'[quagga-users 11994] Re: Is this supposed to work (or actually,'

[prev in list] [next in list] [prev in thread] [next in thread] 

List:       quagga-users
Subject:    [quagga-users 11994] Re: Is this supposed to work (or actually,
From:       "Tyler J. Wagner" <tyler () tolaris ! com>
Date:       2010-12-09 11:16:37
Message-ID: 1291893397.2328.36.camel () baal ! talia ! net
[Download RAW message or body]

My constant refrain: disable rp_filter before spending any more time on
this. If it hasn't screwed you already, it will.

http://www.tolaris.com/2009/07/13/disabling-reverse-path-filtering-in-complex-networks/

Just out of curiosity, why failover between two bonds? Why not just make
one big bond and use it?

Regards,
Tyler

On Thu, 2010-12-09 at 10:51 +0000, Arie_Blumenzweig@Dell.com wrote:
> Hi All, sorry if this is a re-post (had some e-mail mishaps lately :-)
> 
>  
> 
> I'm trying to fail over and back a route between two Linux bonds, and
> it doesn't work.   The setup is as follows, the configuration files
> follow:
> 
>  
> 
> 1. I have two nodes, each with two bonds (bond0 and bond1), each with
> three links underneath (bond0: eth20, eth21, eth22 and bond1:eth0,
> eth1, eth2)
> 
>  
> 
> 2. The nodes run zebra and ospfd (config files below)
> 
>  
> 
> 3. I configure IP addr 117.1.1.1/24 and 117.1.1.2/24 on bond0 on the
> two nodes
> 
>  
> 
> 4. I set the links of bond0 on node0 down, one after the other.  As
> expected, packets pass through to the other
> 
>     node (a few netperf UDP_STREAMs from node0 to node1) , being sent
> by the bond driver via one of the remaining available links.
> 
>  
> 
> 5. As soon as all the links in bond0 are down zebra moves the route
> for network 117.1.1.0/24 from bond0 to be routed via one of the 
> 
>     networks on bond1.
> 
>  
> 
> 6. Packets stop passing through (I monitor the counters
> in /sys/class/net/bond[01]/statistics/tx_packets).  I expected the
> packets
> 
>     to be routed via bond1 (was I wrong?)
> 
>  
> 
> 7. As soon as I bring any of bond0.s links up packets get flowing
> again
> 
>  
> 
> I'm aware to the fact that the bonding module doesn't report its link
> status to zebra, so I used netplugd to 
> 
> monitor /sys/class/net/bond0/slave_eth2[0-2]/operstate
> and /sys/class/net/bond0/operstate and:
> 
>  
> 
> 1. ip link set bond0 down - when all its underlying links are down
> 
> 2. ip link set bond0 up - when any of its links comes up
> 
> 3. ifconfig bond0 down - when bond0 gets either out or probed.
> 
>  
> 
> Is my setup supposed to work, is it known to have worked anywhere?
> 
>  
> 
> Your help is greatly appreciated!
> 
>  
> 
> My configuration:
> 
>  
> 
> Node0:zebra.conf
> 
>  
> 
> ! -*- zebra -*-
> 
> hostname node0
> 
>  
> 
> interface eth30
> 
>    ip address  117.192.0.1/10
> 
>  
> 
> interface bond0
> 
>   ip address 117.1.1.1/24
> 
>  
> 
> interface bond1
> 
>   ip address 172.41.52.0/16
> 
>  
> 
> Node0:ospfd.conf
> 
>  
> 
> interface eth30
> 
>   ip ospf cost 100
> 
>  
> 
> interface bond0
> 
>   ip ospf cost 100
> 
>  
> 
> interface bond1
> 
>   ip ospf cost 33
> 
>  
> 
> router ospf
> 
>   ospf router-id 0.0.0.1
> 
>   network 117.192.0.0/10   area 0.0.1.0
> 
>   network 117.1.1.0/24     area 0.0.1.0
> 
>   network 172.41.52.0/16   area 0.0.1.0
> 
>  
> 
> Again, many thanks!
> 
>  
> 
> Arie
> 
>  
> 
>  
> 
>  
> 
>  
> 
> 
> _______________________________________________
> Quagga-users mailing list
> Quagga-users@lists.quagga.net
> http://lists.quagga.net/mailman/listinfo/quagga-users

-- 
"If one would give me six lines written by the hand of the most honest
man, I would find something in them to have him hanged."
  -- Armand Jean du Plessis, Cardinal Richelieu

_______________________________________________
Quagga-users mailing list
Quagga-users@lists.quagga.net
http://lists.quagga.net/mailman/listinfo/quagga-users

[prev in list] [next in list] [prev in thread] [next in thread] 