[j-nsp] Juniper mpls fast-reroute and pfe

[Piotr Marecki]

ehlo

It is stated in documentation that whenever some of fast-reroute features is
used ( one-to-one or facility backup) is used , pfe has local information
precomputed so it can use detour or bypass when  a path failure occurs
without further RE signalization.Documentation also states that PFE local
repair is "enabled by default and requires no additional configuration".Let
see..

I have rsvp signalled lsp with fast-reroute option enabled
label-switched-path to-lab-m5-2-from-gw1 {
    to 10.127.0.44;
    bandwidth 150m;
    adaptive;
    fast-reroute;

On next router , as expected we can see that cspf algo found detour for that
path and following label values are active on RE :

[from show mpls lsp extensive]
  Resv style: 1 SE, Label in: 102000, Label out: 103648
  Detour Label out: 102448

Quick check in mpls.0 table
mpls.0: 18 destinations, 18 routes (18 active, 0 holddown, 0 hidden)
102000 (1 entry, 1 announced)
        *RSVP   Preference: 7
                Next hop: via so-0/0/0.0 weight 1, selected
                Label-switched-path to-lab-m5-2-from-gw1
                Label operation: Swap 103648
                Next hop: via so-1/0/0.0 weight 10001
                Label-switched-path to-lab-m5-2-from-gw1
                Label operation: Swap 102448

So , on RE everything is ok.Then we check PFE table :
[from show route forwarding-table label 102000 extensive]
Destination:  102000
  Route reference: 0                   Route interface-index: 0
  Flags: sent to PFE
  Next-hop type: Swap 103648
  Next-hop interface: so-0/0/0.0

So , actually there is no information on PFE about detour label.After adding
per-flow load-balancing export policy in
[edit routing-options forwarding-table] , everything is ok
[from show route forwarding-table label 102000 extensive]
Destination:  102000
  Route type: user
  Route reference: 0                   Route interface-index: 0
  Flags: sent to PFE
  Next-hop type: unilist               Index: 296      Reference: 1
  Next-hop type: Swap 103648
  Next-hop interface: so-0/0/0.0       Weight: 1
  Next-hop type: Swap 102448
  Next-hop interface: so-1/0/0.0       Weight: 10001

So question is  if  PFE local repair indeed requires some additional
configuration ( per flow load-balancing) which is not stated in doc's or
there is something more that can't be seen by show route forwarding-table
( and in that case per flow load-balancing is not required) ?

thanks in advance

Piotr Marecki
Netia S.A