We have a cluster of 5 dual socket cpu servers with each running some vms each. We have the following questions regarding the max. usage of cpu-cores that we can allocate.
E.g. if we look at one server that has two sockets, 32 cpu cores (16 real cores, and 16 hyperthreading cores) and two VMs using about the same amount of cpu.
Via Hardware - Processors, should we both give:
- Sockets: 2
- Cores: 32
- VCPUs 16
(we would like to use cpu hotplug, so thats why we use VCPUs)
Or should we keep some cores reserved for the proxmox hypervisor?
E.g. if we look at one server that has two sockets, 32 cpu cores (16 real cores, and 16 hyperthreading cores) and two VMs using about the same amount of cpu.
Via Hardware - Processors, should we both give:
- Sockets: 2
- Cores: 32
- VCPUs 16
(we would like to use cpu hotplug, so thats why we use VCPUs)
Or should we keep some cores reserved for the proxmox hypervisor?