Ethernet RX overruns and dropped packets
Hi,
Can someone explain why an interface would start showing dropped
packets and overruns? I have about six machines on a local LAN (the IP
is associated with the br0 device), and all have at least some amount
of dropped packets. This is one example from one of the machines on
the LAN; the LAN interface on the gateway machine is very similar.
eno1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet6 fe80::ec4:7aff:fe7a:73f4 prefixlen 64 scopeid 0x20<link>
ether 0c:c4:7a:7a:73:f4 txqueuelen 1000 (Ethernet)
RX packets 2294973231 bytes 1227551884960 (1.1 TiB)
RX errors 0 dropped 159933 overruns 2252 frame 0
TX packets 2707484667 bytes 1948072588485 (1.7 TiB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
device memory 0xc7200000-c727ffff
I recently rebooted the gateway and noticed it there first. It's a
fedora25 system acting as a gateway with shorewall. The LAN side is a
1Gbs ethernet on a gigabit switch. The WAN side is a 10mbit ethernet
link in a colo. I suspect this machine is the cause, as nothing's
changed on the LAN machines for a while, and the dropped packet count
isn't incrementing fast enough to coincide with greater than 1TB of
traffic.
I have IPMI access to the machines on the LAN, so can do testing, but
I don't have IPMI access to the gateway, so can't really do much
without having to drive to the colo first.
What's the typical cause of these errors? I thought it was perhaps the
duplex mode or other link setting, but they all appear to be the same
(1000/full).
There aren't any dropped packets or overruns on the WAN interface on
the gateway, but could some signal or other data from the WAN side be
causing this?
I can run wireshark or something similar, but it's been a while, so if
that's your recommendation, I'd really appreciate it if you could
provide specific traces you'd think were best.
Ideas greatly appreciated.
Thanks,
Alex