Critical issue with NLB Cluster in ESX 4.1i

[santh]

Hi,

We are using 2 different NLB clsuter (2 Node cluster) in the Virtual machines. Each virtual machine having 2 different network card.

The first network card using Server IP address and second network card using NLB IP address.

Now the problem with NLB communication with 2 virtual machines. We are trying to ping continue in the NLB & Server IP addresses. But the server IP address is working fine. Only the issue NLB IP address is pinging and after some time giving request time out.

The Server IP address & NLB IP address running same VLAN. We have configured UNICAST NLB cluster. As per vmware document we have configured UNICAST NLB cluster in the vswitch & port group. Pls let me know any solution for the same.

Regards,

Santh

[Hoschi201110141]

Hi

I do not work with NLB.. but maybe the following is helping you:

If you like to use UNICAST it is neccesary to create a Portgroup or a vSwitch where you set "Notify Switches" to "No".

Connect only the NLB interfaces to this Portgroup / vSwitch.

I mean this is becaus NLB uses an diffrent MAC then the MAC on the NIC itself. When "Notify Switches" is enabled the MAC of the NIC itself wil be sent to the Switch... thats not what you need for NLB. NLB would send it's own MAC to the Switch => thats the Balancing-Mechanism.

See: http://www.vmware.com/files/pdf/implmenting_ms_network_load_balancing.pdf

Greetings

[santh]

Hi Hoschi,

This settings already done in the vswitch & port group. But still we are facoing same issue.

Regards,

Santh

[Hoschi201110141]

Hi Santh

I asked my Netadmin... He say that he configured UNICAST NLB (non-VM) on Cisco Switches.

He ment: If you have a inteligent switch like these one from Cisco they register that Traffic

goes to a Multicast Adress on a Unicast IP or something like this. He think it's a Problem in
the MS NLB Unicast Concept itself. It's not a realy god solution but he set static ARP Entries on the NLB Port's... it works.

If you have intelligent Routers / Switches maybe your Netadmin can help you.

good luck

[santh]

Hi Hoschi,

Yesterday we have analyed one thing for the NLB IP Address pinging result (NLB IP Address sigment for the vlan 3). Once the NLB IP addresses coming request time out, that time we tried to ping same vlan 3 different ESX host virtual machine perfectly pinging the NLB IP addresses. But other vlan's virtual machine not pinging nlb IP address.

Now we are confusing where is the problem for the network side or vmware side. Pls let me know you have any idea to solve this issue.

Regards,

Santh

[Hoschi201110141]

Hi Santh

Im runnig out of knowledge about NLB. But what i have understand you should check the ARP Tables

and maybe you need to make some static ARP entrys. This sems not to be a vmware issue, see the
Link above:

http://www.cisco.com/en/US/products/hw/switches/ps708/products_configuration_example09186a0080a07203...

There are interresting sentences like: "However, since the incoming packets have a unicast destination IP       address and multicast destination MAC the Cisco device ignores this entry and       process-switches each cluster-bound packets."

Or: "For Cisco Catalyst 6000/6500 series switches, you must add the           disable-snopping parameter."

If you not running Cisco, try to finde a similar Document by the Vendor of your choice.

Sorry that i can not do more for you

Hoschi