dedicated NIC for HA

[nace_tech]

hello all, I have 2 esx4 update 2 hosts that have been in prod for about

3 years. recently i upgraded my license and i want to get involved in

vmotion and eventually HA. the only part that i'm conceptually stuck on

is the networking. i have 6 nics per host. 1 for SC, 2 internal

network, 3 external network. To meet the networking best practice i want

to dedicate one of the external nics to vmotion. I belive this nic

will then also be for the HA heartbeat or am i mistaken? The easy option would be to cross over connect the two

machines together but i plan on getting 2 more hosts shortly that i

would like to add to the vmotion and HA cluster.

assuming i create a

vlan on my main internal switch and put the vmotion nics on it, what do i do when that switch fails or I need to take it offline? If it's planned can i disable HA? if it's not planned and i come in to find my servers duplicated on another host, do i just shut one of them down after the switch is repaired and continue life as normal?

[AndreTheGiant]

One simple solution for your case is put vMotion and COS on the same vSwitch and have 2 NICs on it.

Then configure vMotion portgroup team policy to use one NIC as active and the second as passive, and COS portgroup to use NICs in a dual way.

To improve security you can use VLAN tagging at portgrup level to isolate this two networks.

Andre

[nace_tech]

AH, because they never need to travel all the way back to the main internal switch and everything can be done 'locally' in the v network? then that will address the issue of the physical switch going down because the heartbeats will be going over the virtual network and stay local to the v infrastructure?

if so, why ever create a vlan for vmware traffic and not just use v networking routes? are vlans or physical routing only needed when on different subnets?

[AndreTheGiant]

If you loose physical networking for HA, then hosts goes in isolation mode.

By default (on vSphere) the isolation response is to power-off the VMs (not very nice).

Be sure to have a reliable networking. Or change isoltation response. Or add other interface to do additional HA test.

Andre

[nace_tech]

thanks andre, in theory if our 1 main switch goes down (we have 3 switches total internal, external, n gateway) then our operations are pretty much on hold anyway so the vms going down is the least of the worries at the moment. it sounds like vsphere has precautions or atleast options to deal with the switch going down and preventing dual images being live perhaps by using the vcenter server to do additional testing to determine if things are royally in the doghouse besides lost communication between an HA pair? (just to help my understanding) thanks much for your help.

[AndreTheGiant]

You can add additional COS and use VMware HA advanced settings to add other "test point".

The parameters is "das.AllowNetwork"

http://kb.vmware.com/kb/1006421

Andre