Duplicate nonce generation on retransmitted interests
I am running NDN on top of NFD 0.6.2. On the consumer end, I have the same interface drop packets/connection with NACK duplicate messages. On further investigation, I figured that some of the interests with interrupted connections generate "same" nonce for same interest packets (probably retransmitted) and whenever that happens, the connection over NFD completely drops.
I came across this conversation thread on issues/3888 on redmine and found that this was a design flaw that was to be corrected in v0.6 but I still see this in 0.6.2.
I have attached the ndndump trace file where you can trace the duplicate interests and how the NFD connection drops after it.
Could someone help me with this?
#1 Updated by Davide Pesavento 10 days ago
- Tags deleted (
- Subject changed from Duplicate nonce generation on retrasnmitted interests to Duplicate nonce generation on retransmitted interests
- Category changed from Build to Forwarding
- Priority changed from Urgent to Normal
- Start date deleted (
#2 Updated by Junxiao Shi 10 days ago
Additional information is required to diagnose this issue:
Capture packet traces with tcpdump using filter "udp port 6363", not
Provide NFD logs of the node where Nack packets originate. Set log levels as follows:
default_level INFO Forwarder TRACE
Explain what makes you believe "the connection over NFD completely drops", and provide relevant logs to prove your claim.
See also: HOWTO write a bug report effectively.
Please provide requested information within 15 days.
#3 Updated by Ishita Dasgupta 4 days ago
- File tcpdump6 tcpdump6 added
- File ndndump6 ndndump6 added
- File nfd.log nfd.log added
- File ndnperf.png ndnperf.png added
Here is the tcpdump log as you asked. I couldn't make sense of it as it doesn't say anything about the nack-ed packets (or I may have missed). I have attached the ndndump log for the same session if you need to compare/ for more information.
I am providing the default NFD log. I couldn't get the log with the FORWARDING trace on. (These nodes are containers that run nfd 0.6.1 and somehow "nfd-stop" doesn't run with error "command not found" even though I can see it. Thus, I am unable to modify the nfd.conf and restart the nfd daemon -- I am working on getting this fixed so that we can have more information)
I also noticed that other nodes that interact with these containers have NFD-0.6.2 so I'm trying to get them all working on the same version.
Apologies for saying that "the connection over NFD completely drops", I meant the ndnperf drops with a nack message. To give more backgorund, I am trying to download the same file on multiple containers hosted on the same node via ndnperf. And ndnperf on one of these containers nack-s while the file is not completely downloaded. And NDNPerf log and ndndump log led me to the duplicate nonce generation.
I am downloading NFD from https://github.com/named-data/NFD on to these containers that run Ubuntu 18.04, gcc version 7.4.0.
PS: As soon, as I can restart the nfd daemon on this containers, I will update the NFD log trace.
#4 Updated by Junxiao Shi 3 days ago
Here is the tcpdump log as you asked.
It's not captured correctly. Use
-w flag to generate
.pcap files. stdout is useless.
I couldn't get the log with the FORWARDING trace on.
Then it's useless.
I meant the ndnperf drops with a nack message.
Report a bug on ndnperf issue tracker, as they are not handling Nack properly.
It's not NFD closing the face.