Re: 4507R Log Message

rtjensen4 · ‎12-09-2009

Hello!

I'm recieving this log message

%C4K_EBM-4-HOSTFLAPPING: Host 00:50:56:B6:10:0A in vlan 102 is moving from port Gi4/2 to port Gi4/39

It occurs fairly often and not with just these two ports. I traced cables and the ports mentioned in log message run into the same ESX Host server. I haven't had a chance to touch base with my Server guy yet, but I'd assume those ports are in the same network. Would I be correct in my assumption? Has anyone had this issue before? I'm assuming this is a problem with how the ESX host is configured?

Output interpretter suggests a spanning-tree loop:

C4K_EBM-4-HOSTFLAPPING (x1):Host [mac-addr] in vlan [dec] is flapping between
port [char] and port [char]

Explanation: The specified host is detected as a source address on multiple ports.
Typically, a host is  supposed to be learned on only one port. A spanning tree
loop is the most common cause of this  condition. All traffic from the specified
host is temporarily dropped. After 15  seconds, forwarding  is reenabled. If the
problem persists, the traffic from the problem host continues to be dropped  indefinitely.

Recommended Action: Make sure that spanning tree is enabled to prevent spanning
tree loops. If you  have spanning tree disabled, make sure that you have no spanning
tree loops in your network.

Giuseppe Larosa · ‎12-09-2009

Hello Rtjensen,

this can be the result of not matching NIC teaming configuration:

if on the server side they have used active/active the ports should be configured for an etherchannel bundle and terminated on the same switch.

you neek to check with them their settings to fix this.

Hope to help

Giuseppe

rtjensen4 · ‎12-09-2009

Ok great. Thanks for pointing me in the right direction. Would we use an acutal "Etherchannel" or use something standards-based like Pagp? Auto negot

iate the channel or statically configure it?

vvasisth · ‎12-09-2009

ok if you have a server on the other side and assuming the NIC teaming is done as it should be done (please cross check this )

in this case you need to use MODE "on" for etherchannel as LACP and PaGP wont work.

Mode on will force etherchannel on those ports irrespective of the device on the other side as a Server wont understand Pagp or LACP.

regards,

Varun

Giuseppe Larosa · ‎12-09-2009

Hello Varun,

LACP is standards based (802.3ad) and can be supported by modern server operating systems we use it with servers and it is the best choice for the reasons I've explained in my previous post.

PAGP is proprietary so it is not applicable.

mode on has the big drawback that it is not able to detect who is connected on the other side of each member link.

Hope to help

Giuseppe

vvasisth · ‎12-09-2009

ok, i was not aware of servers can understand LACP thanks for the infor

however i have seen mode on working fine for lot of my customers. I do second your opinion about the drawback of mode on.

Great thanks.

rtjensen4 · ‎12-09-2009

I did some more digging. Looks like we need to use LACP. WHen I configure the channel-group # on the switch, that's only significant to the switch correct? For example, if i configure channel-group 6 on the switch, the server doesn't need to match that right?

Giuseppe Larosa · ‎12-09-2009

Hello,

sorry I've missed these details.

LACP is strongly recommended, PAGP cannot be used because it is Cisco proprietary and yes bundle-id is locally significant

We are moving most new servers to this kind of setup.

to be noted if for any reason the remote end stops to send out LACP message frames the ports will be suspended after 90 seconds.

This is indeed a security measure that allows to detect if cables are moved.

using channel-group mode on is not recommended because it has no detection capabilities.

Hope to help

Giuseppe