relationship between connected cdrom device and ability to use HA in vSphere

[lethargos]

Hello,

We are running vSphere 6.7.0.54000 (rather old, I know) where I'm provisioning virtual machines starting from ova templates using terraform. In order to do that, I need to create virtual machines with CDROMs attached. As far as I could tell, there is no way I can automatically disconnect the CDROM after provisioning the virtual machine. And the CDROMs remain connected to the virtual machine.

In any case, our sysadmin suddenly tells me that in order for those virtual machines to profit high availiability, the CDROMs need to be disconnected, otherwise they won't be moved to another host in case of failure and he makes references to these 'remediation pre-check' tasks done on the cluster scope without pointing exactly to the error (I've gone through the logs and all these tasks completed successfully).

It is true that, if you try to disconnect the CDROM, the virtual machine get blocked/frozen and you need to explicitly answer whether you want to force eject the CDROM or not, as the virtual machine might still be using that information and vSphere doesn't know about it. But I am able to live migrate the virtual machine without any issues from one host to another. This argument hasn't really convinced him, because he says (and he is right in the strict sense of the word) that it's different when a failure occurs and in that case the virtual machine won't be moved to another host because that "nasty" CDROM device is connected to the virtual machine. The image file is on a shared datastore to which all hosts have access.

The CDROM uses emulate device mode (so it's not passthrough), which intrigues all the more. I'm attaching the relevant screenshots of the CDROM settings.

Any help is much appreciated.