guest freeze with 100% cpu usage on pve 1.3

[nextime]

Hello.

Im investigating an issue where apparently in a semi-random timing, when a guest have some load ( > 0.50 ) sometimes it gives some kernels messages and it hangs, no way to do anything in the guest.

On the host side, i can see the kvm process using 100% cpu, this apart, all the rest is working good on host, and event other guests are working.

It happen more times with a low cpu settings, it is more rare with cpu setting higher than 40000

ENVIRONMENT:

PVE: latest 1.3
KERNEL HOST: latest pve 2.6.24 from apt
DISTRO HOST: debian lenny with proxmox packages from apt
KERNEL GUEST: 2.6.30-1-686 from debian package
DISTRO GUEST: debian sid

Real HW:

CPU: AMD Opteron quad core 2200 Mhz
RAM: 8 gigs
DISKS: 2x 1TB sata disks with raid1 hardware

Guest "hw":

Network: virtio
block device: virtio

Guest config:

virtual disk on raw file (20 GB)
1 GB of ram
fair scheduler: 50000

NOTE: guest has on the disk 2 partitions, an ext2 fs on /boot, and a dm-crypt for the second partitions, with lvm on top of the encrypted volume, on which there are other ext3 filesystems

NOTE2: maybe related to this bug of virtio block device: https://bugzilla.redhat.com/show_bug.cgi?id=510304

I'm try to see if using other block device than virtio it happen. Now i'm trying with IDE disk.

 

[dietmar]

Yes, please try without virtio.

 

[nextime]

UPDATE:
this morning ( i'm in italy timezone ) i found the guest with very high load, caused probably by a lost of connectivity on the primary ethernet interface.

On this guest there are 2 eth interfaces, both virtio, one bridget on vmbr0 with the public ip, and the second bridged with vmbr1 on the host, with a local LAN IP.

I have a VPN with the host, so, i can reach from my home the LAN interface.

The public interface was dead, the lan one was working, but after few commands in ssh, also the lan interface was freezed.

Logging in by VNC on the console, the machine was responsive, but no networking.

So, i have rebooted the machine changing the two ethernet interfaces from virtio to rtl8139.

First i tryed e1000, but the machine was freezing during boot with 100% cpu (on the guest), so, i've tryed also with rtl8139 that seem to work like a charm.

NOTE: i have another guest with same identical kernel/distro, that have 2 virtio network interfaces and 2 virtio disks, with now 5 days of uptime. It work good. This machine have less load.

 

[nextime]

Ok, another update. Also the second guest suffer the issue, it is just freezed.

 

[nextime]

Bad news. Even with IDE block device, my machine have just freezed...

 

[tom]

Bad news. Even with IDE block device, my machine have just freezed...

thanks for reporting.

another question:
why do you expect a stable system when you use Debian Sid (unstable) as a guest?