VMs with scsi don't start anymore

[crazy-to-bike]

Hallo,

I have created 2 VMs (one with an IPCop and one with an Debian Server) with each one scsi Disk. That went fine until today. When I try to start them, I get an error message in the console:

Booting from local disk...
Boot failed... Press a key to retry

When I create an new VM with an IDE Disc and install the IPCop again, it works fine.

Unfortuantely, Proxmox installs an scsi Disk again, when I restore a snapshot.

Yesterday, after I had shutdown the VMs, I updated Proxmox. I think that these Updates occur the error.

So, how can I solve the Problem?
The Debian Server contains so many configurations and data, that I should be able to use it further.

Please help me to solve the Problem.

Edit:
Now the messages are different:

Booting from Hard Disk...
Could not read the boot disk

The Problem is the same.

 

[tom]

the issue is discussed on the developer list, see http://pve.proxmox.com/cgi-bin/mailman/private/pve-devel/ (you need to be a member to see the archive).

as workaround, do a downgrade of the pve-qemu-kvm package (to 0.15).

 

[crazy-to-bike]

I never had to do a downgrade before.

Can you tell me how to do this?

 

[crazy-to-bike]

I tried to downgradw with the following command:

dpkg -i pve-qemu-kvm_0.15.0-2_amd64.deb

The output was

(Reading Database [...]
Preparing zo replace [...]
Unpacking replacement pve-qemu-kvm ...
Setting up pve-qemu-kvm (0.15.0-2) ...
Processing triggers for man-db ...

No errors.

But now no VM starts anymore - no matter which kind of hd.

Please help me to get a working Installation again.

 

[crazy-to-bike]

now I tried

dpkg -i pve-qemu-kvm_0.15.0-1_amd64.deb

and the VMs start again. But the Problem with scsi-HDs is the same.

What can I do?

 

[tom]

read the dev list.

 

[crazy-to-bike]

read the dev list.

Fist: I'm not a member, so I can't read it.

Further, I feel overextended with a developer mailinglist.
I'm just a user who made the mistake to
1. chose the beta
2. make an update too early

Will there be an update that solves the Problem soon?

 

[dietmar]

Will there be an update that solves the Problem soon?

No, that problem is not easy to solve.

 

[dietmar]

Cant you switch from scsi to ide (edit /etc/pve/qemu-server/VMID.conf)?

 

[crazy-to-bike]

No, that problem is not easy to solve.

ok, but is there a possibility to downgrade the system, that scsi will work again?

Downgrading of pve-qemu-kvm alone didn't take effekt.

The attempt to downgrade some of the other packages which were updated when the problem occured, fails.

 

[crazy-to-bike]

Cant you switch from scsi to ide (edit /etc/pve/qemu-server/VMID.conf)?

I tried this with the IPCop-vm. But it ends in a kernel panic.

If I am lucky, it's only because of wrong entries in te fstab (sda instead of hda)

I'll try it again with chancged fstab.

 

[crazy-to-bike]

The try to migrate the IPCop from scsi to ide failed.

in the fstab there are entries like

/dev/harddisk1

A change to

/dev/hda1

doesn't take any effect.

I will now try this with the debian server.

If you have any more ideas to reactivate my vms please tell me.

 

[crazy-to-bike]

At least the debian server boots from the virtual ide hd.

That reduces the expenditure of reconfiguration very much.

Can I assume that such a problem will not appear again using ide hds in the vms?

Otherwise it would imply not to update the proxmox system any more. But that would come into beeing a security leak.

 

[tom]

I highly recommend to use the stable Proxmox VE for production systems. The beta is for testing only.

 

[alain]

Why not usung virtio disks with Linux machines ? It is my default choice for (KVM) linux VMs, and should show as good (or almost identical) performances as SCSI, as it is included in kernel.

Alain

 

[crazy-to-bike]

I highly recommend to use the stable Proxmox VE for production systems. The beta is for testing only.

The installation is not a production system. It's a testing system which should switch to a production system during 2012. According to your roadmap version 2.0 will be stable until Q1/2012.

 

[crazy-to-bike]

Why not usung virtio disks with Linux machines ? It is my default choice for (KVM) linux VMs, and should show as good (or almost identical) performances as SCSI, as it is included in kernel.

Alain

Well, I have no experiences with virtio. By experience with real hds I thought scsi would be the best choice.
When creating a new vm, the default recommendation ist ide. So after my experience with scsi I'm thinking that this is the kind of hd the developers would prefer / suggest to use.

Therefore my question in post #13, if using ide as virtual hd will ensure that such a problem will not appear again.

 

[alain]

It will work with IDE, but performances (IO) will be much lower than virtio. You should try. Virtio is stable for Linux VMs. It is just a matter to remove the disk, then re-add it as virtio...

Alain

 

[crazy-to-bike]

It will work with IDE, but performances (IO) will be much lower than virtio. You should try. Virtio is stable for Linux VMs. It is just a matter to remove the disk, then re-add it as virtio...

Alain

can the developers confirm this?

 

[spirit]

The problem is we cannot "boot" anymore on scsi without using non-gpl lsi rom, since qemu 1.0.

As workaround,you can you boot from an ide disk (or floppy , then mount root on a scsi disk.