[j-nsp] IS-IS bug in 7.3?

[Gordon Smith]

Hi all,

Just came across something interesting while testing IGP convergence and
configuration.

I have a backbone network segment that has both Juniper and Cisco
devices on it.
The routers exchange loopback address routes via IS-IS, then stand up
BGP adjacencies to the loopbacks of neighbouring routers. This works as
intended.

The problem I've come across is in JUNOS 7.3 code, and concerns IS-IS
hold timers.

The default should be 27 sec hold-time with a 9 sec hello-interval.
JUNOS 7.2R1.7 and Cisco IOS honor this.
In JUNOS 7.3R1.6 however, the hold-time is drastically reduced and will
not accept a change via config.

This is what I should see:

fe-0/0/0.0
  Index: 66, State: 0x806, Circuit id: 0x2, Circuit type: 2
  LSP interval: 100 ms, CSNP interval: 10 s
  Level Adjacencies Priority Metric Hello (s) Hold (s) Designated Router
    2             7       64     10     9.000       27 jcore2.02 (not
us)

But on the mis-behaving box I get:

IS-IS interface database:
fe-0/0/0.0
  Index: 75, State: 0x806, Circuit id: 0x2, Circuit type: 2
  LSP interval: 100 ms, CSNP interval: 10 s
  Level Adjacencies Priority Metric Hello (s) Hold (s) Designated Router
    2             7       64     10     3.000        9 jcore2.02 (us)

Doing this:

interface fe-0/0/0.0 {
    checksum;
    level 2 {
        hello-interval 9;
        hold-time 27;
    }
}

Results in no change whatsoever...

Bug in the code? Anyone else seen this? Even restarted rpd, but to no
avail...

An old piece of documentation on Juniper's site (for 4.4) states that 9
sec hold-time is the default, but that's not what I'm seeing here.

Cheers,
Gordon