IDE problems in my Windows KVMs after upgrading to 3.1

[JMoreno]

Hi all,

I am experiencing problems with ALL my Windows KVMs machines after upgrading to 3.1.

More preciesly, when I connect to the console, machines are self recovered and auto rebooted. Minimum, once a day.

Here you can find two screenshoots, which may help:

1.- Error message at boot:"System error: system halted"



2.- Event viewer: You can see all recorded "atapi" related events.

Error message sais: "Device, \Device\Ide\IdePort0 din't replay during the permited time"



It must be said, I have the feeling, KVMs are working slowly after the upgrade. Perhaps it is also "IDE" related.


I would appreciate any help you could provide me with this issue.

Many thanks in advande.

 

[JMoreno]

Do you think applying this reg modifications may help?

http://pve.proxmox.com/wiki/Migration_of_servers_to_Proxmox_VE#inaccessible_boot_device

I never need it in the past with previous ProxMox versions. I was not even aware of it...

You know, I do not like touching the Windows registry except when needed.

Thanks again.

PS: my KVMs do not come from a migration. They were installed directly over new ProxMox KVMs.

 

[screenie]

can you post a qemu config file of such an vm?
and also a screenshot of the device manager (uncollapsed)

i moved a w2k3 vm from PVE 1.9 to 3.1 without any issues two weeks ago;

you could try also moving to virtio where i do it that way:
add a new harddisk with virtio
shutdown/start the vm that the new disk is detected
install the virtio drivers, make sure new harddisk is working
shutdown vm
remove the existing ide harddisks, and re-add it as virtio disk
boot vm and verify hardware in device manager
don't forget to remove the not needed additional harddisk you created as it was only to install the virtio disk drivers;

also, you may want to create a backup before;

if the issues are related to the cdrom drive, you can also try removing the cdrom drive from the vm, and assign a new on after a shutdown/start of the vm - maybe this helps too....

and i did no modifications in the registry, i installed it with ide driver - have still one w2k3 server without virtio disk driver which runs fine

--- snip ---
onboot: 1
boot: cdn
ostype: w2k3
keyboard: de
name: seth2003.domain.local
cpu: qemu32
sockets: 1
cores: 2
memory: 2048
scsihw: virtio-scsi-pci
bootdisk: ide0
ide0: local-StorageSAS:101/vm-101-disk-1.qcow2,format=qcow2,size=50G
ide1: local:iso/virtio-win-0.1-65.iso,media=cdrom,size=60558K
ide2: local-StorageSAS:101/vm-101-disk-2.qcow2,format=qcow2,size=100G
net0: virtio=FA:B8:C2:7C:A2:A0,bridge=vmbr0
--- snip ---

 

[hotwired007]

it might also be worth giving some background information on host & storage (is it local / NFS / iSCSI etc)

 

[JMoreno]

Hi Screenie and Hotwired

Thanks for your reply.

Let me copy the .conf file for one of the failing KVMs, AFTER removing the CD drive:

acpi: 1
balloon: 0
boot: cad
bootdisk: ide0
cores: 2
cpu: qemu32
cpuunits: 50000
freeze: 0
ide0: local:552/vm-552-disk-2.qcow2,format=qcow2
ide1: local:552/vm-552-disk-1.qcow2,format=qcow2
kvm: 1
localtime: 1
memory: 2048
name: sv-Aplicaciones
net0: virtio=CA:19:13:29:2B:BE,bridge=vmbr0
onboot: 1
ostype: w2k3
sockets: 1
startup: order=400
vga: cirrus

After removing the CD drive, I shutdown/started the KVM. Unfortunatelly, I can see the same events in the Windows Event Viewer.

From Screenie's reply, I can understand your proposal is to "convert" the IDE drives into VIRT-IO, but I can not understand how you do it. I mean, I can not see where you say something like "Convert this IDE into VIRT-IO".

All, KVMs are in the server local storage (SCSI Raid 5).

Thanks again for your kind help.

 

[screenie]

From Screenie's reply, I can understand your proposal is to "convert" the IDE drives into VIRT-IO, but I can not understand how you do it. I mean, I can not see where you say something like "Convert this IDE into VIRT-IO".

There is no converting of the disk images, with virtio you just change the way how your disk images are accessed;
Virtio gives you more performance than ide and therefore a good choice or necessary on some installations;

And you can easily switch back from virtio to ide if you don't need or won't it - simply shutdown the vm, remove the virtio disks and re-add them as ide disks...

 

[hotwired007]

how many VMs do you have running on the server and how many disks are in the raid5 array?

 

[JMoreno]

Hi again,

- screenie:

I have followed your advice (ie replace IDE by VIRTIO hard disks). I am affraid I still get same error messages but as virtio. Please, see attached screenshot.



Any other advice your may have for me? Thanks.

- hotwired007

My hardware server is an IBM System x3650 M3 with 48 GB and 6 x 600 SAS harddisks under a RAID 5.

root@pmox1:~# df -h
Filesystem Size Used Avail Use% Mounted on
udev 10M 0 10M 0% /dev
tmpfs 4.8G 400K 4.8G 1% /run
/dev/mapper/pve-root 95G 32G 59G 36% /
tmpfs 5.0M 0 5.0M 0% /run/lock
tmpfs 9.5G 38M 9.4G 1% /run/shm
/dev/mapper/pve-data 2.6T 331G 2.3T 13% /var/lib/vz
/dev/sda2 494M 159M 310M 34% /boot
/dev/fuse 30M 92K 30M 1% /etc/pve
pmox12:/var/lib/vz/nfs/pmox 2.6T 2.2T 464G 83% /mnt/pve/sv-NFS
/var/lib/vz/private/106 500G 162G 339G 33% /var/lib/vz/root/106
tmpfs 5.0G 0 5.0G 0% /var/lib/vz/root/106/lib/init/rw
tmpfs 5.0G 0 5.0G 0% /var/lib/vz/root/106/dev/shm

I am running 2 OVZ and 5 KVM (Windows). There are another 3 more stopped KVMs, which are running on a very few occassions in a month. Most efforts are required to one of the OVZs.

In most of the hours, it is NOT a heavy used ProxMox server. The "Load average" value is below 0.5. The "IO delay" value is 1.5%.

I had no issues under 1.9 or 2.X, but since August (after the server migration to 3.1), I am experiencing these problems, specially in the last couple of weeks.

Thanks you both for reading my posts and for your comments.

Regards.

 

[JMoreno]

Hi, any idea on how to deal with this problem?

Thanks.

 

[Extcee]

Is the disk image to this stored locally on the server or across a NFS/SAN?

If the disk image is stored external to the server can you try moving it locally?

 

[JMoreno]

Hi Extcee,

Yes, the disk images are stored in the local RAID5 system.

Thanks.

 

[JMoreno]

Hi all,

Just to let you know how this issus is finnished. The problem was with one physical hard disk in the RAID. It was not displayed as damaged by the server, but it was failling.

Finally, one day it died. I replaced it by a new one and all the issues descrived above are over.

Thanks you all for your help in any case.

Regards.

 

[term]

What type of RAID card are you using?

 

[JMoreno]

The issues were in a IBM System x3650 M3 server.

Regards.