Update 2010-10-21: Even with the new kernel module (r8168) the Hetzner monitoring reported short offline times now and then (about one ping failure every day). Compared to the old module (r8169) the problems occurred much less often. To examine the problem, I set up a cronjob to report the network status on the host system with a small script:
#!/bin/bash echo -n $(date +%Y%m%d-%H%M) echo -n " `/sbin/ethtool eth0 | grep "Link detected"`" if ! ping -c 1 -q google.de >/dev/null then echo " google ping: FAIL" else echo " google ping: OK" fi
Surprisingly, this script did not report any network errors and since then also the Hetzner monitoring stopped reporting errors. In result my solution till this day is to use the module r8168 and to run my “network keep alive” script periodically (every minute).
Recently, I noticed some network problems (package losts) with an almost fresh XenServer 5.6 installation on a Hetzner EQ root server running a (SSH sessions disconnects, host and client system are not pingable for some minutes). After some research I found some resources that point out problems with the r8169 kernel module:
- http://wiki.hetzner.de/index.php/Installation_des_r8168-Treibers (German)
- http://forum.hetzner.de/wbb2/thread.php?postid=138116#post138116 (German)
Summarized the resources suggested to replace the driver r8169 with the r8168 driver to eliminate the network problems. Of course, the driver has to be replaced only in the host.
lspci and lsmod show that the root server has a Realtek network card and that the problematic r8169 driver is used:
# lspci [...] 06:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168B PCI Express Gigabit Ethernet controller (rev 02) [...] # lsmod | grep r81 r8169 29108 0 mii 5040 1 r8169
Replace r8169 module with r8168
As suggested, I replaced the r8169 driver with a new r8168 one. Unfortunately, there is no pre-compiled module for the XenServer 5.6. Here is a step-by-step introduction how to manually build the r8169 module and replace the r8168 module in the host system:
- Download XenServer’s DDK (Driver Development Kit) virtual machine from
https://www.citrix.com/English/ss/downloads/details.asp?downloadId=2301285&productId=683148 (Additional Components -> XenServer Support Components)
- Copy the downloaded ISO to XenServer.
- Mount the ISO:
# mount -o loop /PATH/TO/ISO/XenServer-5.6.0-DDK.iso /mnt/PATH
- Import the DDK virtual machine:
# xe vm-import filename=/mnt/PATH/ddk/ova.xml
- Wait some seconds/minutes for the import job to be finished.
- In XenCenter add a network device to the newly created DDK virtual machine, start the machine and enter a new root password.
- Configure the network in the DDK VM; for a Hetzner EQ server this the typical settings are:
DEVICE=eth0 BOOTPROTO=static IPADDR=IP-ADRESS-OF-HOST-SYSTEM NETMASK=255.255.255.224 ONBOOT=yes TYPE=ethernet
IP-ADRESS-OF-HOST-SYSTEM dev eth0 scope link default via IP-ADRESS-OF-HOST-SYSTEM
nameserver 184.108.40.206 nameserver 220.127.116.11
- Download the newest r8168 sources from ftp://WebUser:2mG8dBW@18.104.22.168/cn/nic/. I used the file r8168-8.018.00.tar.bz2 (49 KB 01.04.2010 09:59:00) (Download link: ftp://WebUser:2mG8dBW@22.214.171.124/cn/nic/r8168-8.018.00.tar.bz2)
- Unpack the sources and build the module (this creates the file r8168-8.018.00/src/r8168.ko):
# tar vjxf r8168-8.018.00.tar.bz2 # cd r8168-8.018.00/ # make all
- Copy the created r8168.ko to /lib/modules/126.96.36.199-0.1.1.xs188.8.131.52.111158xen/kernel/drivers/net/r8168.ko in the host system and set permissions to 0744.
- Rename the old r8169 module to prevent loading:
# mv /lib/modules/184.108.40.206-0.1.1.xs220.127.116.11.111158xen/kernel/drivers/net/r8169.ko /lib/modules/18.104.22.168-0.1.1.xs22.214.171.124.111158xen/kernel/drivers/net/r8169.backup.ko
- Edit /etc/modprobe.conf:
alias eth0 r8168 alias r8169 off
- Unload the old module, load the new one and reboot the system (use one line for all commands, because rmmod will disconnect the server from the network):
# rmmod r8169 && depmod -a && modprobe r8168 && reboot
After the system is again up and running again, lsmod should no longer list the r8169 module:
# lsmod | grep r81 r8168 95780 0
- I found some useful information in the howto “Compiling the Realtek RTL8111 driver on XenServer by Martin Guppy”.
- Here is a link to my post “XenServer 5.6: Netzwerkprobleme / Realtek 8169” (German) in the “Hetzner Online Diskussionsforum”. The user “drieschel” has attached a install script that can be used alternatively for step 10. to 13.
- The r8168.ko file I’ve built can be downloaded here: http://blog.janjonas.net/wp-content/uploads/2010/08/r8168-8.018.00.zip