Wednesday, June 10, 2015

Routing Problem When Running BGP


Real life example of providers that's running own BGP. And did not set it up correct or didn't see the problem. This example will show a SIP provider that we get sip trunks from and what happens when they not got BGP routing right. Our customers started complaining that phone calls had jitter and some times was disconnected. Just before this problem we upgraded all core routers to latest firmware and for sure I started to get suspicious that this was the problem. But after some troubleshooting I went over to be suspicious for the SIP provider. I started to do some tests for one of the SIP server at provider side (we hade up to 300-350ms). And I could see we had very long respons time from our side. So next step was to talk to SIP provider and ask them to do tests. So they did tests and came up that they had no problem (13-25ms). Now you really start thinking it's your problem. But after one or two hours. I call SIP provider again and start to do some detective job talking to them. After littel talking they saying that they just installed I new Upstream and the are running ther own BGP and use the same Upstream provider. Now I start to understand the problem, but how will I prove it for them (SIP Provider). I call Upstream provider and explain the problem and ask if they could do tests and after two days they call me back and say tey find problem and had prove for it.

THIS IS THE ANSWER FROM UPSTREAM PROVIDER:

Nothing in troubleshooting indicating the failure of the customer (US). SIP provider has asymmetric routing for the moment when they only send traffic to their drop-off in City X but the Shortest Path to the route reflector of Upstream provider located in Stockholm and traffic routed then via Kista. As this debugging can not be driven by the (US) against Upstream provider and SIP provider must check its routing. Or make an Error reporting themselves to Upstream provider.
 
Tools used for troubleshooting:
  1. traceroute or tracert
Now I only need to get SIP provider to understand they doing it in wrong way