Skip to content

README.md

Latest commit

 

History

History
81 lines (63 loc) · 3.9 KB

README.md

File metadata and controls

81 lines (63 loc) · 3.9 KB
Raw

Lab 04: ORIGIN, the difference between network and redistribute

Use case. Two edge routers of the same AS advertise the same prefix, 10.10.0.0/16, to one ISP. EDGE1 originates it with a network statement (ORIGIN IGP). EDGE2 pulls the same static route in with redistribute static (ORIGIN incomplete, shown as ?). Nothing else differs, so the ISP prefers EDGE1 only because of ORIGIN, and EDGE2 is a silent second choice. That is the everyday way ORIGIN bites: an accidental ? quietly demotes a path.

The lab shows the fix. Setting the origin to igp as the route is redistributed makes the two paths equal, and the ISP's next tie-break decides. The ISP runs bgp bestpath compare-routerid so that tie-break is repeatable, and EDGE2 has the lower router-ID.

   EDGE1 (AS 65000)                 EDGE2 (AS 65000)
   router-id 10.255.0.11            router-id 10.255.0.2
   network 10.10.0.0/16             redistribute static (10.10.0.0/16 -> Null0)
        \                             /
   172.16.1.0/30                 172.16.2.0/30
          \                         /
               ISP (AS 65001)
Router AS Management Links
EDGE1 65000 192.168.99.111 e1/0 172.16.1.2/30 to ISP; static 10.10.0.0/16 to Null0, advertised with network
EDGE2 65000 192.168.99.112 e1/0 172.16.2.2/30 to ISP; static 10.10.0.0/16 to Null0, advertised with redistribute static
ISP 65001 192.168.99.131 e1/0 172.16.1.1/30 to EDGE1; e1/1 172.16.2.1/30 to EDGE2; bgp bestpath compare-routerid

The two edges have no session between them; all sessions are eBGP and there is no IGP.

Run it

On the EVE VM, from /opt/bgp-attributes-executor (stop the shared lab and the dashboard container first, see labs/labtool.sh):

labs/labtool.sh 04_origin up                 # takes about 8 minutes
labs/labtool.sh 04_origin show ISP "show ip bgp 10.10.0.0/16"
labs/labtool.sh 04_origin apply 04_origin    # EDGE2 redistributes with 'set origin igp'
labs/labtool.sh 04_origin rollback 04_origin

What you should see (captured from this lab)

Baseline. Same AS_PATH, same everything, but ORIGIN differs. ORIGIN is step 5, so IGP beats incomplete and EDGE1 wins (the router-ID step is never reached):

ISP# show ip bgp 10.10.0.0/16
BGP Bestpath: compare-routerid
  65000
    172.16.2.2 from 172.16.2.2 (10.255.0.2)
      Origin incomplete, metric 0, localpref 100, valid, external
  65000
    172.16.1.2 from 172.16.1.2 (10.255.0.11)
      Origin IGP, metric 0, localpref 100, valid, external, best

EDGE2's own view shows where the ? comes from: Origin incomplete ... sourced, from the redistribute static line in its BGP configuration.

After apply. EDGE2's redistribution now goes through a route-map with set origin igp. Both paths are Origin IGP, steps 1 to 9 tie, and step 10 (lowest router-ID, forced by compare-routerid) picks EDGE2 (10.255.0.2):

ISP# show ip bgp 10.10.0.0/16
  65000
    172.16.2.2 from 172.16.2.2 (10.255.0.2)
      Origin IGP, metric 0, localpref 100, valid, external, best
  65000
    172.16.1.2 from 172.16.1.2 (10.255.0.11)
      Origin IGP, metric 0, localpref 100, valid, external

Things to notice

  • The preference for EDGE1 was accidental. Nobody configured a preference. It came from how each edge learned the prefix.
  • The change moved traffic to a different edge, even though the only edit was on the edge that was losing.
  • ORIGIN only decides when Weight, Local-Pref and AS_PATH length tie. Here both paths are the same length, which is why it mattered.

Notes from building this lab

  • rollback first removes redistribute static route-map REDIST-STATIC and then re-adds redistribute static, so the configuration returns exactly to the baseline (confirmed in show running-config | section router bgp).
  • apply and rollback wait 40 seconds before they check, because eBGP sends updates at most every 30 seconds.