Re: Nexus 1000v and LACP

felixjai · ‎03-10-2010

I'd like to know if anyone has successfully configured LACP on 1000V? I'm having problem with LACP where etherchannels are not establishing. If I don't use LACP, and turn on etherchannel unconditionally, then it works fine. My 1000v is connected to 3120 blade switch.

The port-profile where I use for uplinks to the 3120 is-

port-profile type ethernet system-uplink
vmware port-group
switchport mode trunk
switchport trunk allowed vlan 501-502,900
channel-group auto mode active
no shutdown
system vlan 501-502,900
state enabled

On my 3120-

interface GigabitEthernet1/0/x

switchport mode trunk

switchport trunk allowed vlan 501-502,900

channel-group 10 mode active

'show etherchannel summy' shows the ports are stand-alone ports (I). Again, if i don't use LACP and use 'channel-group mode on', then it works fine.

Any help is appreciated.

Robert Burns · ‎03-10-2010

What version of 1000v are you running? If you're using 4.0.4.SV1.1 and trying to use LACP with vPC-HM, it's not supported. LACP is fully supported with vPC-HM on the latest version 4.0.4.SV1.2.

Can you post the Port Configuration of each interface in the Etherchannel (on the 3120 side).

Regards,

Robert

felixjai · ‎03-10-2010

I'm running 4.0(4)SV1(2). I'm not using vPC-HM. There are 4 x 3120 blade switches on each blade chassis, so there are four NICs for each ESX host, so my 3120 for each port-channel looks like this

e.g.

interface GigabitEthernet1/0/4

switchport mode trunk

switchport trunk allowed vlan 501-502,900

channel-group 10 mode active

interface GigabitEthernet2/0/4

switchport mode trunk

switchport trunk allowed vlan 501-502,900

channel-group 10 mode active

interface GigabitEthernet3/0/4

switchport mode trunk

switchport trunk allowed vlan 501-502,900

channel-group 10 mode active

interface GigabitEthernet4/0/4

switchport mode trunk

switchport trunk allowed vlan 501-502,900

channel-group 10 mode active

interface Port-Channel10

switchport mode trunk

switchport trunk allowed vlan 501-502,900

Robert Burns · ‎03-10-2010

Are you positive all 4 3120's are stacked correctly?

From the VSM you can enable LACP debugging and see exactly why its failing.

1. Shut the port channel on the 3120

2. Enable logging on your SSH session

3. On the VSM "debug lacp error" and "debug lacp demux".

4. Unshut the Port Channel on the 3120

Capture the output and attach here.

You can also do the same from the 3120 side as well, but lets start with the VSM side first.

Regards,

Robert

felixjai · ‎03-10-2010

I'm almost certain that the 3120 are stacked correctly since I have many other blade servers in the same chassis running on standard vSwitch with static etherchannels (channel-group mode on). And without using LACP, these port-channels come up fine. But I'll post the debug output here once i have it.

Also, a different issue on 1000v. This has happened twice so far. The port-profiles created on the VSM disappeared from the vCenter (under Inventory -> Networking -> the 1000v DVS). When I looked at the issue on VSM, it seemed to be vCenter and VSM were out of sync. I needed to go into svs connection config to do "no connect" and "connect" to bring those back. Do you know what would cause the issue?

Robert Burns · ‎03-10-2010

Could be any number of things. Basically there's a network communication between the two hosts. VSM -> vCenter communication typcially uses TCP Port 80, once it connects it should remain pretty solid. Otherwise I would look for network connectivity issues between each whenever the VSM shows disconnected.

The VSM logs will also show you if/when the VSM - vCenter SVS connection flaps. Might be able to correlate it with something happening on your network.

Are you running VSM in HA?

Robert

felixjai · ‎03-10-2010

Yes, it's in HA (a pair of primary and standby VSMs). But why would the port-profiles disappear from vCenter even if communication is lost between VSM and vCenter? I thought once the port-profiles are created and vCenter received them, those should stay regardless VSMs are accessible. Am I wrong? And if the port 80 connection is broken for some unknown reasons, does VSM try to reconnect to vCenter by itself? or I have to do "no connect" and "connect" every time there had been a broken connection?