07-24-2003 11:40 AM - edited 03-02-2019 09:06 AM
I have two Internet-facing routers which are connected to each other via dual paths. An iBGP peering needs to be maintained between the routers. If the BGP peering is created with loopback addresses, is it permissible for the loopback addresses to come from RFC1918 (private) address space, assuming the loopbacks are only used for this purpose? The IGP (OSPF) running between the routers will ensure that the private address of each of the loopbacks is reachable over one of the two paths. It may be possible to take the loopback addresses from the subnet allocated to use by the ISP, but we cannot really afford to subnet this address space as it is only as /26.
07-24-2003 11:55 AM
Yes you can do that. Only make sure that you dont advertise this network via BGP.
You will need to use the command,
neighbor
07-25-2003 03:09 AM
Yes, this is fine.... You may want to filter the private address space out of BGP's tables at your edges (eBGP connections). You should be doing this anyway, actually--filtering RFC1918 stuff both inbound and outbound.
Dirty little secret: Some exchange points use RFC1918 space to number the links between the peers in the exchange point to keep people from being able to see the structure of the exchange point, and to keep from using a lot of unneeded address space. At least one I looked at a couple of years ago did, I don't know if this is still common practice or not.
Russ.W
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide