Running Debian, when transferring files from the server to my pc, the transfer is suddenly interrupted. Some smaller files (couple of MB) seem to work most of the time, but bigger files mostly fail at random points. After the transfer fails, the entire TCP/IP stack seems to be out on the server, even ping doesn't work anymore (every host is unreachable).
I tried the follwing:
- Connecting my PC directly to the server, ruling out errors with switches etc in between.
- Trying different client pc's, ruling out a problem with the client pc
- Trying different files from different hard drives on the server, ruling out hard disk issues
- Trying different protocols (ftp, scp, smb), ruling out a problem with a specific protocol or server
The problem appeared in any of the above configurations. So I'm leaning towards either a software issue are a hardware issue with the NIC. However, syslog and dmesg doesn't indicate any problems. Ethtool results also seem normal...
I'm really out of ideas on what's broken and how to proceed...
ip addr
andip route
in your question. In this case also those ofip rule
andiptables -L OUTPUT -nv
. – Hauke Laging Jun 14 '14 at 17:18iptables -L
for any sort of debugging. Useiptables-save
to show unadulterated rules. – pilona Jun 14 '14 at 21:35iptables-save
outputs the rules without pretty printing. It shows them as they are usually stored. If you have any non-trivial ruleset that doesn't fit nicely into the pretty-printed model ofiptables -L
, it becomes unreadable. We don't need to see the packet or byte counters for every rule, nor will every rule necessarily have protocol, iif, oif, saddr, and daddr matches. If we're going to modify the rules anyway, why bother translating them to another form that needs translating back? My 310 lines of rules on my router's firewall are unreadable in this form. – pilona Jun 14 '14 at 22:02