I'm well aware that variants of this question have been asked before, but I still haven't found a particularly good answer.
What I'm trying to do is write a Python script that will measure the current bandwidth use and the total bytes uploaded/downloaded by process and all of its children and kill it if either of these values exceeds a user-defined threshold. Its purpose will be to limit the network usage of a backup client.
I've considered a couple of options for measuring process bandwidth:
bwmon does exactly what I want, but frustratingly it won't work for me in its current form. The problem is that it relies on reading bytes= and packets= fields from /proc/net/ip_conntrack in order to measure bandwidth. For some reason these fields just don't exist for me - I suspect it might be to do different kernel versions.
Currently the most workable solution I've found would be to parse the output of strace -e trace=recvfrom,sendto -p <PID> as per this suggestion. However, this is a bit messy and raises nasty permissions issues - in order to attach to an existing process I would need to sudo strace, and I'd prefer to avoid running shell commands with superuser privileges from within scripts.
I wonder if anyone has a more elegant suggestion?
I am not very familiar with the contents of /proc, but if there is some other log file that will list packet sizes and counts then I could perhaps modify bwmon to use this instead.
Not a programming-related answer, but: try doing this:
sudo modprobe ip_conntrack
sysctl -w net.netfilter.nf_conntrack_acct=1
Enabling the module and sysctl setting globally is distribution-specific, but at least until you reboot, bwmon should work if it needs bytes and packets in /proc/net/ip_conntrack.
Based on http://ubuntuforums.org/showthread.php?t=2046187 and http://forums.gentoo.org/viewtopic-p-6677939.html (and thereby ultimately on Google).