How to connect a VM to a specific processor core through Numa?

[sbin]

Hi,

When starting VM error:

qm start 100

kvm: -numa node,nodeid=1,cpus=1,memdev=ram-node1: CPU index (1) should be smaller than maxcpus (1)
start failed: QEMU exited with code 1

How can this problem be solved? How to assign the necessary cores to VM?

configuration VM:

balloon: 0
bootdisk: scsi0
cores: 1
ide2: none,media=cdrom
memory: 1024
name: eeee1
numa: 1
numa1: cpus=1,hostnodes=1,memory=1024,policy=bind
ostype: l26
scsi0: VMS:vm-100-disk-0,size=9G
scsihw: virtio-scsi-pci
smbios1: uuid=ceb5befb-d88b-4760-aadc-d3d817e99d5b
sockets: 1
vcpus: 1
vmgenid: b38eff65-ea85-4e45-a87c-be5af2f46a55

Thank you.

 

[t.lamprecht]

nodeid=1

try 0 here, those IDs start with zero, not one.

 

[sbin]

nodeid - there is no such parameter in the documentation.

 

[sbin]

For example, if I run the VM with the settings:

Spoiler: Conf_VM

balloon: 0
bootdisk: scsi0
cores: 1
ide2: none,media=cdrom
memory: 1024
name: eeee1
numa: 1
numa0: nodeid=0,cpus=2,hostnodes=0,memory=1024,policy=bind
ostype: l26
scsi0: VMS:vm-100-disk-0,size=9G
scsihw: virtio-scsi-pci
smbios1: uuid=ceb5befb-d88b-4760-aadc-d3d817e99d5b
sockets: 1
vcpus: 1
vmgenid: b38eff65-ea85-4e45-a87c-be5af2f46a55

Before starting the VM:

root@pv1:~# numactl -H
available: 2 nodes (0-1)
node 0 cpus: 0 2 4 6 8 10 12 14
node 0 size: 12017 MB
node 0 free: 11078 MB
node 1 cpus: 1 3 5 7 9 11 13 15
node 1 size: 12067 MB
node 1 free: 10613 MB
node distances:
node   0   1
  0:  10  20
  1:  20  1

###

root@pv1:~# qm start 100
vm 100 - unable to parse value of 'numa0' - format error
nodeid: property is not defined in schema and the schema does not allow additional properties

after starting the VM:

Spoiler: start_VM

root@pv1:~# numactl -H
available: 2 nodes (0-1)
node 0 cpus: 0 2 4 6 8 10 12 14
node 0 size: 12017 MB
node 0 free: 11039 MB
node 1 cpus: 1 3 5 7 9 11 13 15
node 1 size: 12067 MB
node 1 free: 10266 MB
node distances:
node 0 1
0: 10 20
1: 20 10

root@pv1:~# qm list
vm 100 - unable to parse value of 'numa0' - format error
nodeid: property is not defined in schema and the schema does not allow additional properties
VMID NAME STATUS MEM(MB) BOOTDISK(GB) PID
100 eeee1 running 1024 9.00 9421
root@pv1:~#
root@pv1:~#
root@pv1:~# numastat -c 9421

Per-node process memory usage (in MBs) for PID 9421 (kvm)
Node 0 Node 1 Total
------ ------ -----
Huge 0 0 0
Heap 0 0 0
Stack 0 1 1
Private 15 303 318
------- ------ ------ -----
Total 15 304 319
root@pv1:~#

running VM state:

root@pv1:~# ps aux | grep 9421
.....
-smp 1,sockets=1,cores=1,maxcpus=1 -nodefaults -boot menu=on,....
-cpu kvm64,+lahf_lm,+sep,+kvm_pv_unhalt,+kvm_pv_eoi,enforce -m 1024
-object memory-backend-ram,id=ram-node0,size=1024M -numa node,nodeid=0,cpus=0,memdev=ram-node0
....

The VM was attached to NODE0 and CPU # 2, but the VM uses RAM from NODE1.
From this we can conclude that Numa does not work correctly or does not work at all.

If I made the settings incorrectly, then tell me how to fix the CPU and memory to the desired node?

 

[sbin]

If I specify any cpus in the VM configuration, then proxmox in any case uses only cpus0
This can be seen from the post above.

 

[thomasgug]

Hello
We are having the same issue as sbin
our numa configuration line is :

numa0: cpus=0;1;2;3;4;5;6;7;8;9;10;11;12;13;14;15;16;17;18;19;40;41;42;43;44;45;46;47;48;49;50;51;52;53;54;55;56;57;58;59,hostnodes=0,memory=40960,policy=bind

which is what is described in the documentation.

We are also getting the
vm 100 - unable to parse value of 'numa0' - format error


Thanks

 

[dcsapak]

please post the complete vm config (qm config ID) and the error on start (qm start ID)

 

[thomasgug]

Hello

Configuration:

bootdisk: scsi0
cores: 40
cpu: host
ide2: none,media=cdrom
memory: 40960
name: osl-pi1-wrk20a
net0: virtio=C2:8B:A3:29:7B:A8,bridge=vmbr10,firewall=1
numa: 1
numa0: cpus=20;21;22;23;24;25;26;27;28;29;30;31;32;33;34;35;36;37;38;39;60;61;62;63;64;65;66;67;68;69;70;71;72;73;74;75;76;77;78;79,hostnodes=1,memory=40960,policy=bind
ostype: l26
scsi0: r1-pool:vm-100-disk-0,backup=0,size=50000M
scsihw: virtio-scsi-pci
smbios1: uuid=553a5e55-b812-4723-af72-ee77902932dd
sockets: 1
unused0: local:100/vm-100-disk-0.qcow2
vmgenid: aff0af2a-bab3-409e-9a0b-c4ecf16c27dc

Error:

Feb 28 09:31:49 osl-host36 qmeventd[2317]: vm 101 - unable to parse value of 'numa0' - format error

Doc:

numa[n]: cpus= [,hostnodes=] [,memory=] [,policy=] NUMA topology.
cpus= CPUs accessing this NUMA node.
hostnodes= Host NUMA nodes to use.
memory= Amount of memory this NUMA node provides.
policy= NUMA allocation policy.

 

[MisterY]

I tried to allocate cores to a VM (16 of 24) as well. The VM starts and everything is running. BUT if I run a stress test on the VM for 16 cores then it will utilize all 24 cores of the host to ~70 %. Even if I stress only one core the thread is spread over all 24 cores on Proxmox.

cores: 8
cpu: host
sockets: 2
memory: 16384
numa: 1
numa0: cpus=0;2;4;6;8;10;12,hostnodes=0,memory=8192,policy=bind
numa1: cpus=1;3;5;7;9;11;13;15,hostnodes=1,memory=8192,policy=bind

numactl -H
available: 2 nodes (0-1)
node 0 cpus: 0 2 4 6 8 10 12 14 16 18 20 22
node 0 size: 72548 MB
node 0 free: 47012 MB
node 1 cpus: 1 3 5 7 9 11 13 15 17 19 21 23
node 1 size: 72553 MB
node 1 free: 43520 MB
node distances:
node   0   1
  0:  10  20
  1:  20  10