Network load balancing in vmware environment problem

beke_josue
beke_josue used Ask the Experts™
on
Hi ,

I would like to tell you about a problem occured when we were trying to
configure a network load balancing (NLB) in vmware.

We were using the foundry switch like the picture I have attached to this mail.
(The switch is Brocade fastIron Edge GS648P and serial number is A02539F1E1
1-We had installed and configured NLB on windows server on virtual machines cluster.
2- We were using vmware ESX

After that, all the mac addresses of our vmware machines went to 00:00:00:00:00:00 .
We are trying to find the reasons but every thing seem to be well configured.


Did you have any recommandations for that issue?

thank you

Thank you
nlbpblem.jpg
Comment
Watch Question

Do more with

Expert Office
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
did the actual MAC address which appears in the VM settings, for the virtual network card change to zeros?

how did you setup NLB in Windows Unicast or Multicast?

Author

Commented:
The NLB was configured to multicast mode. Yes, the Mac address of the virtual machines change to zeros and every thing went down.
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
zeros in the actual VMX files?

or on the physical switch?

if everthing when down, how did you establish that the Auto Generated MAC addresses had changed in the VMX file registered on the servers?

did you have to reconfigure all the VMs?

Author

Commented:
After we made changes in the ESX infrastructure, nothing was working. And we saw in the arp table that all mac address of the virtual machines in where we had configured the nlb were zeros. We didn't check in the arp of the physical switch, there is a mistake in the picture I have attached.
After the problem we reconfigure all the VMs like they were before, that is say without nlb. But now we are looking for the reasons of that because we need to install the NLB.
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
okay, so it didn't actually change the VM MAC addresses? (in the VMX files) (physical virtual configuration)

all that happned was NLB multicast flooded the switch and reset the ARP cache.

when checking your physical switch, NLB Multicast had caused the ARP table to report all the machines as zeros?

I've never had this happen, but have you made any modfifications to to the switch to configure static arps for the NLB nodes and Cluster IP address?
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
here is a very long question, where I discussed NLB with another EE users, if it helps

http://www.experts-exchange.com/Networking/Windows_Networking/Q_27096829.html

Author

Commented:
I saw articles where people told me to configure static ARP entry in the physical switch. Is there any reason for that?
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
Yes, you must otherwise the nodes will not converge, if using mukticast on Cisco switches.

Author

Commented:
We use a foundry switch in multicast mode, we will see with the provider how to configure static ARP entry. But We saw in another article that we should make the changes to the physical layer 3 switch or router that acts as default gateway (note the default gateway from within Windows) for the NLB nodes.  In our sample the switch is layer 3 switch but it hasn't the role of defaut gateway.
Is it a must that the switch has to be the default gateway?
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
we usually create static arp entries in all switches on the network.

Author

Commented:
What I want to know is :
Do you create static ARP entry specifically in the default gateway ? Usually it is a router?

Do you create it only in the physical switches?
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
yes, in all network equipment, we provide a static arp entry on the port the mac is likely to appear, so all trunk ports.

yes, only on physical switches.

Author

Commented:
If you see my picture, there is a router. We are not the owner of the router, I mean all the changes at the router level were made by our service provider. I understand now that we must make changes in our switches.
My question is : Is it necessary to ask to our service provider to make changes at the router? (The router is the default gateway)
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
where is the access to the cluster, internal or external?
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
can if you do not have static arps, clustering will not be working for your clients, which you can test.

Author

Commented:
There is cluster between the host. Here we simplify the design in the picture. Now I known that static ARP is a must. And I sure that we have to add it in our switches. What I would like to know is if we need to ask our provider to make the same changes at the router level or not.
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
what is the service you are providing?

e.g. web facing, public server?

or internal usage on private lan
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
looking at your picture NO.

as long as the arp in the switch is correct, because traffic for the clusterip will come in through router and then hit switch.

Author

Commented:
Ok that is what I thought too and I wanted to be sure.  That's what I intend to try.
We will try it next week and I will tell you the results.

Thank you very much hanccocka.
Andrew Hancock (VMware vExpert / EE Fellow)VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017

Commented:
Test NLB correctt, by drain stopping each node. (and correctly flag each node, so you can identify which node is responding).

Author

Commented:
I have attached the picture to show you exactly what is the problem. We want to load balance our microsoft biztalk 2010. For any reason it still doesn't work.
vmproblem.JPG

Author

Commented:
I have attached the picture to show you exactly what is the problem. We want to load balance our microsoft biztalk 2010. For any reason it still doesn't work.
blade.vsd
VMware and Virtualization Consultant
Fellow 2018
Expert of the Year 2017
Commented:
what is the problem, I do no understand what Snapshots has to do with NLB?

Author

Commented:
the snapshot is another problem. I made a mistake. Forget it.

Author

Commented:
The problem was the switch configuration. It was important to add the arp entry mannually. Thank you hanccocka.

Do more with

Expert Office
Submit tech questions to Ask the Experts™ at any time to receive solutions, advice, and new ideas from leading industry professionals.

Start 7-Day Free Trial