VM random shutdown

[Laimonas Vijūnas]

I have problem with one of the proxmox servers as it randomly stops VM. It all started few days after update to 7.4-15. First of all i thought maybe VM needs more RAM ( common situation) but it didn't help. After update to 7.14-16 situation is same. While digging through logs i found that it's always same error:

Jul 23 12:08:40 nev-intel QEMU[56726]: kvm: ../target/i386/tcg/emit.c.inc:1504: gen_pinsr: Assertion `vec_len == 16' failed.
Jul 23 12:08:41 nev-intel kernel: vmbr0: port 2(tap101i0) entered disabled state
Jul 23 12:08:41 nev-intel kernel: vmbr0: port 2(tap101i0) entered disabled state
Jul 23 12:08:41 nev-intel kernel: zd32: p1 p2
Jul 23 12:08:41 nev-intel systemd[1]: 101.scope: Succeeded.
Jul 23 12:08:41 nev-intel kernel: zd16: p1
Jul 23 12:08:41 nev-intel systemd[1]: 101.scope: Consumed 10h 5min 46.183s CPU time.
Jul 23 12:08:41 nev-intel qmeventd[461678]: Starting cleanup for 101
Jul 23 12:08:41 nev-intel qmeventd[461678]: Finished cleanup for 101
Jul 23 12:17:01 nev-intel CRON[469088]: pam_unix(cron:session): session ope


Jul 17 04:43:14 nev-intel QEMU[1841671]: kvm: ../target/i386/tcg/emit.c.inc:1504: gen_pinsr: Assertion `vec_len == 16' failed.
Jul 17 04:43:15 nev-intel kernel: vmbr0: port 2(tap101i0) entered disabled state
Jul 17 04:43:15 nev-intel kernel: vmbr0: port 2(tap101i0) entered disabled state
Jul 17 04:43:15 nev-intel systemd[1]: 101.scope: Succeeded.
Jul 17 04:43:15 nev-intel systemd[1]: 101.scope: Consumed 1w 1d 21h 16min 24.442s CPU time.
Jul 17 04:43:15 nev-intel kernel: zd0: p1
Jul 17 04:43:15 nev-intel qmeventd[3932776]: Starting cleanup for 101
Jul 17 04:43:15 nev-intel qmeventd[3932776]: Finished cleanup for 101


There is only one VM on this server.

 

[Philipp Hufnagl]

can I see the complete log with journalctl -b -1 ?

 

[fabian]

also please include "pveversion -v" output and the configuration of the VM in question, as well as any details that could make it "special"

 

[Laimonas Vijūnas]

Added logs as attachments. About configuration:

Nothing special

 

[fabian]

please actually post the config file (or "qm config XXX" output)!

 

[Laimonas Vijūnas]

please actually post the config file (or "qm config XXX" output)!

Added config file

 

[fabian]

is there a reason why you disabled hardware virtualization ("kvm: 0") and use the "max" cpu type? both is definitely not recommended - see https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_cpu_type for an explanation of possible values..

 

[Laimonas Vijūnas]

HW server is Intel R1000WF and we had problems with starting up migrated VM from other server. This configuration was stable for more than a year.

 

[fabian]

yeah, I think it is a bug in Qemu, but the combination of

- software virtualization (which means severely reduced performance, not something you should ever use in production!)
- CPU type max (which means "enable all CPU features supported", instead of a sensible set)
- Windows as guest OS

is not a good idea irrespective of the crash.

please enable KVM and change the CPU type to something suitable, depending on wich physical CPUs you have in your cluster