First time here? Check out the FAQ!
THIS IS A TEST INSTANCE. Feel free to ask and answer questions, but take care to avoid triggering too many notifications.
0

weird issue while uploading (POST) to GitHub only

  • retag add tags

Hi,

While I'm uploading an image to an issue to github.com from my ubuntu 16.04 LTS computer (say computer A), it fails after some minutes and an error is reported : sometime a 400's and I see in the dev pane of firefox a "connection reset".

This computer A is directly connected to the ISP's box (which do the NAT job). Now, if I configure this same computer A to use my phone as a Internet gateway : it works !

I have another Windows computer B on the same network (connected to the same box), no issue either.

yet another laptop computer C with a 16.04 LTS connected to the same network : FAIL I tried to achieve the same upload with a ubuntu 18.04 LTS on computer A and it also fails.

So might it be a problem with Ubuntu, or a TCP issue somewhere ? I did capture the traffic with wireshark while uploading : see http://dl.free.fr/gY7kZ1b9B

Could it be something specific to Ubuntu ?

EDIT - I came to the fact that set mtu value to 1470 solve the issue on computer A. 1471 won't work. What is this problem ?

mh-3's avatar
1
mh-3
asked 2019-01-18 09:40:30 +0000, updated 2019-01-18 10:46:30 +0000
edit flag offensive 0 remove flag close merge delete

Comments

add a comment see more comments

1 Answer

0

The problem is that the negotiated MSS is too large that it would go through the network un-fragmented.
As your outbound packets ae the Dont Fragment bit set in the IP header you should normally see inbound ICMP packets if fragmentation was required along the path (Path MTU Discovery).
Those ICMP packets would carry the netx hop's MTU size so the sender can learn and dynamically adjust/reduce the MTU size of the host route.
Those ICMP packets however are not arriving at the sender so it continues to send (re-transmit) with too large IP packets that all suffer the same death somewhere along the path.
So the underlying problem is that necessary ICMP packets for PATH MTU Discovery process are either not sent or blocked somewhere in the network.
Manually reducing the MTU below the bottleneck's MTU size circumvents this problem.
Hope this explains it to your satisfaction.
Regards Matthias

mrEEde's avatar
4k
mrEEde
answered 2019-01-20 10:45:18 +0000
edit flag offensive 0 remove flag delete link

Comments

add a comment see more comments

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account. This space is reserved only for answers. If you would like to engage in a discussion, please instead post a comment under the question or an answer that you would like to discuss.

Add Answer