Channel bonding on SLES 10

To get a higher throughput and redundancy it is possible to combine several network interfaces. There are different algorithms available for the load balancing between the network cards. Some need special features or configuration on the switch others don't. In this example I use "balance-rr" which means that the traffic is sent equally through all participating cards (round robin).
The nice thing is that you can pull any network cable and plug it in again. As long as there is always one cable connected you will not see any interruption of the traffic. You have redundancy and a higher throughput. The switch does not need any special configuration. Other algorithms for example use the second card as a backup only but use it only if the first card fails.

Our example looks like this:
Static IP 192.168.1.35 on the virtual bonding interface bond0
Algorithm balance-rr
All traffic is routed through bond0.

The easiest way to start is to use Yast for the initial configuration of the cards. By selecting dhcp you get the appropriate files in /etc/sysconfig/network.
These will be modified as follows (this is my scenario, of course you have to adapt it for yours accordingly):

# cat /etc/sysconfig/network/ifcfg-eth-id-00\:00\:cb\:69\:06\:ca
BOOTPROTO='static'
BROADCAST=''
ETHTOOL_OPTIONS=''
IPADDR=''
MTU=''
NAME='Realtek RT8139'
NETMASK='255.255.255.0'
NETWORK=''
REMOTE_IPADDR=''
STARTMODE='auto'
UNIQUE='JNkJ.IQxIdIhhuH7'
USERCONTROL='no'
_nm_name='bus-pci-0000:05:04.0'

# cat /etc/sysconfig/network/ifcfg-eth-id-00\:11\:85\:0f\:18\:cd
BOOTPROTO='static'
BROADCAST=''
ETHTOOL_OPTIONS=''
IPADDR=''
MTU=''
NAME='Hewlett-Packard Company HP d530 CMT (DG746A)'
NETMASK='255.255.255.0'
NETWORK=''
REMOTE_IPADDR=''
STARTMODE='auto'
UNIQUE='rBUF.7LBxAzvZZlA'
USERCONTROL='no'
_nm_name='bus-pci-0000:05:02.0'

Then we create the file ifcfg-bond0 for the bonding interface:

# cat /etc/sysconfig/network/ifcfg-bond0
BOOTPROTO='static'
BROADCAST='192.168.1.255'
IPADDR='192.168.1.35'
NETMASK='255.255.255.0'
NETWORK='192.168.1.0'
REMOTE_IPADDR=
STARTMODE='onboot'
BONDING_MASTER='yes'
BONDING_MODULE_OPTS='mode=balance-rr miimon=100'
BONDING_SLAVE0='eth0'
BONDING_SLAVE1='eth1'
# or BONDING_SLAVE0='hwcfg-bus-pci-0000:05:02.0'
#    BONDING_SLAVE1='hwcfg-bus-pci-0000:05:04.0'

To route everything through bond0 we need a static route:

# cat /etc/sysconfig/network/routes:
default 192.168.1.1 - bond0

You don't have to change /etc/modprobe.conf.local.

Finally it should look like this:

gpfs:~ # ip a
1: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth1: <BROADCAST,MULTICAST,SLAVE,UP> mtu 1500 qdisc pfifo_fast master bond0 qlen 1000
    link/ether 00:11:85:0f:18:cd brd ff:ff:ff:ff:ff:ff
    inet6 fe80::211:85ff:fe0f:18cd/64 scope link
       valid_lft forever preferred_lft forever
3: eth0: <BROADCAST,MULTICAST,SLAVE,UP> mtu 1500 qdisc pfifo_fast master bond0 qlen 1000
    link/ether 00:11:85:0f:18:cd brd ff:ff:ff:ff:ff:ff
    inet6 fe80::211:85ff:fe0f:18cd/64 scope link
       valid_lft forever preferred_lft forever
5: sit0: <NOARP> mtu 1480 qdisc noop
    link/sit 0.0.0.0 brd 0.0.0.0
6: bond0: <BROADCAST,MULTICAST,MASTER,UP> mtu 1500 qdisc noqueue
    link/ether 00:11:85:0f:18:cd brd ff:ff:ff:ff:ff:ff
    inet 192.168.1.35/24 brd 192.168.1.255 scope global bond0
    inet6 fe80::200:ff:fe00:0/64 scope link
       valid_lft forever preferred_lft forever

gpfs:~ # route -n
Kernel IP routing table
Destination            Gateway          Genmask         Flags Metric Ref    Use Iface
192.168.1.0     0.0.0.0         255.255.255.0   U     0      0        0 bond0
127.0.0.0       0.0.0.0         255.0.0.0       U     0      0        0 lo
0.0.0.0         192.168.1.1     0.0.0.0         UG    0      0        0 bond0

Warning! If you try to shutdown the network at this point or you try to reboot your server the whole system will hang. You have to pull the plug. The reason for this is that the network init script tries to shut down bond0 without detaching the slaves first.
Therefore it is absolutely necessary to modify the init script:

In /etc/init.d/network, at line 820
from:

        stop)

                echo Shutting down network interfaces:

                if [ -z "$INTERFACE" ] ; then

to:

        stop)
                echo Shutting down network interfaces:

                echo "First detach bonding slaves from master:"
                echo "ifenslave -d bond0 eth0"
                ifenslave -d bond0 eth0
                echo "ifenslave -d bond0 eth1"
                ifenslave -d bond0 eth1

                if [ -z "$INTERFACE" ] ; then

To simplify matters I have inserted ifenslave for both slaves directly. A more generic solution would be nice but it works like this.

SLES Administration