VMWare 4.1 with NX1000V and MS NLB?

[tiwang]

Hi out there

We have a setup based on a blade center with 8 blade servers running vmware esxi 4.1 - we are using ms win2k3 terminalsers which is load-balanced with NLB - as long as we keep the VM's on the same ESXi host the NLB works fine in unicast mode as long as we are using VMWare's native vswitch - but if we replace the VMWare switch with Cisco's NX 1000V we can only address one server and I can see that the clusers mac is found in the cam in one of the veth's.

we have also tried the multicast setup but it is not useable.

Has some out there got a running setup with MS NLB under VMWare & Cisco NX 1000?

best regards /ti

[tiwang]

Really no-one out there which have tried to run MS NLB under the VMWare switch?

[helenio]

I have the same problem with NLB configured on N1K and on standard virtual switch you avoid the problem disabling notify switch that avoid to send RARP packet at the ESX VM power up.

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1556

Said that I'm also looking the same feature/configuration to apply on N1K work. any idea ?

[tiwang]

Hi Helino

I haven't found any solution yet neither - I have removed the NX 1000V sw from those ESXi hosts where we are going to use NLB

best regards /ti

[helenio]

Finally to make it working I confiured the NLB in mul

ticast mode and activated the igmp snooping. The drawback is that you have to

configure arp resolution for the multicast MAC on the Router on the subnet.

[tiwang]

hi again

Are you running this under the Cisco 1000V or the VMWare switch?

When you are running in multicast mode are you then using SessionDirectory also ? - if so do you then also observe a short disconnect when logging where the Session Directory redirects you to another server ?

Best regards /ti

[helenio]

hi again

Are you running this under the Cisco 1000V or the VMWare switch? I'm running it under Nexus 1000V

When you are running in multicast mode are you then using SessionDirectory also ? Yes

if so do you then also observe a short disconnect when logging where the Session Directory redirects you to another server ? No

[tiwang]

hmm - strange - try to take a look here - wireshark dump - here in packet #1094 is the last packet seen from the cluster-address and the next packet is a reconnect on the primary lan-addresse - do you observe something similary? I cannot find some documentation on this behavior in multicast mode..

Are you running on a single adapter?

btw - we are running this on windows 2k3 servers

best regards /ti