For reasons unknown my internal routing protocol 'lost' it's route to our provider independant address space this morning. The symptoms were a complete loss of access to the Internet. Logging a fault with the relavent ISP returned a 'no problem here' response.
We have redundant Internet feeds so I was confounded that both of them should fail at the same time. Further investigation determined that access to the Internet worked fine using an address from the /30 each ISP provided for connectivity. So it was just our address space.
Whilst waiting for the ISPs to get back to me (I had made it clear it was a routing issue not a service issue) I double checked and lo and behold my gateway routers no longer had a route for our address block. I'm sure I had one before but not having one now meant BGP had nothing to advertise.
BGP is different from other routing protocols in that it will only advertise to other parties (people with a different ASNUM) that which is already in your routing table.
I'm sure I had a route before but thinking about it, I believe the route came from BGP in the first place. It's like a circular argument, it works as longs as you don't stop. So some network change stopped BGP advertising to itself our block of address space and thus it no longer existed in the routing table of my gateway routers, and thus BGP could no longer advertise our block to the upstream providers.
Talk about a chain of events.
So I now have a static route to Null0 for our block to ensure BGP always has something to advertise. I should make it a backup route (high admin distance) so my IGP can insert something if it needs to.
Oh the joys.
No comments:
Post a Comment