simultaneously running of vms crashing the hole pve host system

[rudi1234]

Hello there

i'm having a wired problem regarding the combination of two specific vms.
Here is the setup:

vm_1 (libreelec):

bios: ovmf
boot: order=ide0;ide2
cores: 1
efidisk0: machines:vm-201-disk-1,size=1M
hostpci0: 0000:03:00,pcie=1,x-vga=1
hostpci1: 0000:00:1f.3
ide0: machines:vm-201-disk-0,size=7717M
ide2: none,media=cdrom
machine: q35
memory: 2048
name: LibreELEC
net0: virtio=06:69:0C:AD:FD:C7,bridge=vmbr0,firewall=1
onboot: 1
parent: working
smbios1: uuid=5f62d1f0-e01f-4efa-b5a3-7acc5d78c36b
usb0: host=1-5.4
usb1: host=1-5.1
vmgenid: ee647be4-bcde-4d3f-9d95-c5fd405fce4a

vm_2 (batocera):

args: -set drive.drive-scsi0.readonly=on -drive 'file=/dev/zvol/machines/vm-202-disk-0,if=none,id=drive-usb0,format=raw,cache=none,aio=io_uring,detect-zeroes=on' -device 'usb-storage,drive=drive-usb0,bootindex=1,removable=on'
bios: ovmf
boot: order=ide2;net0
cores: 8
cpu: host
efidisk0: local-lvm:vm-202-disk-0,size=4M
hostpci0: 0000:03:00,pcie=1,x-vga=1
ide2: none,media=cdrom
machine: q35
memory: 12048
name: Batocera
net0: virtio=7A:6D:FE:FE:72:EF,bridge=vmbr0,firewall=1
numa: 0
ostype: l26
parent: slow_down_testing
protection: 1
scsi0: machines:vm-202-disk-0,size=100G
smbios1: uuid=4ceb3776-919b-4b51-8e5e-b2d409da4c16
sockets: 1
usb0: host=1-5.3
usb1: host=1-5.4
vmgenid: 645f0a9a-81c3-4159-ac9f-376a557d85d3

vm_3 (RaspAp):

bios: ovmf
boot: order=scsi0;ide2;net0
cores: 1
efidisk0: machines:vm-101-disk-0,efitype=4m,pre-enrolled-keys=1,size=1M
hostpci0: 0000:06:00 # Intel 802.11ac WiFi + BT 4.2 pci adapter
ide2: local:iso/debian-10.10.0-amd64-netinst.iso,media=cdrom,size=336M
machine: q35
memory: 2048
meta: creation-qemu=6.1.0,ctime=1638785039
name: RaspAP
net0: virtio=92:AE:5B:60:FB:10,bridge=vmbr0,firewall=1
numa: 0
ostype: l26
scsi0: machines:vm-101-disk-1,size=6G
scsihw: virtio-scsi-pci
smbios1: uuid=52b1b4e8-642c-45d7-90af-4501c0c201ef
sockets: 1
vga: qxl,memory=128
vmgenid: 33e15c78-9ae0-4a48-93e4-48dd5ef381e0

When I run vm_1 and vm_3 simultaneously everything is fine. I can watch videos via libreelec on my tv with gpu passthrough and be able to use the wifi provided by vm_3.
But when I run vm_2 and vm_3 together the hole pve host system crashes. When running vm_2 alone everything works and I can play some retro games on my tv via gpu passthrough.
I also cannot find anything special in the syslogs when running vm_2 and vm_3 in parallel.
My motherboard is an H570M-ITXac

I also tried to change the disk type from scsi to ide with no difference.

The fact that this setup works with vm_1 and vm_3 but not with vm_2 and vm_3 confuses me and I hope to find some help here.

Regards Rudi

 

[Neobin]

You can't passthrough the same device to different VMs at the same time.

 

[rudi1234]

I'm not passthrough the same device to different VMs at the same time (i guess you mean hostpci0: 0000:03:00,pcie=1,x-vga=1). I'm not running vm_1 and vm_2 at the same time. it's about vm_1 and vm_3 OR vm_2 and vm_3.

 

[Neobin]

Misread/-understood that, sorry.

 

[dietmar]

Maybe you run out of memory?

 

[rudi1234]

Yes, I ran out of memory. Thank you very much. Reducing the memory of the Batocera instance solves it completely.