IP device tracking (duplicate IP address)

[Jason Flory]

Hello Everyone,

We are having issues with our Virtual server environment where servers come up with duplicate IP address for address 0.0.0.0 and server will stay offline until disable and re-enable nic.

My research has led me to ip device tracking in our switches. 

Host servers are IBM blades that are connected to Cisco switch modules within the blades.  These switches are not stackable and are layer 2.   All switches have 4 uplinks to our core switch which handles all layer 3 functionality.  My problem is that the ip device tracking commands do not exist on the switches directly connected to the host servers.  These commands are only available on our layer 3 switches which are 3850s.  However when I show device tracking on the core it seems to be only tracking for servers directly connected not the blades.

What is recommended is we change the delay for the ip device tracking probe.  If the ip device tracking commands are not available what do you recommend that we do.

Thanks

[Reza Sharifi]

Hi,

For test purpose, can you disable IP tracking and see if the problem goes away?

On the Cisco switch modules, do you have a default route?

If yes, what is it pointing to?

Is this happening an all servers/blade switches or specific servers connecting to specific switches?

HTH

[Jason Flory]

Yes it seems to be happening on all VMs connected to hosts that are on the IBM blades which have cisco switch modules that do not have the IP device tracking commands. 

They do have ip default-gateways set but not a default route because they are only layer 2.  The Ip default-gateway is pointing to our core switch.

I cannot disable because the switch does not have the commands available.  Our core switch does have the commands but does not seem to be tracking for anything on the blades.

Thanks

[Reza Sharifi · Accepted solution]

Have you seen this document:

3. Duplicate IP address - Part 1
Several configuration choices can result in the BladeCenter reporting duplicate IP addresses, even when there are no known conflicts. Duplicate IP address - Part 1 discusses an issue with blade servers reporting a duplicate IP address.

The most common cause of a blade server reporting a duplicate address is the result of placing one of the interfaces of a blade server in the same VLAN as the Management VLAN of the Cisco Systems CIGESMs (default is VLAN 1), with an IP address in the same subnet as the Management address used internally by the Management Modules to communicate with the Cisco Systems CIGESMs.

http://www.redbooks.ibm.com/abstracts/tips0423.html

Also, have you opened a TAC case to make sure this is not a bug in the IOS?

HTH

[Jason Flory]

Are servers are not in our management vlan.  I need to be clear these are only the virtual machines that are hosted on the blade servers.  I am pretty sure I have found the cause which is ip device tracking needs to be tuned to allow a delay however the version of IOS does not have the IP device tracking commands available.  Are these commands typically available on layer 2 switches. 

TAC may not support the blade switches.  I suppose I could open it on the core switch.

[Jason Flory]

Oh yeah one other thing is that the nics on the blade hosts are trunked to switch

[Jason Flory]

So I looked at our cisco blade modules and sure enough they had a management IP on the same subnet as the servers.  The main management IP was not on the same subnet but looked like someone tried to setup this up on our management network which is VLAN 2.  They added a vlan 2 ip address which was on the same subnet as the servers.  Deleted int vlan 2 and presto problem gone.  Made no sense to me but it worked.

Thanks for sending that redbook information.