[Nfd-dev] Help with Interests and Nonces

[Junxiao Shi]

Hi John

I believe it’s still #1953, but at a large timescale.

Per ndnmap, SRRU node has 5 peers. With multicast strategy used on /ndn/broadcast namespace, an incoming Interest from one of these peers will be forwarded to 4 other peers.
You are observing 1089 incoming and outgoing Interests. Divide that by 5 (1 incoming + 4 outgoing), it’s roughly 218 incoming Interests.
They are spread over a duration 1071 seconds. Divide that by 218 incoming Interests, and we can see the interval between two incoming Interests is 4.9 seconds on average.

The solution for #1953 is the Dead Nonce List, which is designed to prevent these persistent loops.
A persistent loop can be detected if the Nonce is still in DNL when the Interest loops back; it cannot be detected if the Interest loops back after Nonce lifetime.
The expected lifetime is 6 seconds, but it’s probablistically adjusted, so the actual lifetime can possibly fall to less than 4.9 seconds, and therefore a persistent loop at 4.9 seconds interval cannot be detected.


If this ever happens again, you can try to find out the entire loop by following the source IP field in ndndump trace, and then run ndndump at that peer.
After the entire loop is found, you can adjust NFD log level to DEBUG for Forwarder component (send SIGHUP to nfd process, don’t restart NFD), and the NFD logs can give more information on what’s happening.

Yours, Junxiao

> On Apr 1, 2016, at 8:38 AM, Dehart, John <jdd at wustl.edu> wrote:
> 
> 
> All:
> 
> I am trying to figure out something that is happening on the NDN Testbed and need some help.
> 
> While adding a new node, I misconfigured part of the NLSR keys configuration for the node. In
> tracking that down I noticed that there are some interests that are never satisfied but
> they also never go away. Here is an example from the output of ndndump that I am logging to a file:
> 
> ndnops at octopus-ProLiant-DL380p-Gen8:~$ grep 3880653420 ndndump.srru.log | head -1
> 1459523487.472518 From: 133.1.17.51, To: 210.114.89.49, Tunnel Type: UDP, INTEREST: /ndn/broadcast/KEYS/ndn/th/ac/srru/KEY/ksk-1458752543332/ID-CERT?ndn.Nonce=3880653420
> 
> ndnops at octopus-ProLiant-DL380p-Gen8:~$ grep 3880653420 ndndump.srru.log | tail -1
> 1459524558.534421 From: 210.114.89.49, To: 114.247.165.44, Tunnel Type: UDP, INTEREST: /ndn/broadcast/KEYS/ndn/th/ac/srru/KEY/ksk-1458752543332/ID-CERT?ndn.Nonce=3880653420
> 
> ndnops at octopus-ProLiant-DL380p-Gen8:~$ grep 3880653420 ndndump.srru.log | wc -l
> 1089
> 
> ndnops at octopus-ProLiant-DL380p-Gen8:~$
> 
> The way I read that is that in my log file I have a first occurrence of an interest for this key at time:
> 1459523487.472518
> 
> And a last occurrence at time:
> 1459524558.534421
> 
> Those times are 1071 seconds (over 15 minutes) apart. And there have been 1089 instances of that Nonce
> coming and going from the node I am monitoring.  (And they are continuing. checking now, there are 1471 instances …)
> 
> Does that mean that these interests are bouncing around the world and never timing out? Or perhaps more correctly stated,
> are these interests traveling around the world and by the time they make a cycle back to a node, the PIT entry has timed out
> and that node re-broadcasts it?
> 
> I have shut down nfd and NLSR on the misconfigured node and I have removed the node as a neighbor from all NLSR configuration files
> and restarted nfd and NLSR on all of its neighbors. So, I don’t think any node should be generating new interests for this key.
> 
> Thanks for any insight on this.
> 
> John