NetworkTalk & BGP

A.8.c) Common errors and bug to avoid

  • The two common errors made are:
    1. to point remote loopback route at IP address rather than a specific link,
    2. to forget to specify the physical interface in the static route.
  • If you enter the following routes;
    ip route 30.30.30.127 255.255.255.255 30.30.31.21
    ip route 30.30.30.127 255.255.255.255 30.30.31.25
    ip route 30.30.30.127 255.255.255.255 30.30.31.29

and when one of the interfaces is down, CPU process can high-up continuously to 99%, which of course is very dangerous.
If the interface on which the preferred static route to this loopback is cut, in this case if we shut pos1/1/2, the preferred static route is still not removed from table.
Recollection process is using 99% of CPU. In other words, the router is still trying to reach the next hop of the physical interface cut through another route.
This route is generally learned from another Ebgp session with another ISP. As a result, since the interface is down, the packets are actually sent to a black hole, because it keeps trying to send through pos1/1/2.

Before cut :
RT-B#sh ip route 30.30.30.127
Routing entry for 30.30.30.127/32
Known via “static”, distance 1, metric 0
Routing Descriptor Blocks:
30.30.31.29
Route metric is 0, traffic share count is 1
* 30.30.31.25
Route metric is 0, traffic share count is 1
30.30.31.21
Route metric is 0, traffic share count is 1

After cut:
RT-B#sh ip route 30.30.30.127
Routing entry for 30.30.30.127/32
Known via “static”, distance 1, metric 0
Routing Descriptor Blocks:
30.30.31.29
Route metric is 0, traffic share count is 1
* 30.30.31.25
Route metric is 0, traffic share count is 1
30.30.31.21
Route metric is 0, traffic share count is 1

As a result:
RT-B#sh processes cpu | e 0.00%
CPU utilization for five seconds: 100%/0%; one minute: 53%; five minutes: 17%
PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process
7 17467240 12822226 16960 5.91% 0.86% 0.64% 0 Check heaps
10 62638380 670931689 93 0.07% 0.15% 0.21% 0 ARP Input
66 43237700 105520440 409 4.55% 10.36% 3.54% 0 BGP Router
69 5854600 69807223 83 0.95% 0.23% 0.05% 0 BGP I/O
265 6829624 29879692 228 0.07% 9.92% 2.94% 0 IP RIB Update
329 175304 1082 162018 87.92% 21.50% 5.32% 0 Collection proce
471 152 428 355 0.23% 0.08% 0.02% 1 Virtual Exec

  • The solution is, in order to remove the static preferred route from the routing table; to precise that next-hop 30.30.31.25 is related to its own physical interface.If the interface is down and so the next-hop is removed from the routing table, the router will not try to reach anymore this next-hop through another way.
    ip route 30.30.30.127 255.255.255.255 POS1/1/2 30.30.31.21
    ip route 30.30.30.127 255.255.255.255 POS1/1/2 30.30.31.25

    ip route 30.30.30.127 255.255.255.255 POS1/1/2 30.30.31.29

Before cut:
RT-B#sh ip route 30.30.30.127
Routing entry for 30.30.30.127/32
Known via “static”, distance 1, metric 0
Routing Descriptor Blocks:
30.30.31.29, via POS1/1/3
Route metric is 0, traffic share count is 1
30.30.31.25, via POS1/1/2
Route metric is 0, traffic share count is 1
* 30.30.31.21, via POS1/1/1
Route metric is 0, traffic share count is 1

After cut:
RT-B#sh ip route 30.30.30.127
Routing entry for 30.30.30.127/32
Known via “static”, distance 1, metric 0
Routing Descriptor Blocks:
30.30.30.29, via POS1/1/3
Route metric is 0, traffic share count is 1
* 30.30.30.25, via POS1/1/2
Route metric is 0, traffic share count is 1

top

Come back to Tutorial Index”