[j-nsp] EXP based LSP selection

Dragan Jovicic draganj84 at gmail.com
Sun Aug 14 18:08:03 EDT 2016


Hi,

But since the policy is matching everything, then all routes in inet.3 will
> be altered right?
>

More importantly all the routes in mpls.0. Since this is most likely a
bgp-free router it could very well not use inet.3. But you can be more
specific in your policy, see bellow.

I suspect that this is how the inet.3 table looks like for all the prefixes
> in it.
> But then I'm wondering how come the return traffic for ping to
> 120.10.255.1 is not caught by the policy and looped back? Do you know what
> I mean please?
> Or maybe if the tunnel is not terminated at the PE (120.10.255.1) but
> exists only in the core and all /32 routes in inet.3 have these two LSPs
> associated with them wouldn't the traffic loop when passing through this
> core router?
> Or is there some kind of split horizon rule, i.e. if packet cam through
> interface-A it can't be sent out via int-A even if inet.3 says it should?
> If the policy would have been applied to an interface then you'd have a
> direction but if it's applied in the forwarding table it has to be applied
> to all directions no?
>

I understand. No worries, there can be no loop. The policy is forced only
for those LSPs which are best path for the specific prefix/label only. In
our case the policy does not even consider two LSPs for returning traffic.

So I added two more LSPs for returning traffic in other direction:

# run show route table inet.3

inet.3: 20 destinations, 26 routes (4 active, 0 holddown, 20 hidden)
+ = Active Route, - = Last Active, * = Both

120.10.255.1/32    *[RSVP/7/1] 2d 08:29:37, metric 1000
                    > to 120.10.50.2 via ge-1/0/9.500, label-switched-path
lsp-be
                      to 120.10.50.2 via ge-1/0/9.500, label-switched-path
lsp-realtime
                    [LDP/9] 11:41:34, metric 1000
                    > to 120.10.50.2 via ge-1/0/9.500, label-switched-path
lsp-be
                      to 120.10.50.2 via ge-1/0/9.500, label-switched-path
lsp-realtime
120.10.255.5/32    *[RSVP/7/1] 00:01:47, metric 1000
                      to 120.10.110.2 via ge-1/0/9.1100,
label-switched-path lsp-be-return-traffic
                    > to 120.10.110.2 via ge-1/0/9.1100,
label-switched-path lsp-realtime-return-traffic
                    [LDP/9] 00:00:13, metric 1000
                      to 120.10.110.2 via ge-1/0/9.1100,
label-switched-path lsp-be-return-traffic
                    > to 120.10.110.2 via ge-1/0/9.1100,
label-switched-path lsp-realtime-return-traffic

The returning traffic uses another set of LSPs trough our transit core.
Modified policies:


# show policy-options policy-statement lsp-policy1
from rib mpls.0;
then cos-next-hop-map cos-map1;

# show class-of-service forwarding-policy
next-hop-map cos-map1 {
    forwarding-class BE {
        lsp-next-hop [ lsp-be lsp-be-return-traffic ];
    }
    forwarding-class REALTIME {
        lsp-next-hop [ lsp-realtime lsp-realtime-return-traffic ];
    }
}

Test:

# run ping 120.10.255.1 source 120.10.255.5 count 200 rapid tos
160
PING 120.10.255.1 (120.10.255.1): 56 data bytes
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
--- 120.10.255.1 ping statistics ---
200 packets transmitted, 200 packets received, 0% packet loss
round-trip min/avg/max/stddev = 0.613/0.738/4.228/0.305 ms

# run show mpls lsp statistics
Ingress LSP: 4 sessions
To              From            State     Packets            Bytes LSPname
120.10.255.1    150.1.3.1       Up              0                0 lsp-be
120.10.255.1    150.1.3.1       Up            200            17600
lsp-realtime
120.10.255.5    150.1.3.1       Up              0                0
lsp-be-return-traffic
120.10.255.5    150.1.3.1       Up            200            17600
lsp-realtime-return-traffic


# run ping 120.10.255.1 source 120.10.255.5 count 200 rapid tos 0
PING 120.10.255.1 (120.10.255.1): 56 data bytes
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
--- 120.10.255.1 ping statistics ---
200 packets transmitted, 200 packets received, 0% packet loss
round-trip min/avg/max/stddev = 0.588/0.742/2.125/0.225 ms

# run show mpls lsp
statistics
Ingress LSP: 4 sessions
To              From            State     Packets            Bytes LSPname
120.10.255.1    150.1.3.1       Up            200            17600 lsp-be
120.10.255.1    150.1.3.1       Up            200            17600
lsp-realtime
120.10.255.5    150.1.3.1       Up            200            17600
lsp-be-return-traffic
120.10.255.5    150.1.3.1       Up            200            17600
lsp-realtime-return-traffic


Regards

Dragan


More information about the juniper-nsp mailing list