On 21/02/12 01:10, Jonas Björklund wrote:
> policy-statement my-export-routes {
> term t10 {
> from {
> protocol bgp;
> route-filter 1.2.3.0/21 prefix-length-range /21-/24;
> route-filter 4.5.6.0/20 prefix-length-range /20-/24;
> }
> then accept;
> }
> term t100 {
> then reject;
> }
> }
>
> Some of my network begin to be announced. But not all of them.
>
> The one which was learned by OSPF *AND* BGP was not able to pass the
> policy. Only those which was learned from BGP.
>
> I want all networks learned from BGP (even those from OSPF) pass the
> policy.Then remove "from protocol bgp". The router is doing exactly what you asked, you're simply missing two things: 1. Only active routes are considered in filters (for values of only that suffice for this) 2. IGP (OSPF, IS-IS) routes are preferred over BGP in JunOS, unlike IOS The way I do self export is a little more complex, but prevents route spam: 1. For my large blocks (really anything in more then one OSPF area) I configure an aggregate route covering them, for example: set routing-options aggregate route 10.23.64.0/18 Unless you specifically add a policy these will never be advertised outside the router. 2. For all blocks I'm advertising I create a prefix-list: set policy-options prefix-list self 10.23.64.0/18 set policy-options prefix-list self 10.2.3.4/24 3. Inside my outbound chain I simply do: set policy-options policy-statement self-out from prefix-list self set policy-options policy-statement self-out then accept Which means the prefixes I expect (and *only* those exact prefixes) get advertised. (This only makes sense in smaller networks with few edge BGP speakers, larger networks should use their existing confederation / reflector boundries as points to inject that regions self into BGP) -- Julien Goodwin Studio442 "Blue Sky Solutioneering"
signature.asc
Description: OpenPGP digital signature
_______________________________________________ juniper-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/juniper-nsp
