One of my colleagues accidentally moved our master VM to an another vCenter (with XVCvMotion) and promptly after this all cloning duties on pools published from the image immediately failed. This quickly took down our pools with frequent logins and logoffs.
I've been in an assumption that master has no importance in cloning process after the template has created. AFAIK the process goes like
- master -> internal template ( with a dependent disk)
- internal template -> replica ( full clones, no navel cord)
- replica -> parent -> clone ( or replica -> clone depending of pool setup)
I had to check some logs to get a grip from this and indeed the loss of the master did kill the operations
2023-02-01T12:48:07.747+02:00 ERROR (0730-28B8) <PendingOperation-/DC1/vm/Horizon/win/vdi-win03-CloningNGVC> [NgvcHelper] Master vm vm-169012 not present in cache
2023-02-01T12:48:07.747+02:00 ERROR (0730-28B8) <PendingOperation-/DC1/vm/Horizon/win/vdi-win03-CloningNGVC> [CloneNgvcVmPO] Cloning of VM vdi-win03 has failed. null
at java.base/java.lang.Thread.run(Unknown Source)
I don't get it. What is the design justification here, what extra does the Ngvc process needs from the master after the publish has been done?