Good Afternoon, We currently have VCL 2.5 deploying VMs on VMware ESXi. One of our images spins up a Linux VM that has HDP containerized in docker. We use "unload/power off after reservation" fo rhte predictive loading module so that VMs are shut down when reservations are complete. However, this does not seem to work if users only connect to the docker container and never connect to the native SSH port of the base Linux VM.
It appears that the issue is that the VCL does not detect that the reservation has ever been used, so it leaves it running in order to re-use it for the next reservation that is made. Yet, this is not possible since the SSH credentials have already been set for the previous user, and thus this stale VM cannot be reserved again successfully. Hopefully I am explaining that clearly, as it represents our best guess regarding what is actually occurring. The vlcd logs do not show any errors, and users can initiate their own workaround by terminating the "broken" reservation (which I believe properly destroys the VM) and reinitiating a new reservation to get a fresh image spun up. Unfortunately, ordinary users do not necessarily figure out this workaround. Has anyone experienced something similar and/or can anyone offer a configuration change to address this issue? Thanks in advance.
