Is there an easier way to cold migrate a VM from vDS to vDS without removing the vNIC?
My current environment has 2 physically separated datacenters with about 15 hosts in each. vCenter is configured with each physical datacenter having a single 15-host Datacenter/Cluster and vDS defined for each.
The current process prior to cold migration is to remove the vNIC(s), then migrate the VM to the desired location. Is there a better or easier way of doing this? Would labeling my dvPgs exactly the same resolve this issue?
For such scenarios we have created the following workaround in our environment:
Create an extra vDS (we named it TransferSwitch), add all hosts from all datacenters/clusters to it, but without any physical uplink.
A cold migration process from one datacenter/cluster to another is then done in three steps:
1. Connect the VM's NIC to the TransferSwitch
2. Cold migrate the VM to the destination
3. Connect the VM's NIC to the destination vDS
This avoids removing/re-adding the VM's NICs and all consequences (like a new device being detected in the Guest OS, with a different MAC address).
You may create the TransferSwitch just temporarily for each VM migration, but we decided to just keep it permanently.
- Andreas
For such scenarios we have created the following workaround in our environment:
Create an extra vDS (we named it TransferSwitch), add all hosts from all datacenters/clusters to it, but without any physical uplink.
A cold migration process from one datacenter/cluster to another is then done in three steps:
1. Connect the VM's NIC to the TransferSwitch
2. Cold migrate the VM to the destination
3. Connect the VM's NIC to the destination vDS
This avoids removing/re-adding the VM's NICs and all consequences (like a new device being detected in the Guest OS, with a different MAC address).
You may create the TransferSwitch just temporarily for each VM migration, but we decided to just keep it permanently.
- Andreas
Andreas,
Thanks for the response. I'm a bit confused here. Bare in mind that I'm still learning about dvSwitches and their intricacies. With that, isn't it true that a dvSwitch needs at least one uplink for each host for communication with one another? If I create a dvSwitch with no uplinks wouldn't vcenter give a warning requesting the uplink?
No, you won't get a warning. It's perfectly okay to create a vDS without any physical uplinks (just as it is okay to create a standard vSwitch without uplinks).
Of course, a VM attached to this vDS won't be able to communicate over the network. However, you only attach a VM to that vDS for moving it between clusters while it is powered off.
Andreas Peetz wrote:
...just as it is okay to create a standard vSwitch without uplinks