[c-nsp] NCS-5001 - MPLS L3VPN Issue

[James Bensley]

On 9 March 2016 at 20:27, Tom Hill <tom at ninjabadger.net> wrote:
> Presumably to "prove" this (or rather, to add any evidence at all) it's
> wipe & reinstall time for the affected 9001?
>
> I'm not trying to get your back up here - I'm more concerned about FUD
> vs. actual, operational experience. XR is not perfect, but this doesn't
> appear to be a problem affecting anyone other than yourself?

Sure I understand. I've just done an erase and reinstall on the live
router and now its fixed.

Also based on the pre and post upgrade data working with TAC, we think
we have identified a test we didn't run in the lab that would have
triggered the same issues (sadly we don't have infinite time to test
every topology and scenario and feature etc). However TAC had seen
this issue with other customers but couldn't replicate nor could we,
but we do come across issues they haven't seen with other customers so
they might be topology specific.

I'm not going to spend more time trying and replicate this issues now
to see if we did miss a possible method of replication because we have
a fix now which is more important.

On 9 March 2016 at 20:29, Jared Mauch <jared at puck.nether.net> wrote:
> We’ve seen odd issues which we have not yet root caused around the software
> installation and troubles.  Cisco has not been able to reproduce, but we have
> seen it numerous times.  It’s still possible there is some PBCAK but given
> the nature of other issues we’ve seen, it’s unlikely.

>From time to time we have had PEBCAK issues; like you're not supposed
to run vanilla IPv4 unicast sessions and IPv4 labelled unicast session
to the same peer, if you source traffic from a BVI in a VRF instead of
a loopback (such as SSH/FTP/TFTP/NTP souce interface) you should be
using per-VRF labelling for that VRF at least etc. But those are never
major issues.

IOS-XR is exactly what Cisco needed/needs I think but it's not quite
mature enough yet.


Cheers,
James.