Subsequent to the last routing wg seminar I have just been looking at out parameters for bgp dampening. To summarise What we (1849) currently have : 1- don't dampen any routes we use for special purposes 2- don't dampen any root nameserver routes 3- Heavy dampening on /24 and greater (>24 filtered anyway) 4- not so heavy dampening on /22 and /23 5- All else using default for standard protection. I am seriously think of adding 192.* into 3 above. People should be renumbering !!! Parameters to go with above are : ! Savage dampening of all networks which have a mask of ! /24 and above. These are supressed into a datastructure ! with a half life of 45 minutes, only re-use when reaches 250 ! Max outage of 255 minutes. no route-map expo-flap-dampen permit 10 route-map expo-flap-dampen permit 10 match ip address 181 set dampening 45 250 2000 255 ! dampen /23 /22 ! half life is now 25 minutes and reuse at 650 no route-map expo-flap-dampen permit 20 route-map expo-flap-dampen permit 20 match ip address 182 set dampening 25 650 2000 60 We should redress these figures in light of discussions last week. Personally I think that the concept is about right as we have it, we just need to agree the figures/times. It means that all major aggregated space is not really affected any more than it was before. In fact we could improve it as per below. Should we try 3000 as the suppress at figure so that minimum of 4 flaps is required? The max-suprress time needs to be changed to 60 max as we said. Hence new figures would be :::: ! /24 and above. set dampening 30 750 3000 60 ! dampen /23 /22 set dampening 20 1000 3000 45 all else default which is set dampening 15 750 2000 30 but change the minimium count to 3000 and reuse at to 1000 --Tony
Tony Barber writes:
I am seriously think of adding 192.* into 3 above. People should be renumbering !!!
Should we or should we just wait for IPv6 and migrate to translating our addresses to that when communicating externally? Pete
participants (2)
-
Petri Helenius -
Tony Barber