We have an issue with our environment based on vCenter 6.5, NSX and VRA 7.4.
When we deploying a windows based blueprint it is possible that somewhere in the process the network adapter won't be connected to a NSX logical switch. The machine vmx file shows then the following line "ethernet0.startConnected = "FALSE"" when a machine is hit with the issue. The blueprint deploys without any visible error message.
I'm sure that the line is not in our template machine. We are deploying blueprints based on linked clones and a customization spec with sysprep enabled. We could not reproduce it on Linux blueprint.
We have tried the a golden image with a E1000 or VMXNET adapter but both adapters have sometimes the issue
Does anyone know the problem? Or could provide some steps to identify which component in our infrastructure causes the problem/
when a machine is hit with the issue.
When you say this, are you implying that it is never connected to start with or that it is connected to start but at some point thereafter it becomes disconnected? If you have the same blueprint but deploys to a VLAN-backed port group instead, does this behavior still occur?
I guess that it has never been connected to the logical switch, but I'm not sure. I mean with that sentence that is virtually random when a deployed machine has the issue. I could deploy 6 instances of the same machine and only 1 machine could be malfunctioning of the 6.
I have not tested it yet with a port group, can be a good way to exclude any (faulty) NSX configuration.
I have seen this many times. over the years with vCAC 6 right through to vRA 7.x and have found this happens with the machine vMotions and moves hosts during the configure stage.
Thanks for the possible solutions.
We ran some tests to rule out any nsx components and to narrow down to issue.
Our usecase at this moment is as follows. We want to deploy any deployment without network and internet access. To prevent that a machine could compromise the security of our infrastructure.
The owner of the deployment need to select a nsx virtual wire after he secured the machine.
We are trying to achieve this goal with a port group connected to a virtual distributed switch and a blocked ports security policy. We have included in the blueprints a custom property (VirtualMachine.Network0.Name) to deploy the machine in this network.
We guess this setup could be the root cause of our problems but right now we don't have a other way to provide this use case.
This just seems like a highly strange use case to me anyway and I really question why you're even doing it to begin with. I've just never heard of anyone doing this. It kind of defeats the purpose of automation to begin with. If you NOT do what you're doing and deploy to a normal VLAN-backed portgroup do you still have the issue? You say it is Windows templates only?