This is a pretty simple scenario:
R1--------------- R2
| |
| |
+-------R3----------+
R1 and R2 are directly connected, one-hop MPLS LSP established between the 2 routers. The primary path is a one-hop path and the secondary path is a de-tour path via R3.
Every time when I put the link between R1 and R2 down, the LSP path will be swing to the secondary path. Then I brought the link back, then it will take quite a while (usually tens of seconds) for R1 to mark the original primary path to be "up":
For example:
===========
The interface was up at 10:54:12:
Apr 22 10:54:12 m120-b-re0 mib2d[1571]: SNMP_TRAP_LINK_UP: ifIndex 628, ifAdminStatus up(1), ifOperStatus up(1), ifName so-5/0/1.0
But the primary path was up at 10:54:46:
18 Apr 22 10:54:46.701 Selected as active path
17 Apr 22 10:54:46.699 Record Route: 1.2.3.4(flag=0x20) 10.0.0.1(Label=3)
16 Apr 22 10:54:46.699 Up
============
Here is my MPLS configuration:
lab@m120-b-re0> show configuration protocols mpls
inactive: traceoptions {
file mpls-trace_2 size 10m files 10;
flag error;
}
no-propagate-ttl;
ipv6-tunneling;
label-switched-path NDLTT-ER1-to-CHENTT-ER2 {
traceoptions {
file mpls size 10m files 10;
flag all;
}
to 202.163.59.107;
ldp-tunneling;
retry-timer 5;
revert-timer 5;
no-cspf;
link-protection;
primary to-R2-direct;
secondary to-R2-via-R3;
}
path to-R2-direct {
10.0.0.2 strict;
}
path to-R2-via-R3 {
10.1.0.2 strict;
10.2.0.2 strict;
}
interface all;
interface fxp0.0 {
disable;
}
I knew that the signaling for establish the primary path back would take time, but for a one-hop path with static strict next-hop configured, 30+ seconds could be a little bit longer than expected?