Port Configuration for NIC Teaming

[James Davies]

I dont do servers! so dont have much idea on this one, any help appreciated.

Server has been installed with 4 NICs, these have been placed into a team, using LACP,

My job was to do the switch side of things, this server is in vlan 10 , so I created 4 interfaces as such:

int range gi0/5 - 8

switchport mode access

swithcport access vlan 10

channel-group 5 mode active.

and for the port channel

int po5

switchport mode access

switchport access vlan 10.

all interfaces were shutdown, configured the team NIC with its IP then brought up the port channel, no luck though.

the interfaces stay in suspend, do I need to specify the vlan on the teamed NIC? its a HP blade I believe

Thanks

[John Blakley]

You shouldn't have to do the vlan on the other side since you're not configuring these ports for trunking. I'm not versed on HPs at all, but first thing I would check is for a firewall. If there's something blocking the negotiation frames, that would keep it from coming up. You could also try "mode on" to see if it comes up. The recommendation on these forums is not to leave it in the "on" state, but instead let it negotiate the protocol. I've run into situations though that simply don't work any other way, so I've had to leave some etherchannels configured in the "on" state...

HTH,

John

[Steve Fuller]

Hi,

If you specify the VLAN on the teamed NIC then you'd also need to change the switch configuration to use switchport mode trunk rather than switchport mode access as you have at the moment. I would leave the settings as they are for the moment unless you can confirm that IEEE 802.1Q trunking has been enabled on the server.

You say the interfaces are suspended. Is there any error message you see on the switch or do you get an indication of why the interface is suspended?

Also is the server correctly configured for LACP? Can you post the output of show etherchannel summary and show lacp counters to make sure the switch is actually receiving LACPDU from the server.

Regards

[Steve Fuller]

Hi James,

Can you post the output as requested in my previous response?

Regards

[James Davies]

I work away from the network so have to keep reversing the changes,

there are no LACP counters being recieved, and sh ether chann summ, shows the ports as suspended.

I just dont know whats wrong? do I need to configure the load balancing method on the cisco switch?

[Steve Fuller]

Hi James,

If the switch is showing a value of 0 for the LACP counters then the server is not configured to operate as an IEEE 802.3ad LAG running LACP. You need to get the server team to check their configuration.

Depending upon the switch model and IOS version, you might try adding the command no port-channel standalone-disable on the port-channel interface. This will make the member interfaces go into an Individual (I) rather than Suspended (s) state, but at least one link of the LAG should ten work and you'd connectivity to the server. You might need to bounce the port-channel interface for this command to take effect though.

Regards

[James Davies]

There is no VLAN specified on the teaming end, its just set to LACP teaming. The virtual adapter has been assigned its IP and I have connectivitiy through one port on the switch and it seems to be sharing the traffic accross the 4 NICS (Round Robin)not teaming them.

I dont have access to the switch today, but the etherchannel summary shows the port as suspended.

Im sure I tried trunking the switch end, instead of access mode., but didnt specify a VLAN on the server end, perhaps this is worth a try.

[Steve Fuller]

Hi,

At this stage I wouldn’t change the switch port mode. Even if the server is configured for IEEE 802.1Q trunking and the switch is configured as an access port, that should not suspend the port(s).

I think the problem is in respect of the Link Aggregation (LAG) configuration and the switch and server not having compatible settings. We know you have an “active” IEEE 802.1AX (or 802.3ad) LAG on the switch as you’ve configured channel-group 5 mode active. The active keyword is what tells the switch to use the standards based Link Aggregation Control Protocol (LACP) and to send LACPDU.

What we don’t know is how the server is configured. The simplest way to check that is using the show lacp counters command when you can access the switch. If the receive counters are zero, then the server is not configured as an active LAG, or perhaps not even as a LAG. The output from show etherchannel summary is also useful as we can see the state of the LAG and each member link.

If you can also find the OS the server is running that would help. If it’s Linux then ask the server team to send you the output of cat /proc/net/bonding/bond0 (or whatever name they’ve given the LAG).

Regards

[Dennis Mink]

Is this for an ESXi host?  I am guessing it is.

If it is ESXi, I would usae the following load balance algorithm:

sw1(config)#port-channel load-balance src-dst-ip, which translates to IP Hash on the ESXI side

check this:  http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1004048

[James Davies]

Its not a VM, its a physical box