Routing

Hi, 

After digging into this forum for answers, none worked on our scenario.

Currently we have a MX10003 PE  and  lots of non-JPNR MPLS capable switchs. As this MPLS switchs are new, there a lot of missing features, like FRR and other ways to do a fast LSP convergence.

The current convergency time is about 1~2 seconds. But there is a problem. We terminate BGP customers from those Switches to our PE via L2CIRCUIT. But I noticed that when there is a fiber cut and the l2circuit needs to be restabilished to the active LSP, the BGP sessions flaps, because the L2circuit with stitched L3 LT-Interfaces-tunels get down and Junipers default behavour is to disable the L3 interfaces IP and BGP session down as soon as it detects interface or l2tunnel down.  I tried with VPLS with IRB with all connectivity knobs, even with that, the IRB IP address get down and BGP sessions get down as soon as VPLS connection between PE and the P Switch goes down.

Is there any way to force BGP and IRB or LT ip address to keep up for a little time before gets removed from routing table and killing BGP session?  We have a lot of redundancy but the customers complains about the BGP session flap that is painfull on some customer routers..

Regards.

Hi Softdados,

Greetings, If I am understanding correctly, please consider these options.

For interfaces:

When a hold-down timer is configured and the interface goes from up to down, the down hold-time timer is triggered. Every interface transition that occurs during the hold-time is ignored. When the timer expires and the interface state is still down, then the router begins to advertise the interface as being down.

https://www.juniper.net/documentation/en_US/junos/topics/reference/configuration-statement/hold-time-edit-interfaces.html

For BGP: In good theory, the session should go down until the dead interval is expired, have you considered increasing the hold-time ?

https://www.juniper.net/documentation/en_US/junos/topics/reference/configuration-statement/hold-time-edit-protocols-bgp.html

Regards,
Lil Dexx JNCIE-ENT#863

If this solves your problem, please mark this post as "Accepted Solution" so we can help others too \:)/

Hi,

It may be a little tricky. You can trying following:

1. Your irb goes down when l2circuit/vpls session goes down. The reason irb goes down is because there's no active layer 2 interfaces any more. You may try to add some dummy interfaces in your bridge domain where the irb belongs to

2. For l2circuit, if you configure a backup neighbor to somewhere else, eg another PE. Can it keep the BGP session up for a bit longer timer 

Thanks,

VPLS + Dummy LT-Interface worked.

Regards!

Great to know. If you have VPLS with multiple PEs, as long as there's one remote PE is up, your irb should stay up till BGP hold timer expires