I would look for an MTU issue.
With an MPLS tunnel the packets must be slightly larger than the payload -- 4 bytes for the MPLS label, for each encapsulation (ie if you're going to put MPLS tunnels into an MPLS tunnel, the outside tunnel will be 8 bytes larger than the payload). Here's a discussion of this.
Hopefully you're not filtering ICMP messages -- if you are you'll be dropping the "use a smaller MTU" packets the network will be sending to your hosts if indeed you are sending too-large packets over the wire. This discusses issues related to MTU...
One thing you can try on your end without involving your ISP is you can shrink your MTU on your border device -- that way it negotiates a proper MTU with your inside network for traffic going over the MPLS link.
If I understand the question correctly you have an IGP (or local) route to the network, and you annouce it over BGP. When the route vanishes in the IGP (or local) you want BGP to pull the route.
If that is the case you are doing stuff wrong(TM), and Quagga will not let you easily do this. From the manual for the network command:
BGP: network A.B.C.D/M
This command adds the announcement network.
router bgp 1
network 10.0.0.0/8
This configuration example says that network 10.0.0.0/8 will be announced
to all neighbors. Some vendors' routers don't advertise routes if they
aren't present in their IGP routing tables; bgp doesn't care about IGP
routes when announcing its routes.
This is due to the increased flapping you can easily get if you export IGP information in BGP. We have enough of a route churn on the internet allready, and it's considered bad practice to redistribute routing information from IGP into BGP. BGP is not an IGP, and don't abuse it as one ;)
Also I can't really see any good cases for pulling the route from the Internet (it will cause flapping and you risk getting dampened for hours or days), unless you are ending up in a split-AS situation if this specific route is gone and want to protect yourself from the weird routing issues this can cause. (In this case, you should consider if you want the router to stay online at all. Split-AS situations are nasty!)
The correct solution(TM) is to leave the route up and as stable as possible, regardless of what your IGP is doing. If you lose the connection to the network just drop the traffic locally. Make sure you don't loop it back to your transit provider if the IGP route to the network is down.
The basic rule is "never change your BGP announcements unless it's something the whole Internet has to know about". That your IGP is flapping is not something the rest of the Internet cares about.
Edit:
From what I understand your network looks like this:
Provider (AS 5555) --------------------- Provider (AS 5555)
(12.12.12.12) |
| eBGP |eBGP
| |
Router1---------15.15.15.0/24---------------Router2
172.16.14.1 172.16.14.2
| iBGP |
--------------------------------------------
And your problem is that if you take down the interface on Router1 towards 15.15.15.0/24 you want it to stop announcing the network so you shift the data over to 172.16.14.2. This type of automatic changes to your peering policy is not something you usually do, and is as far as I know not something supported by Quagga. Instead you are expected to reroute the data over the IGP and keep your peerings static. If you were to do changes to your peerings you would change the MED (MULTI_EXIT_DISC) to steer the traffic to the right router.
Note that if taking down 15.15.15.0/24 splits your AS you have additional failure modes, none of them good.
Best Answer
BGP Site-of-Origin (SoO) is used to prevent looping in the following set-up.
1) Customer sites connected to the MPLS/VPN provider use a single AS (thus needing as-override configured). 2) A customer site (usually the hub or other important location) has multiple connections to the provider.
The SoO filtering is automatic once SoO is configured. No BGP filtering is needed.