VMs crash randomly

[dgrivelet]

Hello all,

Some of my VMs (2022 and 2019) crashes randomly.

Logs :

"Aug 06 04:27:38 proxmox QEMU[2096]: KVM: entry failed, hardware error 0x80000021
Aug 06 04:27:38 proxmox QEMU[2096]: If you're running a guest on an Intel machine without unrestricted mode
Aug 06 04:27:38 proxmox QEMU[2096]: support, the failure can be most likely due to the guest entering an invalid
Aug 06 04:27:38 proxmox QEMU[2096]: state for Intel VT. For example, the guest maybe running in big real mode
Aug 06 04:27:38 proxmox QEMU[2096]: which is not supported on less recent Intel processors.
Aug 06 04:27:38 proxmox QEMU[2096]: EAX=00011d2a EBX=d19e2180 ECX=00000001 EDX=00000000
Aug 06 04:27:38 proxmox QEMU[2096]: ESI=1c0d6040 EDI=d19ee440 EBP=00000000 ESP=a2237d40
Aug 06 04:27:38 proxmox QEMU[2096]: EIP=00008000 EFL=00000002 [-------] CPL=0 II=0 A20=1 SMM=1 HLT=0
Aug 06 04:27:38 proxmox QEMU[2096]: ES =0000 00000000 ffffffff 00809300
Aug 06 04:27:38 proxmox QEMU[2096]: CS =aa00 7ffaa000 ffffffff 00809300
Aug 06 04:27:38 proxmox QEMU[2096]: SS =0000 00000000 ffffffff 00809300
Aug 06 04:27:38 proxmox QEMU[2096]: DS =0000 00000000 ffffffff 00809300
Aug 06 04:27:38 proxmox QEMU[2096]: FS =0000 00000000 ffffffff 00809300
Aug 06 04:27:38 proxmox QEMU[2096]: GS =0000 00000000 ffffffff 00809300
Aug 06 04:27:38 proxmox QEMU[2096]: LDT=0000 00000000 000fffff 00000000
Aug 06 04:27:38 proxmox QEMU[2096]: TR =0040 d19f2000 00000067 00008b00
Aug 06 04:27:38 proxmox QEMU[2096]: GDT= d19f3fb0 00000057
Aug 06 04:27:38 proxmox QEMU[2096]: IDT= 00000000 00000000
Aug 06 04:27:38 proxmox QEMU[2096]: CR0=00050032 CR2=83e306e0 CR3=001ae000 CR4=00000000
Aug 06 04:27:38 proxmox QEMU[2096]: DR0=0000000000000000 DR1=0000000000000000 DR2=0000000000000000 DR3=0000000000000000
Aug 06 04:27:38 proxmox QEMU[2096]: DR6=00000000ffff0ff0 DR7=0000000000000400
Aug 06 04:27:38 proxmox QEMU[2096]: EFER=0000000000000000
Aug 06 04:27:38 proxmox QEMU[2096]: Code=kvm: ../hw/core/cpu-sysemu.c:77: cpu_asidx_from_attrs: Assertion `ret < cpu->num_ases && ret >= 0' failed.
Aug 06 04:27:38 proxmox kernel: set kvm_intel.dump_invalid_vmcs=1 to dump internal KVM state.
Aug 06 04:27:38 proxmox kernel: vmbr1: port 5(tap104i0) entered disabled state
Aug 06 04:27:38 proxmox kernel: vmbr1: port 5(tap104i0) entered disabled state
Aug 06 04:27:38 proxmox systemd[1]: 104.scope: Succeeded.
Aug 06 04:27:38 proxmox systemd[1]: 104.scope: Consumed 21h 34min 28.453s CPU time.
Aug 06 04:27:40 proxmox qmeventd[1659116]: Starting cleanup for 104
Aug 06 04:27:40 proxmox qmeventd[1659116]: Finished cleanup for 104"

Any ideas ?

Thanks in advance for your replies

Daniel

 

[leesteken]

This post suggest updating Proxmox to at least kernel 5.15.39-3. Have you tried that?

 

[Neobin]

Aug 06 04:27:38 proxmox QEMU[2096]: KVM: entry failed, hardware error 0x80000021

Seems to be fixed with the 5.15.39-3 PVE-kernel [1]. But not sure, if it is already available in the enterprise repository.

Edit: Too slow.

[1] https://forum.proxmox.com/threads/v...e-error-0x80000021.109410/page-13#post-487275

 

[dgrivelet]

Thanks for your replies. This kernel version is only available in the test repository. I'm reluctant because of side effects.

 

[leesteken]

Thanks for your replies. This kernel version is only available in the test repository. I'm reluctant because of side effects.

It's also in the no-subscription repository but that does not help you, I guess. This post, in that same thread, suggest a work-around.

 

[dgrivelet]

It's also in the no-subscription repository but that does not help you, I guess. This post, in that same thread, suggest a work-around.

I tested first to update the intel microcode as described below :

https://wiki.debian.org/Microcode#Debian_11_.22Bullseye.22_.28stable.29

I'll tell you if it does the trick

Thanks !

 

[dgrivelet]

Bad news, the crashes continue, but less than before. Tested the kvm.tdp_mmu=N trick and wait.