I have seen the same probelm with Linux and Solaris machines (both use
MTU path discovery, IMA) to certain sites around the globe. Resetting the
MTU down (so much as one byte) fixes the problem.
I am not the only person seeing this problem... Several of our customers have
seen the same thing. When this happens do the following:
0> traceroute to the site (several times) [save the trace for reference]
1> ping the site to see if it is just general packet lose.
2> pattern ping the site
a> ping -p 0000 <site> (all 0 data)
b> ping -p ffff <site> (all 1 data)
c> ping -p aaaa <site> (alternating 10)
d> ping -p 6666 <site> (alternating 01)
3> 'ping -s 1492' to see if someone is droping full packets.
4> pattern ping the site with big packets
a> ping -s 1492 -p 0000 <site> (all 0 data)
b> ping -s 1492 -p ffff <site> (all 1 data)
c> ping -s 1492 -p aaaa <site> (alternating 10)
d> ping -s 1492 -p 6666 <site> (alternating 01)
For example, I cannot talk to melkki.cs.helsinki.fi (ftp2) without dropping the
MTU to 1499 first. I have a detailed trace of the activity clearing showing
a packet being lost repeatedly.
--Ricky