CentOS - Jan 2010 - Lockup using stock r8169 on 4.8 in gigabit mode during heavy transfer on lan

I have no output in the logs, but /etc/init.d/network restart fixes the issue.
I am rsync/scp (10GB) data from one 100MB Full duplex host to an ASUS M3A78-EM
(RTL8111B/C). It locks up around the 1.5GB Tx mark.
If I ethtool -s eth0 autoneg off speed 100 duplex full it does not happen again.

Ideas?

--
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
-                                                               -
- Jason Pyeron                      PD Inc. http://www.pdinc.us -
- Principal Consultant              10 West 24th Street #100    -
- +1 (443) 269-1555 x333            Baltimore, Maryland 21218   -
-                                                               -
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
This message is copyright PD Inc, subject to license 20080407P00.

Jason Pyeron wrote:
> Ideas?
realtek sucks(massive evidence on the net over the past 10+ years),
get a real NIC(no pun intended).

nate

----- "Jason Pyeron" <jpyeron at pdinc.us>
wrote:
> I have no output in the logs, but /etc/init.d/network restart fixes
> the issue.
> I am rsync/scp (10GB) data from one 100MB Full duplex host to an ASUS
> M3A78-EM
> (RTL8111B/C). It locks up around the 1.5GB Tx mark.
> If I ethtool -s eth0 autoneg off speed 100 duplex full it does not
> happen again.
> 
> Ideas?
Part of the problem is that you mention r8169 kernel module in your subject but
your hardware is RTL8111B. Realtek drivers are notorious for loading on improper
hardware and appearing to work properly. You need to get the proper driver for
your NIC. I've run into this countless times with the r8169 driver and
non-r8169 Realtek NICs.

And, on a side note, Realtek may possibly be the worst NIC out there, especially
for server use. It may link at gigabit speed, but I bet a trusty old 3Com or
Intel Pro 100 will get better throughput. Install a different NIC as fast as you
can and never look back.

--Tim