Creating ip alias on bonded interface ie. bond0:1

Question

System: HP Proliant DL360 G5 running CentOS 5.4

Bonded interface is working fine for a long time. I just went to add an alias the way I always have on a regular interface, and on first check it works (pinging on the local box) but it is not accessable from outside (iptables is turned off). In addition with this setup the normal network response started to decline, hanging for around a minute before I could get a prompt on login.

Here are my config files:

[root network-scripts]# cat ifcfg-eth0 
DEVICE=eth0
BOOTPROTO=none
ONBOOT=yes
MASTER=bond0
SLAVE=yes
USERCTL=no

[root network-scripts]# cat ifcfg-eth1
DEVICE=eth1
BOOTPROTO=none
ONBOOT=yes
MASTER=bond0
SLAVE=yes
USERCTL=no

[root network-scripts]# cat ifcfg-bond0 
DEVICE=bond0
BONDING_OPTS="mode=1 miimon=100"
BOOTPROTO=none
ONBOOT=yes
NETWORK=10.2.1.0
NETMASK=255.255.255.0
IPADDR=10.2.1.11
USERCTL=no

[root network-scripts]# cat ifcfg-bond0:1 
DEVICE=bond0:1
BOOTPROTO=static
ONBOOT=yes
NETWORK=10.2.1.0
NETMASK=255.255.255.0
IPADDR=10.2.1.12
USERCTL=no

any thoughts?

Solution 1

Does the bonding kernel module loads properly or not ?

If not, it could be related to a dependency issue w/ ipv6, appeared in rhel5.4 with bonding driver 3.4.0.

the following commands should resolve this issue:

# touch /etc/modprobe.d/disable-ipv6
# echo "options ipv6 disable=1" >> /etc/modprobe.d/disable-ipv6 

Solution 2

You do not have a GATEWAY line. The only thing that can reach those interfaces are other devices on the same vlan

Comments

[edgars]

So did you solve it?

[Govindarajulu]

Rubbish. Adding aliases for eth0 and eth1 will not solve anything. It is perfectly ok to create an alias for bond0 without aliasing its slaves.

[voretaq7]

Please read blog.serverfault.com/2011/06/09/press-the-green-button-twice and consider improving your answer.