[j-nsp] VPLS NSR in 9.1

[Harry Reynolds]

Hello, 

I did some functional testing of bgp-signaled vpls nsr with ldp
transport on most JUNOS platforms, to include the MX. My tests expected
no control plane hits to protocol peers and "more or less" hitless
forwarding to pass. AFAIK, we do not guarantee vpls nsr to be 100%
hitless in the data plane, but the majority of my tests show 0 loss, and
I will fail a case with more than 2 rapid pings being lost.

You mention "ldp signaled" below. Are you using ldp or rsvp signaled lsp
for vpls transport? RSVP nsr is not supported, but should not lead to
chronic loss of forwarding either.

What you describe is not normal. Things that often cause vpls/nsr
problems are:

1. Not waiting until the system has reached synchronized state before
performing the nsr. Use show task replication on master and show system
switchover on BU RE to confirm. Also, make sure that all routing
instances are stable and ready with a show route instance detail, making
sure no instances are pending. I think newer versions of JUNOS software
are enforcing a minimum wait between switchover events.

2. Having an ASP present (unless configured not to provide tunnel
services for vpls). This should not be an issue on the mx as it has
built-in tunnel services.  For giggles you might try using
no-tunnel-services under the vpls instance to see if eliminating the vt
interface improves things.

Perhaps you can unicast me your routing-options, interfaces, protocols,
and instances configurations, as well as the output of a show chassis
hardware? 

Also, if you get in this state again also send me output for:

show vpls connections extensive

show route table <name>.l2vpn.0 detail

show route forwarding-table family vpls extensive

Regards and HTHs

Harry








> -----Original Message-----
> From: juniper-nsp-bounces at puck.nether.net 
> [mailto:juniper-nsp-bounces at puck.nether.net] On Behalf Of Marlon Duksa
> Sent: Thursday, July 24, 2008 2:28 PM
> To: juniper-nsp at puck.nether.net
> Subject: [j-nsp] VPLS NSR in 9.1
> 
> Does anyone have experience with Non Stop Routing in 9.1 for 
> VPLS? Release Notes says that VPLS NSR is supported but in 
> our case we clearly see packet drops, as a matter of fact 
> traffic does not even recover after a failure.
> 
> We have LDP signaled and BGP signaled VPLS (single VPLS 
> instance) and after we reset the active RE, the MX960 sends 
> LDP Withdrawal message to the LDP peer and BGP Update msg to 
> the BGP peer. Aren't the states supposed to be synchronized 
> between two REs in NSR case? Peers should have no knowledge 
> of the RE failure.
> 
> In the same setup we run (IPv4) BGP and ISIS and those two 
> protocols are just fine with NSR.
> 
> 
> Thanks,
> Marlon
> _______________________________________________
> juniper-nsp mailing list juniper-nsp at puck.nether.net 
> https://puck.nether.net/mailman/listinfo/juniper-nsp
>