I'm running a supermicro CSE-829U-10 + X11DPU dual Xeon 4110 system, which itself is very stable. Running as a node in the cluster with several weeks of uptime with no issues. I tried to create a VM, and it runs for between 10 seconds and 2 minutes, then the green arrow symbol changes to a yellow exclamation mark and gives qm status: internal-error. I've tried with both mintOS and ubuntu 24.04 iso's and have the same issue. The system has 4x 32gb DIMMs, and I've tried running with 1 DIMM per CPU, testing, then swapping both DIMMs and testing again, same issues. So if it's a memory issue multiple DIMMs would have to be bad which seems unlikely.
I've tried with dual CPUs + NUMA, and single CPU with no NUMA, same issue.
I'm using the same method/configuration I use for setting up VM's on my DL360 gen 10 (which is the same hardware but HPE branded), and I've never had any issues with VM's on that machine. The only difference is that system is installed on a sata SSD whereas this system is using a PCIe to 2x m.2 nvme adapter for two nvme SSD's in ZFS raid 1 config for the proxmox install.
I don't know if this is obsolete, but I can't find where any relevant logs would be? /var/log/syslog doesn't exist.
Summary:
System info:
VM configuration
Error codes:
I've tried with dual CPUs + NUMA, and single CPU with no NUMA, same issue.
I'm using the same method/configuration I use for setting up VM's on my DL360 gen 10 (which is the same hardware but HPE branded), and I've never had any issues with VM's on that machine. The only difference is that system is installed on a sata SSD whereas this system is using a PCIe to 2x m.2 nvme adapter for two nvme SSD's in ZFS raid 1 config for the proxmox install.
I don't know if this is obsolete, but I can't find where any relevant logs would be? /var/log/syslog doesn't exist.
Summary:
- VM's launch but stop with status: internal-error after 10 seconds to 2 minutes.
- Rotated DIMMs to test for memory issues so this seems unlikely.
- IOMMU enabled and passthrough configured.
- Host configured on 2xNVME drives in ZFS raid 1.
- Host shows no ECC or MCE errors.
- Secure Boot toggling had no effect.
- Tried with SR-IOV disabled and enabled.
- Tried with CPU in host and x86-64-v2.
- Tried with dual and single CPU, NUMA on and off.
System info:
Code:
proxmox-ve: 9.1.0
pve-manager: 9.1.4
pve-kernel: 6.14.11-4-pve
pve-qemu-kvm: 10.1.2-5
CPU: Dual Intel Xeon Silver 4110
RAM: 128GB ECC RDIMM
Code:
bios: ovmf
machine: q35
cpu: host
memory: 16384
sockets: 1
cores: 5
efidisk0: local:101/vm-101-disk-0.qcow2,efitype=4m,ms-cert=2023,pre-enrolled-keys=1
scsi0: /dev/disk/by-id/nvme-INTEL_SSDPE2KE016T8...,size=1.4TB
boot: order=ide2
ide2: local:iso/linuxmint-22.3-cinnamon-64bit.iso,media=cdrom
net0: virtio, bridge=vmbr0
Code:
QEMU[7591]: KVM: entry failed, hardware error 0x80000021
QEMU[7591]: If you're running a guest on an Intel machine without unrestricted mode support...
...
kvm_intel: VMCS ..., last attempted VM-entry on CPU 29
kvm_intel: *** Guest State ***
...
kvm_intel: VMExit: reason=80000021 qualification=0000000000000004