[c-nsp] Bad routes in MPLS

[Saku Ytti]

On (2013-11-23 14:06 -0800), Tony wrote:

> No, we don't but sounds like perhaps it would allow us pick up stuff like this sooner ? Apart from added complexity are there any reasons not to ? 

Downside is, you'll notice more problems. Obviously the problems are there
regardless of mls ratelimiters, but they make the problems more pronounced.
Without them problem might be some hard to troubleshoot issue, where customer
complains jitter or latency. With MLS ratelimit the problem becomes much more
pronounced.

> > I would also have been interested in flags of the adjacency.
> 
> Do you means the flags as shown below (with the "detail" option), or something else ?
> Index: 311374  smac: 0013.1abf.2280, dmac: 0012.7fee.ed40
>                mtu: 1548, vlan: 1099, dindex: 0x0, l3rw_vld: 1
>                format: MPLS, flags: 0x8418

Yes, these too look fine:
X#sh mls cef adjacency flags 8418 0

TTL_DEC NO_CACHE L3_RW_VALID L3_RW

> Possibly next time when it isn't 0700 Sunday morning and I don't have children jumping all over me :)
> I'm also hoping there isn't a next time before upgrade in 3 days time...

ELAM is super fast to do too, takes like 1min if you can differentiate traffic
with simple triggers like DADDR/SADDR. If packet is coming from MPLS side,
it's lot more work, as there isn't simple trigger for label, so you must do
bit-pattern match.

-- 
  ++ytti