NIC Bonding on Blade BL25

[Scott W Cook]

I have two BL25's in a blade enclosure. Both are running RedHat ES v4 Update 1. One of them will bond all 4 nics into link aggregate mode no problem. The other won't. All nics are plugged into the same Cisco switch. I'm using the same script (chaged the ip address though) and it tells me the following:

Bringing up interface bond0: /sbin/modprobe device bond0 does not seem to be present, delaying initialization.

The PSP has been installed and the nics are running BCM5700 drivers.

Here is the script I run to set up the bonding:

#!/bin/bash

cp /etc/modules.conf /etc/xx-modules.conf
echo "alias bond0 bonding" >> /etc/modules.conf
echo "options bond0 miimon=100 mode=4" >> /etc/modules.conf
cp /etc/modprobe.conf /etc/xx-modprobe.conf
echo "install bond0 /sbin/modprobe bonding -o bond0 miimon=100 mode=4" >> /etc/modprobe.conf

cd /etc/sysconfig/network-scripts
cp ifcfg-eth0 xx-ifcfg-eth0

echo "DEVICE=bond0" > ifcfg-bond0
echo "TYPE=ethernet" >> ifcfg-bond0
echo "BOOTPROTO=none" >> ifcfg-bond0
echo "ONBOOT=yes" >> ifcfg-bond0
echo "NETWORK=192.168.64.0" >> ifcfg-bond0
echo "IPADDR=192.168.64.230" >> ifcfg-bond0
echo "GATEWAY=192.168.64.1" >> ifcfg-bond0
echo "NETMASK=255.255.254.0" >> ifcfg-bond0
echo "USERCTL=no" >> ifcfg-bond0

echo "DEVICE=eth0" > ifcfg-eth0
echo "USERCTL=no" >> ifcfg-eth0
echo "ONBOOT=yes" >> ifcfg-eth0
echo "MASTER=bond0" >> ifcfg-eth0
echo "SLAVE=yes" >> ifcfg-eth0
echo "BOOTPROTO=none" >> ifcfg-eth0

echo "DEVICE=eth1" > ifcfg-eth1
echo "USERCTL=no" >> ifcfg-eth1
echo "ONBOOT=yes" >> ifcfg-eth1
echo "MASTER=bond0" >> ifcfg-eth1
echo "SLAVE=yes" >> ifcfg-eth1
echo "BOOTPROTO=none" >> ifcfg-eth1

echo "DEVICE=eth2" > ifcfg-eth2
echo "USERCTL=no" >> ifcfg-eth2
echo "ONBOOT=yes" >> ifcfg-eth2
echo "MASTER=bond0" >> ifcfg-eth2
echo "SLAVE=yes" >> ifcfg-eth2
echo "BOOTPROTO=none" >> ifcfg-eth2

echo "DEVICE=eth3" > ifcfg-eth3
echo "USERCTL=no" >> ifcfg-eth3
echo "ONBOOT=yes" >> ifcfg-eth3
echo "MASTER=bond0" >> ifcfg-eth3
echo "SLAVE=yes" >> ifcfg-eth3
echo "BOOTPROTO=none" >> ifcfg-eth3

service network restart

end


And here is an unbonding script I use when the bonding doesn't work...

#!/bin/bash

cd /etc/sysconfig/network-scripts
mv xx-ifcfg-eth0 ifcfg-eth0
rm -rf ifcfg-eth1
rm -rf ifcfg-eth2
rm -rf ifcfg-eth3
rm -rf ifcfg-bond0

cd /etc

mv xx-modules.conf modules.conf
mv xx-modprobe.conf modprobe.conf

service network restart

end

Thanks in advance for any help.

[Steven E. Protter]

Shalom,

Try the first bond attempt with an active-passive configuration.

Note the only NIC cards I've been able to bond active-active are Intel. Broadcom cards explode in such a configuration.

SEP

[Scott W Cook]

I will try that.

I have 8 blades in the enclosure. Two are BL25 and the rest are BL20. All of them came right up using the scripts I provided. The only one that is causing me grief is one of the two BL25's. Weird.

[Scott W Cook]

I just tried it. Same result. Tells me the same thing as before. bond0 doesn't seem to be present on the system.

I also tried it with just two nics in the bonding team. Same result.

It seems like it's a modprobe issue. I'll have to do some more research.

Thanks again.

[Al Licause]

When your bond0 is up or at least configured, you might want to look at the information in the /proc/net area for details:

cat /proc/net/bonding/bond0
or
cat /proc/net/bond0/info

[Alexander Samad]

What if any messages are there in messages.

You can hand debug the steps. do a modprobe -v to test the bonding module insert and then a bash -x /sbin/ifup.

I have found numeres typing errors this way

[Aaron_125]

think we had the same issue and got by changing the order of entries in modprobe.conf->our modprobe

alias eth0 bcm5700
alias eth1 bcm5700
alias eth2 bcm5700
alias eth3 bcm5700
alias scsi_hostadapter cciss
alias usb-controller ohci-hcd
alias net-pf-10 off
options bonding miimon=100 mode=1max_bonds=3
alias bond0 bonding
alias bond1 bonding