On 13-03-24 2:08 PM, Payam Tarverdyan Chychi wrote:
Hey,

I'm not sure what the actual exact request from the user was since i don't really participate much in the AMS-IX anymore ...

maybe the attack was destined towards their actual nei ip on the exchange (initially i assumed /22 was their network, sounds like maybe they meant the /22 that is shared for connectivity by the exchange members) and they wanted to drop traffic destined to them? You have to remember that traffic routed via a router is not the same as traffic destined for a router and if I recall the email paste by Tobias, it sounded like the use on AMS-IX was getting attacked on their bgp ip and asked everyone to either stop carrying traffic from their networks to the x.x.x.x/22 or setup a filtering so only BGP protocol is allowed and everything else is dropped (someone correct me here if im wrong)


--> Sorry, this below was in response to your previous statement of being multi-homed (if you null route on not directly connected routers). The directly connected router will not drop the bgp sesions as its 'directly connected' Yes, If you nullroute /22 that belongs to the peering session you are going to kill your nei adj with the exchange. Since only valid traffic is identified to be BGP, i would simply setup an ACL and discard anything being sent to the x.x.x.x/22 exchange subnet except BGP packets, applied on the "output" (which carry the routing table/updates...) and perhaps add your own network mgmt interface ips for ICMP ping to help for troubleshooting down the line.

On a side note ...the X0Changes really need to some up with process and procedures to help people deal with issues like this, leaving it open in this day and age is ... (stupid?) Not all network admins are the same nor share the same knowledge and leaving it up to the network admins to figure things out sometimes just means bad news for everyone.

Simple link to look at for constructing an ACL on a Juniper (im sure google has more!) heh
http://kb.juniper.net/InfoCenter/index?page=content&id=KB16685

cheers,
-Payam

On 13-03-24 1:24 PM, Zehef Poto wrote:
Thank you Payam. I think I got what you mean.

In this particular case however, the X/22 route is not a customer or anything. It is the IXP's peering LAN !

So... It means that the person requested all the IXP's members to null-route the whole peering LAN ? How can you possibly ask for this ?

I peer with several members within this LAN. If I null-route the X/22 LAN, we agree that my peering sessions will go down, right ?

Thanks again,

2013/3/24 Payam Chychi <pchy...@gmail.com <mailto:pchy...@gmail.com>>

    Carry a route is the same as accepting a route and having it
    become active, allowing traffic to traverse your network to the
    destination. In this case the user is asking you to drop the
    route (attack traffic) at your edge if possible and not to carry
    it through your network and deliver it to the end destination(his
    network) because its probably saturating or causing him
    performance issues.

    Normally networks well have a global community string that they
    can tag a route with and it will send it to null0, dropping that
    traffic at the edge v.s the user withdrawing its -/24 route from
    the advertise table. You can also go on the peering router and
    set the next hop route for the attacked destination ip to null0
    (discard) and only traffic traversing that one router well drop
    the traffic (global community well handle this if you  have a
    multi homed network)

    Local nullroute example:
    "Set routing-options static route x.x.x.x/32 discard" ...
    Something like this

    All your doing is dropping traffic for x.x.x.x/x at your edge,
    most cases its a /32 nullroute.

    Google is your friend :)
    Cheers,
-- Payam Chychi
    Network Engineer / Security Specialist

    On Sunday, 24 March, 2013 at 6:47 AM, Zehef Poto wrote:

    Hey guys,

    Thank you all for the very valuable input. Actually yes, Tobias
    is right,
    I'm having this question because of the (quoted by Tobias)
    e-mail we got
    yesterday across several IXPs.

    I just don't understand what is to "carry a route in my
    backbone". Am I not
    supposed to know all of (or most of) the Internet routes, since
    I work with
    tier-1 upstream providers ? As a consequence, it means I'm
    carrying all
    these routes right ?

    A "show route X/22" tells that it was advertised by an eBGP peer
    on one of
    my edge routers, and the three other ones learnt this same route
    via OSPF.

    This is where I'm completely confused. What am I supposed to do
    to "carry"
    a route or not ?

    Thanks again,

    2013/3/24 Tobias Heister <li...@tobias-heister.de
    <mailto:li...@tobias-heister.de>>

    Hi All,

    Am 24.03.2013 00 <tel:24.03.2013%2000>:26, schrieb Jeff Wheeler:
    Whoever that person is that said something about "use
    next-hop-self"
    in this context, either you misunderstood them, or you shouldn't
    listen to them anymore. That has nothing to do with looking to
    see if
    your router knows about a route.

    This sounds like the OP wants to help the cloudfare guys who
    send the
    following mail to DECIX/AMSIX (and probably other IX) yesterday.

    We're currently seeing a very large attack directed to our IP
    on AMS-IX
    (X).

    We request that all peers:

    1) Don't carry this route (X/22) in your backbone. (you can set
    next-hop-self, etc). It'll save other security concerns and
    possible free
    transit you're giving away to others.
    2) Filter any traffic within to the AMS-IX exchange fabric (again,
    X/22), except for your point to [multi]point BGP communications.

    --
    Kind Regards
    Tobias Heister
    _______________________________________________
    juniper-nsp mailing list juniper-nsp@puck.nether.net
    <mailto:juniper-nsp@puck.nether.net>
    https://puck.nether.net/mailman/listinfo/juniper-nsp




_______________________________________________
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

Reply via email to