root/local_proxmox
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Document NUMA socket pinning for VM 100 (socket 0) and VM 101 (socket 1)

Browse Source 
Add cpuset instructions so each VM is actually pinned to its physical CPU
rather than relying on Proxmox to guess. VM 100 → CPU 1 (socket 0),
VM 101 → CPU 2 (socket 1).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CTRZQmUSrYXFC6WkcypYPv
This commit is contained in:
Claude
2026-06-28 01:05:26 +00:00
parent 132c55e4e1
commit d01fe4717d
1 changed files with 27 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+27 -4
View File
@@ -2236,14 +2236,26 @@ Click **Next**.
> **Why 12 cores / 1 socket?** The E5-2660 v4 has 14 cores per socket (28
> total across both CPUs). Keeping Sockets at 1 pins the VM to a single NUMA
> node — all memory accesses stay local to that socket's RAM channels, avoiding
> the ~30–40ns penalty of crossing to the other socket. Give VM 100 12 cores
> from socket 0, VM 101 gets 6 cores from socket 1.
> the ~30–40ns penalty of crossing to the other socket. VM 100 gets socket 0
> (CPU 1), VM 101 gets socket 1 (CPU 2).
>
> **Enable NUMA** tells Proxmox to allocate RAM from the same socket the vCPUs
> are pinned to. Without it the RAM may come from the wrong socket even if
> Sockets is set to 1. The checkbox is in **Processors → Edit → Enable NUMA**
> (not in the Memory dialog).
>
> **Pinning VM 100 to socket 0 (CPU 1) on the Proxmox host:**
> ```bash
> # Find which CPU IDs belong to socket 0
> lscpu -e | grep "^[0-9]" | awk '$3==0 {print $1}' | tr '\n' ','
> # Example output: 0,1,2,...,13,28,29,...,41,
>
> # Pin VM 100 to socket 0 cores (replace the CPU list with your actual output above, no trailing comma)
> echo "cpuset: 0-13,28-41" >> /etc/pve/qemu-server/100.conf
> ```
> After adding `cpuset`, restart the VM. Proxmox will then schedule all vCPUs
> only on those physical cores, keeping memory traffic on socket 0's RAM channels.
>
> **nested-virt flag** — you will see this in Extra CPU Flags. Leave it at
> Default; it is automatically enabled when CPU type is `host` on Intel, which
> is what allows VirtualBox/WinApps to run inside this VM.
@@ -2616,8 +2628,19 @@ Click **Create VM** in the web UI.
| Type | host |
| Enable NUMA | checked |
> 6 cores from socket 1 (VM 100 uses socket 0). NUMA keeps memory local to
> the cores doing the work.
> 6 cores from socket 1 (CPU 2) — VM 100 uses socket 0 (CPU 1). NUMA keeps
> memory local to the cores doing the work.
>
> **Pin VM 101 to socket 1 (CPU 2) on the Proxmox host:**
> ```bash
> # Find which CPU IDs belong to socket 1
> lscpu -e | grep "^[0-9]" | awk '$3==1 {print $1}' | tr '\n' ','
> # Example output: 14,15,...,27,42,43,...,55,
>
> # Pin VM 101 to socket 1 cores (replace list with your actual output, no trailing comma)
> echo "cpuset: 14-27,42-55" >> /etc/pve/qemu-server/101.conf
> ```
> After adding `cpuset`, restart VM 101.
**Tab: Memory**