Essentially I have two physical servers (Both identical, both ESXi hosts, each with 2xdual gbit nics) in a private rack in a data center. They will both be running a variety of db/web/storage server guests and each has a large redundant storage array for backing up the other's vm's for redundancy. Each machine has two uplinks (one from each NIC) to the datacenter's switch and the remaining two (again, one from each NIC) connecting them together directly, without a switch, for 3 reasons: data transport security between machines, independence from the datacenter network equipment and speed.

My question is this: How can I configure these hosts so that a guest web server on ESXi host 1 to connect to the DB server on ESXi host 2 over their private connection? Specifically what needs to be configured on the ESXi servers in order to add a new virtual NIC and assign a private IP address for internal communication with other guests.

If you need any additional detail I will be monitoring the thread closely to make updates.

Thank you!


On each ESXi, create a new virtual switch based on the direct connect NICs.

Using the 'Configuration' tab in vCenter on the ESXi, select 'Networking' and 'Add Networking...' Create new 'Virtual Machine' connections. Create a new virtual switch, and associate that switch with the proper NIC. Give the switch a meaningful name -- you'll need it later.

Now, in each VM that needs to talk to over the direct connection, create a new NIC (or modify an existing one), and set it's 'network label' the the switch you just created.

Assign the private IPs to those virtual NICs. Programs running in the VMs that use the private IPs you create will be routed across the direct connect.

  • 1
    This is perfect, thank you Jeff! I'm having some issues with vmware adding the private NIC to the virtual switch I created... Call "HostNetworkSystem.UpdateVirtualSwitch" object "networkSystem" on ESXi "x.x.x.x" failed. Should I start a new thread/question? – tg2 Oct 26 '10 at 20:14
  • NVM, fixed that problem... was a bloated log file (sel.raw -- sensor related) that was filling the vmware MAINSYS drive to 100% (32Mb) truncating it and restarting the host resolved the problem. – tg2 Oct 26 '10 at 20:26
  • Just verified that this works from within the VM's. I can't see where the "accept this answer" button is :\ – tg2 Oct 26 '10 at 20:50
  • Happy to help! "Accept answer" is the big checkmark under the answer's votes. – Jeff Leyser Oct 26 '10 at 20:57
  • Done and done :) – tg2 Oct 26 '10 at 23:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.