Hi
In 4.10.1. Creating a High Performance Virtual Machine, Template, or Pool
https://www.ovirt.org/documentation/virtual_machine_management_guide/inde...
I need some help trying to understand this section, please see my comments ">
inline".
<-- snip
Pinning topology:
The first two CPUs of each NUMA node are pinned.
OK, so the first 2 pCPUs of the NUMA node are used for I/O and
emulator threads.
If all vCPUs fit into one NUMA node of the host:
Yes, they do ...
The first two vCPUs are automatically reserved/pinned
Are the first 2 vCPUs automatically pinned to the I/O and Emulator
thread pCPUs mentioned in previous point?
The remaining vCPUs are available for manual vCPU pinning
<-- snip
So if I create a High Performance VM with 6 vCPUs, I should:
- Don't pin vCPUs 0 and 1
- Do pin vCPUs 2,3,4 and 5
All other High Performance VMs within the same NUMA node have their first 2 vCPUs
automatically pinned (presumably to the I/O and emulator thread pCPUs)
What do you think?
Thanks
Angus