Hi All,

there is a test environment with 3 Oracle Linux PCs and 2 Cisco SG200-08 switch. Each of PCs has got at least 2 NICs (Intel Desktop PCIe). One of this PC should work as a iSCSI Target using Linux software. This PC has got 3 NICs. All PCs are connected to 2 switches and have got IP adresses in different networks. The first network has a gateway and routing to other networks. The second network should only work between these 3 PCs to separate iSCSI traffic. This configuration works without any problems.

Now I would like to configure  2 (of 3 available) NICs together as a bonding interface bond0 on the PC with iSCSI target. The switch has a support for the 802.3ad link aggregation. At the moment I don't use any VLAN etc.I have only configured the corresponding IP for the switch. So I have enabled the configuration for 2 corresponding ports on the switch with dynamic LACP. Then I have configured the PC with the bonding interface with the options: BONDING_OPTS="mode=802.3ad miimon=100". Ater restarting the network on the PC I can see with "sar -n DEV", that both NICs are used. Als "ping"s from other PCs work.

But the used  bonding mode according to the kernel (cat /proc/net/bonding/bond0) is balance round-robin (balance rr) instead of 802.3ad. This is wrong.

I could not find any errors in /var/log/messages. Do I have to enable dynamic LACP  on the Linux PC as well? Should I set any other parameters in BONDING_OPT or in SYSCTL.CONF for such configuration?

I have tried both static and dynamic LAG configuration on the switch. With static configuration the corresponding ports (g5 and g6) are only marked as standby members. I could not get them as active members at all.

Any idea?

Thanks in advance

Andrej