4

Will lowering the carrier-delay on an interface configuration interfere with using BFD on the same interface?

If I configure carrier-delay msec 0 on an Ethernet interface the RIB is automatically notified of the physical link down event. In a scenario where the physical link doesn't go down but communications are not traversing the link (3rd party DWDM lamda for example) BFD would detect this.

Are there dangers to using a low carrier delay value with BFD, or is it recommended? Or will one simply superseed the other from a configuration perspective?

I would be using them in conjunction with interaface dampening and ip routing protocol purge interface in global configuration.

  • Carrier-delay is there for a reason, and I personally don't think it is a good idea to disable it as you're asserting in this question. 100ms is as low as I have deployed; however this all is dependent on platforms and requirements – Mike Pennington Apr 3 '14 at 16:29
  • @MikePennington Sorry for not being clear. I'm NOT looking to remove it, I'm looking to lower it from default down to 0 msec or something very low like 2 msec, with BFD being 10ms intervals with a x3 multiplier. – jwbensley Apr 3 '14 at 16:33
  • @MikePennington I believe the default is 10ms on fiber ports. It should be possible to set it to 0 if combined with dampening. I don't see this interfering with BFD. – Daniel Dib Apr 3 '14 at 18:35
  • @DanielDib, I haven't seen IOS defaults lower than 2 seconds... can you give a reference doc? Interface dampening is a reasonable airbag if you want to be aggressive; however, it only damps IP services (which jwbensley seems to be doing). I think the whole notion of IP routing protocols converging in <50ms is kindof out-there... better to use MPLS TE w/ FRR if you need tight convergence windows – Mike Pennington Apr 3 '14 at 18:53
  • @MikePennington I saw it mentioned in Cisco Live via BRKCRS-2031, slide 22. I haven't found a command to show the default so it's not easy to be certain. These are tools that should be used carefully but I had to tweak some values just this day to be able to meet a two second convergence requirement. – Daniel Dib Apr 3 '14 at 21:38
4

It won't interfere per se, but question is - what you're trying to achieve?

First of all, it's hardware dependent - you may be silently ignored using this command on software platforms. Essentially, you should drill down to supported/unsupported notes for your gear to check if it's available and working. Some platforms can tune up/down delays separately.

If you're trying to achieve fast convergence, use one protocol - by approaching the problem with the KISS principle. BFD seems to be best fit in case multiple protocols are used and they need information about the neighbor loss ASAP.

BTW, carrier-delay will only work on interfaces, that will signal up/down situation properly. Connections to shared media (like Ethernet switch) or to media that doesn't properly propagate up/down signal (some SDH/SONET/ATM gear) will cause problems anyway, so you should fall back to something that YOU control.

3

I am using the following interface configuration, although this is platform depending. This is a 7600 with LAN cards, on some ASR9Ks and ME3800s the timers are lower due to the better hardware support so you need to tune this to your own requirements:

conf t

 ip routing protocol purge interface 

 int te6/5
  carrier-delay up 5
  carrier-delay down 0
  dampening
  bfd interval 500 min_rx 500 multiplier 3

I couldn't find any information in any Cisco docs and that doesn't recommend combining these commands. I did find this thread in which a Cisco employee has contributed to saying that one can mix them and it is recommended: http://www.gossamer-threads.com/lists/cisco/nsp/161219

BFD is performing rapid link down detection. When that happens the link will go straight down. When the link comes back up, the carrier-delay will hold it back from being signal as UP to the routing protocol for 5 seconds. Dampening is also enabled such that if this link continues to flap and violates the damping threshold it will be supressed from being signalled to the routing protocol(s).

It is recommend you tune the damping values to your needs: http://www.cisco.com/c/en/us/td/docs/ios/12_0s/feature/guide/s_ipevdp.html#wp1027265

  • When using BFD on 7600s/6500s you also need to consider if you are using DFCs or not and the command process-max-time. – jwbensley Jul 3 '15 at 9:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.