«

I couldn't find a solution to my problem in this group so I am posting
this in hopes that you may find it useful.

Problem:

I started getting these messages in my cisco 3500xl cluster logs every
minute.

%RTD-CLUSTER_MEMBER_7-1-ADDR_FLAP: FastEthernet0/9 relearning 6 addrs
per min

sh mac showed a mac address of 0201.0000.0000 coming from
FastEthernet0/9 along with the real mac of the nic card bound to the ip
address. The host was a Windows 2003 server running the network load
balancing service. This service was generating the 02:01:00:00:00:00
mac transmission.

Solution:
go to network connection properties and look for the service named
network load balancing and uninstall it. This will cause the server to
disconnect for a moment so beware if you are using a remote console or
your users are logged on.

ADDR_FLAP
RTD-CLUSTER_MEMBER
%RTD
NLB
Broadcom NetXtreme 5751 Gigabit Controller

Hello,

here is what the Cisco Error Message Decoder says:

--> 1. %RTD-1-ADDR_FLAP [chars] relearning [dec] addrs per min
Normally, MAC addresses are learned once on a port.Occasionally, when a
switched network reconfigures, due to either manualor STP
reconfiguration, addresses learned on one port are relearned on
adifferent port. Howeve r, if there is a port anywhere in the switched
domainthat is looped back to itself, addresses will jump back and forth
betweenthe real port and the port that is in the path to the looped
back port. Inthis message, [chars] is the interface, and [dec] is th e
number of addressesbeing learnt.

Recommended Action: Determine the real path (port) to the MAC address.
Use the debug ethernet-controller addr command to see the alternate
path-port on which the address is being learned.Go to the switch
attached to that port. Note that the show cdp neighbors command is
useful in determining the next switch. Repeat this procedureuntil the
port is found that is receiving what it is transmitting, and removethat
port from the network.--<

I guess you could use the 'mac-address-table static' command and add
the MAC address to the table, that way, it would not have to be
relearned...

Regards,

Naz

Hello,

here is what the Cisco Error Message Decoder says:

--> 1. %RTD-1-ADDR_FLAP [chars] relearning [dec] addrs per min
Normally, MAC addresses are learned once on a port.Occasionally, when a
switched network reconfigures, due to either manualor STP
reconfiguration, addresses learned on one port are relearned on
adifferent port. Howeve r, if there is a port anywhere in the switched
domainthat is looped back to itself, addresses will jump back and forth
betweenthe real port and the port that is in the path to the looped
back port. Inthis message, [chars] is the interface, and [dec] is th e
number of addressesbeing learnt.

Recommended Action: Determine the real path (port) to the MAC address.
Use the debug ethernet-controller addr command to see the alternate
path-port on which the address is being learned.Go to the switch
attached to that port. Note that the show cdp neighbors command is
useful in determining the next switch. Repeat this procedureuntil the
port is found that is receiving what it is transmitting, and removethat
port from the network.--<

I guess you could use the 'mac-address-table static' command and add
the MAC address to the table, that way, it would not have to be
relearned...

Regards,

Naz

Hello,

here is what the Cisco Error Message Decoder says:

--> 1. %RTD-1-ADDR_FLAP [chars] relearning [dec] addrs per min
Normally, MAC addresses are learned once on a port.Occasionally, when a
switched network reconfigures, due to either manualor STP
reconfiguration, addresses learned on one port are relearned on
adifferent port. Howeve r, if there is a port anywhere in the switched
domainthat is looped back to itself, addresses will jump back and forth
betweenthe real port and the port that is in the path to the looped
back port. Inthis message, [chars] is the interface, and [dec] is th e
number of addressesbeing learnt.

Recommended Action: Determine the real path (port) to the MAC address.
Use the debug ethernet-controller addr command to see the alternate
path-port on which the address is being learned.Go to the switch
attached to that port. Note that the show cdp neighbors command is
useful in determining the next switch. Repeat this procedureuntil the
port is found that is receiving what it is transmitting, and removethat
port from the network.--<

I guess you could use the 'mac-address-table static' command and add
the MAC address to the table, that way, it would not have to be
relearned...

Regards,

Naz

"nazgulero" <> wrote in message
news: ups.com...
> Hello,
>
> here is what the Cisco Error Message Decoder says:
>
> --> 1. %RTD-1-ADDR_FLAP [chars] relearning [dec] addrs per min
> Normally, MAC addresses are learned once on a port.Occasionally, when a
> switched network reconfigures, due to either manualor STP
> reconfiguration, addresses learned on one port are relearned on
> adifferent port. Howeve r, if there is a port anywhere in the switched
> domainthat is looped back to itself, addresses will jump back and forth
> betweenthe real port and the port that is in the path to the looped
> back port. Inthis message, [chars] is the interface, and [dec] is th e
> number of addressesbeing learnt.
>
> Recommended Action: Determine the real path (port) to the MAC address.
> Use the debug ethernet-controller addr command to see the alternate
> path-port on which the address is being learned.Go to the switch
> attached to that port. Note that the show cdp neighbors command is
> useful in determining the next switch. Repeat this procedureuntil the
> port is found that is receiving what it is transmitting, and removethat
> port from the network.--<
>
> I guess you could use the 'mac-address-table static' command and add
> the MAC address to the table, that way, it would not have to be
> relearned...

But - the address is "locally defined" - so windows load balancer is
probably sending it out on multiple ports from the server, or from multiple
servers in a cluster.

Turning it off will fix the cisco issue, but probably break the application
/ server which cant be good.......

The classic fix was to connect all the load balancer LAN connections to a
hub and then put that onto the switch - but that woudl limit the aggregate
thruput to 100 Mbps 1/2 duplex.

in w2k you could run the load balancing software with different network
setups - maybe the same is true here?

what used to be possible was to use a multicast dest address for the load
balancing and unique MACs from each server, or a multicast as a source
address - but no idea how 2003 has "fixed" this.

maybe your m$oft people can set up the server differently? - as it isnt
going to work well like this.
>
> Regards,
>
> Naz
--
Regards

- replace xyz with ntl

Danny wrote:
> I couldn't find a solution to my problem in this group so I am posting
> this in hopes that you may find it useful.

Setting up an etherchannel to the server will fix this too, and work
better than the M$ load balancing mechanism.