
M> I have two nic bge0 and bge0, and I would like them
M> to have the same IP address in case that my redundant
M> switch stop responding. Bridge is not an option,
M> I tried with netgraph but I did not succeed.
M> Does anyone have some solution?
M>
M> here is my config
M>
M> rc.conf
M> ifconfig_bge0="1.2.3.4 netmask 255.255.0.0 media 100baseTX mediaopt
M> full-duplex"
M> ifconfig_bge1="media 100baseTX mediaopt full-duplex"
M>
M> and /usr/local/etc/rc.d/bonding.sh
M>
M> #!/bin/sh
M> ifconfig bge1 up
M> kldload /boot/kernel/ng_ether.ko
M> kldload /boot/kernel/ng_one2many.ko
M> ngctl mkpeer bge0: one2many upper one
M> ngctl connect bge0: bge0:upper lower many0
M> ngctl connect bge1: bge0:upper lower many1
M> ngctl msg bge1: setpromisc 1
M> ngctl msg bge1: setautosrc 0
M> ngctl msg bge0:upper setconfig "{xmitAlg=1 failAlg=1 enabledLinks =[ 1
M> 1 ] }"
M> I followed the examples from google groups and I stuck with
M> duplicated packets when both ethernets are plugged and
M> 50% packet loss when one ethernet is unplugged, damn !!#@!#@!$%^@$#@
It is very strange that you have duplicates. Do you see duplicates of
incoming or outgoing packets?
50% packet loss is because to ng_one2many can't detect which link is
active and which is not. We have a solution for this in HEAD, but I'm
very sorry we can't MFC this to RELENG_5 due to ABI freeze. You can bring
up patches from HEAD to your system.
Actually, I'd better recommend you the following setup:
ifconfig bge0 up
ifconfig bge1 up
kldload /boot/kernel/ng_ether.ko
kldload /boot/kernel/ng_one2many.ko
kldload /boot/kernel/ng_eiface.ko
ngctl mkpeer bge0: one2many lower many0
ngctl connect bge1: bge0:lower lower many1
ngctl msg bge1: setenaddr ${MAC-address-of-bge0}
ngctl mkpeer bge0:lower eiface one ether
ngctl msg bge0:lower.one set ${MAC-address-of-bge0}
ngctl msg bge0:lower setconfig "{xmitAlg=1 failAlg=2 enabledLinks =[ 1 1 ] "
ifconfig ngeth0 1.2.3.4 netmask 255.255.0.0
Notice that I use failAlg=2, to make ng_one2many listen to flow control messages.
However, ng_ether can't send these messages in RELENG_5, see above.
And this setup is not guaranteed to work flawlessly unless you have a smart switch
and this two cables are plugged into one trunk group on switch.
M> manual for this sucks
I plan to write a handbook chapter on netgraph sometime this year.
--
Totus tuus, Glebius.
GLEBIUS-RIPN GLEB-RIPE
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050218084138.GB4027>