I have a full-mesh LDP LSPs between my MX-104 routers, however, between two specific routers and on the same loopbacks I configured RSVP LSP to be used as the transport for only one l2circuit and no more. The problem is, when the RSVP gets signaled, it gets installed in the inet.3 and gets preferred over any other LDP LSP. So all the traffic destined to RSVP tail-end will prefer the RSVP over the LDP. I have increased the preference of the RSVP, and it has been taken out of the inet.3, so the l2circuit didn't prefer the RSVP path anymore! Do anyone has a working configuration for this? or should I configured another loopback address on every pair of routers for the RSVP signalling? -- mk
tried this Timothy, and the RSVP didn't appear in the inet.3. Failed to work! ----- Original Message ----- From: "Timothy Creswick" <Timothy.Creswick@vorboss.com> To: "Mohamed Kamal" <mkamal@noor.net>, nanog@nanog.org Sent: Friday, May 27, 2016 3:30:08 PM Subject: RE: Preferring RSVP for only one l2circuit.
I have increased the preference of the RSVP, and it has been taken out of the inet.3, so the l2circuit didn't prefer the RSVP path anymore!
Just add "no-install-to-address" to the LSP.
Hi, ( it might be a bit much to look for that here, but meh its Friday ) Goals Usual Multi-point L2 services, with L3 path(s) to the Internet -and/or- inter-site L3 routing (VRF per customer). This is a simple project involving upgrading a L2 MAN into and more flexible and isolated (MACs wise) MPLS infrastructure. Road blocks We're wasting way too much time with sales rep/engineer, for then, ending up finding some irrelevant limitation after having spent ~20h of man power looking at their offering. Question ( Obviously interoperability is just ridiculous when talking MPLS... ) Thus I'm looking for reference designs from different vendor viewpoint. Simple enough, but I would like to put the "make/model of the actual device capable" of doing the function at the P, PE and CE level of their design. Too many times we ended up finding boxes that look the part but are limited in such a way to accomplish only the task that fit the vendor own MPLS paradigm. ( TLDR: Make us spend more $$$ than needed =D long gone is the swiss-knife concept ) Offlist answers are fine.
On 27/May/16 20:36, Alain Hebert wrote:
( Obviously interoperability is just ridiculous when talking MPLS... )
Why would you think so?
Too many times we ended up finding boxes that look the part but are limited in such a way to accomplish only the task that fit the vendor own MPLS paradigm. ( TLDR: Make us spend more $$$ than needed =D long gone is the swiss-knife concept )
What platforms have you worked on before that have caused you grief? Mark.
On 27/May/16 15:46, Mohamed Kamal wrote:
tried this Timothy, and the RSVP didn't appear in the inet.3. Failed to work!
Best option is to create a separate Loopback interface on each router, and use that to signal the RSVP tunnels. It's the only way to keep your LDP and RSVP tunnels from mingling with each other for EoMPLS circuits you want to nail on a given path. Mark.
participants (4)
-
Alain Hebert -
Mark Tinka -
Mohamed Kamal -
Timothy Creswick