Question for all you BGP/OSPF Gurus

Discussion in 'Cisco' started by John, Apr 11, 2005.

  1. John

    John Guest

    Admittedly, I am not a routing guru. However, I have to get some redundancy
    built in to our existing Internet infrastructure. The proposed setup is
    pretty simple; 2 sites w/ different ISPs and a total of 4 routers, switches,
    and firewalls. Please visit www.geocities.com/ji_fisher/bgp.jpg to see the
    logical diagram (for anonymity I have used someone else's external IP
    blocks). Anyway, if you look at the diagram you will see that we have
    routers A and C and routers B and D connected over some high speed metro
    Ethernet links.

    We are peering with Time Warner in Austin, and SBC in Taylor. Within our
    network we want all four routers to peer with each other over iBGP. For our
    IGP, we will use OSPF. My understanding is the BGP will not advertise routes
    learned via another iBGP neighbor unless that route is known by an IGP (in
    this case OSPF).

    What we want is simple, if Time Warner dies, SBC needs to be able to handle
    routes to and from the 67.128.17.0 /24 block, and if SBC dies, Time Warner
    needs to be able to handle routers to and from the 67.128.16.0 /24 block.
    Also, any router and switch failures should be relatively transparent to our
    Internal and External users.

    My question is this; should all four routers be a member of Area 0, or
    should we have disparate areas between sites? The reason I ask is because
    when we brought this topology up last night, we established BGP peering just
    fine. The problem was that as the BGP routes were being propagated between
    iBGP peers, they would climb up to 156,000, and then suddenly drop to
    roughly 4000. They would then slowly creep back up, only to plummet again.
    This, of course, made for a very unstable network. We were forced to remove
    the secondary routers for the time being to stabilize the network.

    Sorry to ramble, but if anyone has any suggestions on how to configure this
    in a highly available and stable manner PLEASE make a suggestion. Thanks in
    advance for the help.

    John
     
    John, Apr 11, 2005
    #1
    1. Advertisements

  2. You can use the "no sync" option in BGP to turn off that check. You
    should *not* redistribute the routes learned via EBGP into OSPF.
    OSPF should handle detecting failures within your infrastructure.
    Do you really need to get full routes from both ISPs? I think you would
    do better to have each of them send you just their customer routes and a
    default route.

    I don't think you need to worry about multiple OSPF areas with your
    simple configuration.
     
    Barry Margolin, Apr 11, 2005
    #2
    1. Advertisements

  3. seeing this i suspect you have not enough memory in your bgp routers,
    and they simply drop bgp session after receiving portion of prefixes,
    then restarts again.
    what routers are you using for bgp and how much memory do they have?
    if there is no reason to have the full bgp table, you should consider
    receiving only router for your provider's customers and default route,
    as suggested by Barry Margolin, otherwise you probably have to upgrade
    your routers or at leas optimize memory usage.
     
    =?ISO-8859-2?Q?Krzysztof_Ciep=B3ucha?=, Apr 12, 2005
    #3
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.