1

Can anyone help me understand how networking works for any live migration of VMs.

Consider two VMs with private addresses 192.168.5.100, 192.168.5.101 with gateway 192.168.5.1. They also might have public IP addresses that they might be hosting.

I have the following questions:

  1. If these VMs were to be live migrated to another host, Does the destination need to have the same subnet configured?

  2. Usually is the L2 network stretched so that if one VM is live migrated, it can still talk to the other. If yes, how is it typically done?

  3. What happens to the public IP addresses? How do clients know about the new location.

I also came across ESX supporting live migration over WAN. In that case how does the connectivity between VMs work?

Thanks in advance!

Undivided
  • 13
  • 2

1 Answers1

0

The VM IP addressing doesn’t make any difference at all. The VMs don’t migrate themselves. The host does.

It is assumed that the hosts can communicate with each other to do the migration. It is also assumed that once the migration is done the new host will make the same networks available to the VM.

Once the VM migrates it continues to operate with the same IP addresses it had. If those same networks aren’t connected to the host, it won’t work.

For that reason, migrating VMs across different networks is an advanced procedure that also assumes you are going to change the IP addresses after migration or you have some type of routing protocols in place that make the original networks reachable at the new physical location.

Appleoddity
  • 3,488
  • 2
  • 13
  • 33
  • Thank you for the answer! Just curious, What kind of routing protocols are setup usually so that the migrated VM can reach the private network on the original source? I am just trying to understand what an admin would typically do in case such migrations are needed. – Undivided Jun 13 '18 at 05:20
  • 1
    It’s not that I don’t want to answer. It’s just too complicated to explain in a brief post. Migrating to a different physical site is usually going to be done for disaster recovery purposes. Which means the site has already been prepped and has been connected in to the company WAN using various routing protocols like BGP to make network segments available in different physical locations. It’s not something you can do across two geographically different and isolated sites. – Appleoddity Jun 13 '18 at 05:27