[j-nsp] Bypass LSP functionality question

[Harry Reynolds]

Hey  Stefan, no, in normal conditions the weight/metric of the bypass is poisoned/set high o traffic will always take the primary.

At time of repair the loss of primary makes for "any port in a storm" and now traffic flows on the bypass. Possible the metric is updated/lowered, but given there is no primary not sure it would matter.

Regards




-----Original Message-----
From: Stefan Fouant [mailto:sfouant at shortestpathfirst.net] 
Sent: Thursday, July 07, 2011 9:48 AM
To: Harry Reynolds
Cc: David Ball; Phil Bedard; Juniper-Nsp
Subject: Re: [j-nsp] Bypass LSP functionality question

On 7/7/2011 12:26 PM, Harry Reynolds wrote:
> Do you have a per-packet lb policy applied to the forwarding table at the point of local repair?
>
> AFAIK, any fast reroute/bypass needs this for optimal switchover as it allows the detour/bypass to be preinstalled in the pfe waiting for its day in the sun.
>
> As I recollect, the ingress will continue to use the bypass until it can signal a new path (MBB). You may want to set the ingress lsp to adaptive to better chances of the new path being established so the lsp can get of the bypass.

Harry is right.  This is covered in the 'Junos MPLS and VPNs' course. 
Only the Active LSP's next-hop is installed in the PFE by default. 
During the time that it takes for the RE to install the detour or bypass next-hop in the forwarding table, it is not unusual to see a small amount of dropped traffic.

Harry, I've always wondered about this however, since the recommended solution is to apply a forwarding-table load-balance policy, does this mean that traffic will *always* use both the Primary path as well as any Detours or Bypasses under normal working conditions?

Stefan Fouant
JNCIE-ER #70, JNCIE-M #513, JNCI
Technical Trainer, Juniper Networks
http://www.shortestpathfirst.net
http://www.twitter.com/sfouant