UDP CONNECTIONS DON'T WORK!!!

[ALEX778899]

Hello,
scaleway.com support team told us to open an issue here.
IOTA IRI can't work in scaleway.com with UDP protocol, we tested IOTA IRI with other VPS companies and works like a charm.
The problem could be you check the size of packets and for security doesn’t accept to fragment.
You can easy test it the UDP problem from YOUR SIDE, just use https://hub.docker.com/r/bluedigits/iota-node/
I advice you to solve the problem ASAP because IOTA is growing, works very fast with UDP protocol and scaleway.com is loosing a lot of IOTA customers.....

Please let me know ASAP
Thanks in advance

[akomelj]

I concur - UDP fragmentation is the problem and IOTA uses larger UDP packets. I haven't tried VPS servers but I can confirm that it does not work on baremetal servers.

Instead of running IOTA in a Docker container, you can use nping to send large UDP packets to Scaleway VPS/baremetal server and monitor received packets with tcpdump. You will see that UDP fragments get discarded.

I.e. nping --udp -p <VPS port> --data-length 1650 <VPS IP> will send UDP packet of 1650 bytes to selected IP and port.

This issue is really annoying as more and more IOTA nodes are using UDP due to Nelson and TCP bugs in IRI.

[brmzkw]

Hi,

We've been in the rush in the last few days, but I'll give an answer to the issue as soon as possible.

[brmzkw]

Hi,

We improve the NAT located in Paris. Can you please retry?

[akomelj]

No, it's still the same. Only 1480 bytes (in a UDP frame of 1514 bytes) arrive when sent from outside network.

Tested with nping --udp -p <VPS port> --data-length 1650 <VPS IP> on the sending side and tcpdump -i eth0 -s 2048 -w udp.pcap udp port <VPS port> on the receiving side (parameter -s 2048: store 2k of packet data to output packet file -w). PCAP file contains shortened packets.

[viossat]

Still the same here.

[carbonimax]

Same issue for me today

[zoran]

same here

[mmarchand]

Hi guys, I can confirm we identified the issue (though a bit tricky), it's not a simple and quick fix, but the team is working on it
I hope to have good news within the next few weeks, thanks for your patience

[DarkShimy00]

Do you have any updates?

[mmarchand]

yes, we have a fix, I am checking for deployment ETA now

[DarkShimy00]

Hi, any news on that?

[mmarchand]

should be in production on all NAT servers within the next week

[justinclift]

@akomelj Have you had a chance to retest this from your end? 😄

[akomelj]

Hi, sorry for the delay - I actually didn't know that the fix was had already been installed. :-(

Anyway, I have just run the tests and now it appears that all fragments are coming through. Great!

Wireshark dump screenshots w/ all fragments:

[ALEX778899]

Too much time, I don't need anymore.

[justinclift]

Excellent. Thanks for testing and confirming @akomelj. 😄