NIC Teaming with two GbE2c switches in C3000

[John van_2]

Hi,

I have a BL460c server with an additional Mezzanine NIC card. My C3000 enclosure has two GbE2c switches.

I have set up a HP Network Team with one port from the embedded NIC and one port from the mezzanine NIC to provide a redundant link through both switches to the LAN.

The team connects but constantly drops one port (embedded) with status "Failed (Rx Path)". It repeatedly tries to pick up the second port but then reverts to the status identified above.

Any suggestions as to what I might have done wrong?

Thanks,

John

[Ken Henault]

What OS is this for? What teaming method did you select?

[John van_2]

Hi Ken.

Windows 2008 X64

Used the HP NCU
Team Type Selection : Automatic
Transmit Load Balancing Method: Automatic

Is that what you were after?

John

[Lmm_1]

Do you have crossconects enabled (Port 17 and 18)? can you run a Network survey? from command propmt: hpnetsvy >filename.txt

rgds,

[John van_2]

Yes. 17 and 18 are enabled. The survey details are attached. I have removed the details of Team 1 which is an iSCSI team connected to a SB600c storage blade.

John

[Miika T]

I'd change the nic teaming to NFT + Transmit load balancing. As the team is split between to swithes you can't select switch assisted LB, which can happen with automatic setting.

-MiikaT

[John van_2]

Hi MiikaT.

I tried your suggestion to set NFT. The load balancing greys out but I assume it still does load balancing. Unfortunately, the problem remains. After a second or two one of the NIC's is disabled. The event log actually says "All but one is disabled". Because there are only two, it means just one of the two NIC's is disabled.

I have found a reference which says Tree Spanning should be disabled on the switch ports to which this server connects. I haven't been able to work out how to disable Tree Spanning on individual switch ports on the GbE2c. I can only find where I can disable Tree Spanning by VLAN.

Would you recommend I try that?

Thanks,

John

[Ken Henault]

That is the correct behavior. NFT is active/passive teaming. Both NICs need to be connected to the same switch for active/active teaming.

See this white paper for a better understanding of the available options:

http://bizsupport.austin.hp.com/bc/docs/support/SupportManual/c01415139/c01415139.pdf

[Lmm_1]

The recommended port configuration is port-fast enabled (Cisco terminology), here is the gbe2 equivalent:

>> # /cfg/l2/stp 1/port 20
>> Spanning Tree Port 20# fastfwd ena
>> Spanning Tree Port 20# apply
>> Spanning Tree Port 20# save

I would try to force the team mode to TLB. By default Port 17 and 18 are members of Trunk group 1, make sure trunk 1 is enabled and can carry traffic of the VLAN where the Ports connected to the teamed NICs belongs.

[John van_2]

I applied fastfwd to the ports connecting the blade server to the switches. These ports are in VLAN1. Ports 17 and 18 are members of Trk1 which is enabled. 17 & 18 are currently not in VLAN1. I also have VLAN10 which is the iSCSI VLAN connected to the SB600c through the other port from each NIC. Ports 17 & 18 are in VLAN10. If I put those ports in VLAN1 as well then the LAN Team works OK but the iSCSI Team fails.

I thought ports could pass traffic for more than one VLAN. I'm guessing I need to change configuration to allow that to happen, but I'm not sure what that may be.

Thanks,

John