Description of problem:
In mixed-use clusters, administrator may want to reserve some resources for important VMs so that they might run anytime, regardless of level of use by less important VMs. The tightest bottleneck IMO is memory so this RFE is centered around this resource. The result should be that if the VM with such reserved memory configured is not running, engine should make sure that there is enough free memory on some host to run the VM (with the catch that if there are more such VMs in the cluster, the free memory on host must align with requirements of VMs - say if there are three VMs with 2 GB reserved each on 2-node cluster with 6 GB of free RAM, the free memory has to be 4 + 2 or 2 + 4 GB, not 3 + 3 GB).
Configuration-wise, there are more possible approaches how to configure this:
* reserve equivalent of Physical Memory Guaranteed automatically for HA VMs
* add a checkbox next to Physical Memory Guaranteed that would reserve the same amount when the VM is off
* add Reserved Memory next to Physical Memory Guaranteed
Version-Release number of selected component (if applicable):
Steps to Reproduce:
We are implementing HA Reservations in 3.4 under BZ 1027697. We want to implement the HA reservation policy, then look to extend to additional use cases such as reservations for Powered Off VMs.
It has been years since this was opened and it never spiked any interest. Closing for now. Please reopen with justification if it is important for you.