[Xorp-hackers] OSPF Problem with Running Multiple Instances of XORP

[Ben Greear]

Pavlin Radoslavov wrote:
> In Linux interface name like "eth0.2" is not real interface, but
> an IP alias address on interface eth0.
> Hence, in your XORP config the interface/vif name should be eth0,
> and have the corresponding IP aliases within the vif block.
>   

ethX.Y can be an 802.1Q VLAN naming configuration, and he mentions
using vconfig, so at least that part of his config should work OK.

>> The problem is OSPF processes can't communicate with each other because OSPF
>> Hello multicast messages are not able to reach other virtual interfaces.
>>     
Two VLAN devices on the same ethernet port cannot talk to each other 
because pkts
transmitted out that physical interface will not normally be returned.  
They will NOT be
routed to other local VLAN interfaces by the local OS.

If you want two XORPs to talk to each other on the same machine, you 
will also
need clever routing table rules and linux patches, use something like 
the 'veth' driver
or redirdev (in my kernel tree only), and so forth.   We (Candela 
Technologies) have
a product that sets up everything, but it's not normally free unless you 
are a student working on
your own hardware.  Each individual piece of getting this to work isn't
so hard, but there are a lot of details, so I can't really explain it 
all in any detail.

If you just want two XORPs to run on the same system, but NOT talk to 
each other, then
that is probably easier and shouldn't require any kernel patches or 
overly complex routing
table rules.

Our kernel and xorp can be obtained with the commands below (no 
passwords required, should be read-only).
I just checked these...I sent out bad syntax that required ssh access in 
the past, so use
this instead:

git clone git://dmz1.candelatech.com/xorp.git
git clone git://dmz1.candelatech.com/linux-2.6.dev.25.git

Thanks,
Ben

-- 
Ben Greear <greearb at candelatech.com> 
Candela Technologies Inc  http://www.candelatech.com