I've been having the same problem with my amd64 system (a Dell 2850) running current for a long time: it's been there at least since August of 2013 (probably longer), and after upgrading to a fresh current from about two days ago, it's still present.
When the machine gets loaded down with work, I get network pauses of varying length, from a few seconds to a couple of minutes, where all network activity just stops, and I get console messages from various subsystems complaining about no buffer space available - e.g. Sep 1 09:32:49 barsoom openvpn[2896]: write UDPv4: No buffer space available (code=55) Note that the condition that causes this to happen is that the machine has a significant non-networking workload. During this last upgrade, it was really bad during the building of the new release before upgrading, and when I aborted the build job, and restarted it without the "-j 4" parameter to build.sh, that didn't change anything. I'm guessing it has to do with disk I/O, and that the build process is doing enough of that without the parallellization to trigger the problem. Any hints as to how to either resolve this, or debug it further? -tih -- It doesn't matter how beautiful your theory is, it doesn't matter how smart you are. If it doesn't agree with experiment, it's wrong. -Richard Feynman
