On Wed, 19 Mar 2025, Roman Riabenko via wrote:
To identify whether MTU was an issue, I tested manually by pinging the
address with packets of different sizes. I do not know whether there is
a better way to do that. I described that in the linked post.
I, too, don't know if there is a better way to do this, as it's what I've
done as well. A small improvement might be to do the size binary search
thing with `mtr`, then you can at least see which hop has the smaller MTU.
As I understand it, one should expect to get an ICMP packet indicating too
large of packets that were not fragmentable (as I also understand, Linux
now sets "don't fragment" by default), so that Linux can dynamically learn
the path MTU. Likely, something is not sending that package that should
be, or it's being filtered erroneously before it can reach your system.
Tip: you might be able to see any smaller MTUs that Linux has learned with
ip route show cache
or
ip -6 route show cache
I don't know how to set the MTU for a particular process, but you can set
it for a particular peer:
ip route add 209.51.188.168/32 via <gateway> mtu <lower-mtu>
or
ip -6 route add 2001:470:142::168/128 via gateway mtu <lower-mtu>
(those are the current addresses for git.savannah.gnu.org
Hope that helps,
Jack
