RHAS 2.1 and bonding

[Rui Amaral]

Hi all,

I have been requested to do NIC failover on a 3 node Oracle RAC Cluster using RedHat AS 2.1 (w/update 2). The machines are all proliants dl 380 g3 with on board nics of NC7781 (broadcom driver 6.2.11a). We also have dual nics intel nics with the 4.4.12-k1 driver. Now the request is to have the broadcom nics as the primary with the intel as the backups. I have checked the bonding info and it seems to be possible. I have set up a pair of nics using the bonding and tested a failover scenario and it did not failover. The bonding driver is bonding-1.0.3-12HP. My config in /etc/sysconfig/network-scripts is the following:
ifcfg-bond0:

DEVICE=bond0
USERCTL=no
BOOTPROTO=none
IPADDR=1.1.1.3
ONBOOT=yes
GATEWAY=172.16.176.1
NETMASK=255.255.252.0
NETWORK=1.1.1.0
BROADCAST=1.255.255.255

ifcfg-eth1:


USERCTL=no
DEVICE=eth1
IPADDR=1.1.1.3
ONBOOT=yes
BOOTPROTO=none
MASTER=bond0
SLAVE=yes

ifcfg-eth3:


USERCTL=no
DEVICE=eth1
IPADDR=1.1.1.3
ONBOOT=yes
BOOTPROTO=none
MASTER=bond0
SLAVE=yes

in /etc/modules.conf:

alias parport_lowlevel parport_pc
alias scsi_hostadapter cciss
alias usb-controller usb-ohci
# do not put bonding at the end of the modules of you gt into an endless loop
alias bond0 bonding
alias eth1 bcm5700
alias scsi_hostadapter1 qla2300
alias scsi_hostadapter2 qla2300
alias eth0 bcm5700
alias eth2 e1000_4412k1
alias eth3 e1000_4412k1
options bond0 mode=1 miimon=100 downdelay=100 updelay=100 multicast=1 arp_ip_target=1.1.1.3 primary=eth1

does anything strike anyone as being incorrect? Have I missed anything in the setup?

TIA

[Rui Amaral]

forgot to mention that I am testing the interconnects on the cluster (ie. IP=1.1.1.3 - this is what is desired for Oracle Rac) not public links at the moment.

[Alexander Chuzhoy]

The gateway seems to be on the wrong subnet.
Not sure that this is the source of the problem though...