[j-nsp] Routing Instance for fxp0

Chris Evans chrisccnpspam at gmail.com
Sat Nov 28 20:15:10 EST 2009


Alfred,

We do exactly what you're saying with our Cisco environment, the only way to
do it currently on any JUNOS platform is to setup a 'virtual router', not a
routing instance... While this might take a tiny bit more resources it will
allow you to put the FXP interface into it and creates a seperate routing
table. I for the life of me cannot understand how this is not a feature yet
of JUNOS.. I've put in a feature request to allow us to apply the FXP
interface into a routing-instance.. We are piloting the EX series of
switches and they do not support virtual routers so we are kinda screwed
with that platform.. Another option, which is another workaround is to put
all of your PFE interfaces into a routing-instance and leave the FXP in the
global default VRF. If you have a ton of interfaces this is quite cumbersome
though.

For the list member that suggested to use a PFE interface, I find this an
unacceptable solution. Interfaces are NOT CHEAP by any means, I just need a
simple 10mb port to manage a box, I shouldn't need to invest 10K+ for an
interface to allow me to manage a box with a seperate routing table.


Let me know if you need help, I can provide a config if need be.. Start off
with 'edit virtual-router' and you should be able to poke your way around. I
have tested this on our M7i and M10i devices and confirm it works.


Chris


______________

Hello

>* So the actual goal is not the OOB vs PFE-forwarded traffic separation but
*>* rather the opposite?
*
My goal is a totaly separation of OOB and PFE.

At a cisco (12k, crs, ...) VRF lite configuration is a (the?) way to do this.
Perhaps juniper has a complete different way to do this ....

>* Even if you are able to achieve what you are aiming at (forward traffic
*>* between PFE interfaces and fxp0 interface) then at certain rate it would
*>* get throttled by PFE CPU. And this behaviour (PFE CPU throttling) is not
*>* configurable AFAIK.
*
Thanks and regards,

-- 
ALF


More information about the juniper-nsp mailing list