[c-nsp] IOS XR and MPLSoMGRE

[mihai at duras.ro]

Thanks Adam.
Rest inline.

On 2018-05-17 10:18, adamv0025 at netconsultings.com wrote:
>> mihai at duras.ro
>> Sent: Wednesday, May 16, 2018 10:57 PM
>> 
>> Hi guys,
>> 
>> I'm trying something strange and funky in a lab and of course ending 
>> up at
> a
>> dead-end :)
>> 
>> I'm trying to configure OpenContrail to peer with ASR9K.
>> I've already found a sample configuration for ASR903 but that one has 
>> IOS
> XE.
>> 
>> Does anyone know if there is an equivalent for the following commands 
>> on
>> IOS XR?
>> So far I can't find anything and am knocking my head against the wall 
>> a
> bit (I
>> did find that MGRE should be supported though and MPLS over it).
>> 
>> l3vpn encapsulation ip MGRE
>>   transport ipv4 source GigabitEthernet0/4/7
>> 
>> route-map SELECT_UPDATE_FOR_L3VPN permit 10
>>    set ip next-hop encapsulate l3vpn MGRE
>> 
> ASR9k does support mGRE but only in Draft-Rosen-MVPNs.
> -so the tunnel endpoints and the necessary overlay encapsulation is all
> determined via BGP (so no route-maps or NHRP).
> 
> You could try to follow this guide:
> https://www.cisco.com/c/en/us/td/docs/routers/crs/software/crs_r6-1/lxvpn/co
> nfiguration/guide/b-lxvpn-cg-crs-61x/b-lxvpn-cg53xcrs_chapter_0100.html
> and just substitute L2TPv3 with GRE encapsulation and see how far you 
> can
> get.
> -but the guide is for CSR and I haven't found anything relevant on 
> L2TPv3 on
> ASR9k.

I have a feeling that in the end my struggle would bring more headaches 
than
results.
The light at the end of the tunnel is not clearly visible :)

> 
> Well you should ask Juniper folks as they claim vendor neutral support
> -would like to see how they tested this with XR :)
> Most likely p2p GRE.

I suspect this was not tested with IOS XR or at least from what can be 
found
via the Opencontrail site.

> 
> I really like the main idea of Contrail (or tungsten) - MPLS L2/L3VPNs 
> all
> the way down to VMs, but the execution is terrible.
> Why MPLSoUnicorns?? -and then you got to have this nonsense at the DCI 
> PEs
> trying to stitch whatever crazy overlay encapsulation DC folks come up 
> with
> with the good old MPLS VPNs in the core (it makes my life miserable).
> 

Totally agreed but you know, it's this crazy industry push..does not 
really
follow logic.

> If you have the chance (new DC POD, green-field deployment), then go 
> with
> the unified fabric - MPLS from DC-to-Core-to-Aggregation it's very easy 
> to
> deploy and you'll love it.

No green-field but rather a PoC with this mixed setup..not too much
flexibility to go around.

> And then if you want to do the contrail thing just spin up vMX or 
> XR9000v on
> the compute hosts -this way you can build your DC fabric using very 
> cheap
> P-router type switches.
> And if VMx or XR9000v eats too much resources on your Host then you can 
> do
> the proper contrail thing but instead of vRouter you can program OVS 
> from
> the ODL (yes the same SDN controller you use for the MPLS core -aint 
> that
> nice) to do the L2/L3VPN label pushing/popping.
> 
> Just waiting for the DC world to wake up and realize Frame-Relay et al. 
> is
> dead -that's why we have MPLS now.

:)) I'm just wondering what other crazy thing will follow and re-invent 
the
wheel :)

Thanks for the long, detailed e-mail and for all the help!!
mihai

> 
> adam
> 
> netconsultings.com
> ::carrier-class solutions for the telecommunications industry::