Server platforms featuring Supermicro mainboards of the types "X9SCL" and "X9SCM" might be unreachable after a certain time of operation (when CentOS 6 is used). The server is reachable only after a reset has been triggered via the customer control panel.

The root cause of this is the relatively old "e1000e" driver (Version 2.3.2-k) for the Intel network chips. This can be verified by reviewing the file /var/log/messages, which should contain the following lines:

Jul 21 19:58:52 x-x-x-x kernel: Hardware name: X9SCL/X9SCM
Jul 21 19:58:52 x-x-x-x kernel: NETDEV WATCHDOG: eth0 (e1000e): transmit queue 0 timed out

The output of ethtool -i eth0 should give the following result:

driver: e1000e
version: 2.3.2-k
firmware-version: 2.1-2
bus-info: 0000:02:00.0

This issue can be fixed by using the more up-to-date driver version (Version 3.1.0.2). For this the current driver module from the "Enterprise Linux" repository is required. This can be achieved with the following commands:

rpm -Uvh http://www.elrepo.org/elrepo-release-6-5.el6.elrepo.noarch.rpm
yum update
yum install kmod-e1000e
depmod -a

After a subsequent reboot the updated driver should be active. The output of ethtool -i eth0 should give the following result:

driver: e1000e
version: 3.1.0.2-NAPI
firmware-version: 2.1-2
bus-info: 0000:02:00.0