BGP with the ISP's demonstrating some wired behavior.

[nikhil.kulkarni]

Folks,

I seen a network with some wired behavior when BGP is enabled with another ISP.

TIll now that network was having BGP enabled with a single Provider(Provider A) who had 2 links coming in and at the back end it was a blended solution, one of those COLO providers kind of setup.

Recently they enabled BGP with a single ISP and seen that some kind of packets drops were being caused.

On one network; e.g. 1.1.1.0 /24 we were trying to ping different hosts from outside. Say 1.1.1.1 and 1.1.1.3 would respond to ping and 1.1.1.5 and 1.1.1.6 would not. In just 30 secs 1.1.1.5 and 1.1.1.6 would respond to ping and 1.1.1.1 and 1.1.1.3 would start giving ping drops. Again 1.1.1.1 and 1.1.1.3 would respond to ping and 1.1.1.5 and 1.1.1.6 would not.

Any suggestion on what to look for? In this case we are not advertising the network 1.1.1.0/24 to the Provider A(i.e. the new provider) for now. Still this network is giving packet drops.

Regards,

Nikhil Kulkarni.

[Peter Paluch]

Hi Nikhil,

My first impression is that this looks like some sort of routing table oscillation, i.e. a periodic change in routing table contents induced by some kind of misconfiguration.

My suggestion is to try repeated traceroute commands to the addresses 1.1.1.1 and 1.1.1.5 from outside in parallel, and observing if their output changes over time - and perhaps if there seems to be a correlation between the particular sequence of hops and packet losses. If there truly is a routing table oscillation they you may very well see the path shown in the traceroute changing every now and then.

Best regards,
Peter