2

I have L2L tunnels, some on marginal circuits, that frequently go down with a message like:

%ASA-3-713123: Group = 50.x.x.x, IP = 50.x.x.x, IKE lost contact with remote peer, deleting connection (keepalive type: DPD)

These are statically defined tunnels, and usually they come back within a few seconds. These have no secondary peers.

I can find lots of information on how to set the threshold and retry for DPD keep-alives, but very little guidance on when to use and how to determine the best settings.

I have tested changing them, and can make the change, and nothing bad happens, but I am unsure if anything good is happening (the up/down bounce while frequent overall, is infrequent on any given tunnels). Before I start changing a bunch in production I would appreciate a sanity check:

  • It seems logical that on marginal circuits (and some of these are, both low quality and occasionally too busy), and notably with no secondary peers, that DPD should go slower (if at all), so I am thinking of changing the retry from 2 seconds (6 total) to 10 (30 total).

  • There does not seem any benefit in extending the threshold, however. Is there any reason to adjust the threshold if you adjust the retry time?

  • While some documents say you must set these symmetrically, I can see no harm in different values from watching it in debug. While I do not intend to be different, it takes time to deploy to each router -- is there any issue if for a period of time these are different (but not turned off) at each end?

  • It appears each tunnel (at each end) needs to be changed individually. I tried changing DefaultL2LGroup (recognizing we have individual static tunnel groups) and as expected it has no impact on them. Am I correct there is no way to change the defaults, and you must change each tunnel (and each end)?

Environment: Generally these are ASA5505's, a few 5525's and 5515's, with versions ranging from 8.2(5)59 to 9.5(2)6. The circuits on the spokes are generally low speed (cable modems, DSL, etc.) and usually not busy, but occasionally too busy.

There are about 85 tunnels that need to be changed, so even if this is relatively safe (and appears to be), I'd rather only do this once. Does it sound like I am on the right track?

  • one usage of DPD keeplive messages is maintaining the NAT translation entry if your firewall is behind nating . – Mr.lock Jul 10 '16 at 11:03
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you could provide and accept your own answer. – Ron Maupin Aug 14 '17 at 21:28
1

Okay here goes:

It seems logical that on marginal circuits (and some of these are, both low quality and occasionally too busy), and notably with no secondary peers, that DPD should go slower (if at all), so I am thinking of changing the retry from 2 seconds (6 total) to 10 (30 total).

I think you are on the right track with regards to your settings - I generally stick with 10s for retry timer - if there are no secondary peers, then it doesn't really matter how fast a failure is detected.

There does not seem any benefit in extending the threshold, however. Is there any reason to adjust the threshold if you adjust the retry time?

No good ones that I can think of. Perhaps if there isn't a lot of traffic on the link to keep it alive, or it's a ridiculously expensive Satellite link, then you might wind out the threshold a bit.

While some documents say you must set these symmetrically, I can see no harm in different values from watching it in debug. While I do not intend to be different, it takes time to deploy to each router -- is there any issue if for a period of time these are different (but not turned off) at each end?

DPD only kicks in when your tunnels are completely idle, so as long as you were using bi-directional mode (the default), then the R-U-THERE/response messages from each node should also count as traffic on the link and reset the threshold on the far end each time (even if the counters are much longer on one end).

It appears each tunnel (at each end) needs to be changed individually. I tried changing DefaultL2LGroup (recognizing we have individual static tunnel groups) and as expected it has no impact on them. Am I correct there is no way to change the defaults, and you must change each tunnel (and each end)?

It's been a long time since I've touched an ASA, but I'm pretty sure that each instance of an L2L tunnel has independent settings rather than inheriting them from any of the default groups, so no unfortunately.

Good luck!

| improve this answer | |
  • Thank you Benjamin, I appreciate the sanity check. Indeed, DPD packets do count as traffic, as I found I needed to be careful which end of the tunnel I pinged to bring it back up in order to reliably see the traffic in debug mode (the other end, starting a bit later, kept seeing traffic from the initiating end's DPD's and not sending its own). Too bad there's no overriding default, I guess I'm going to sit through all that typing on each end. Almost worth writing a script for. – Linwood Jul 10 '16 at 14:55

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.