[one-users] Trouble starting up Virtual Machine from marketplace

[Mark Farragher]

Hi Jamie, 

I fixed the problem!  

I manually created a working virtual machine in virt-manager and then created another one through OpenNebula. After cloning the OpenNebula virtual machine in virt-manager I see its xml file appearing in /etc/libvirt/qemu and this allows me to do a side-by-side comparison of a working VM and a non-working one. 

One of the differences I spotted is that the OpenNebula virtual machine (which does not work) has this line:

	KERNEL = "/vmlinuz"

The working virtual machine does not specify a kernel. So I edited the cloned OpenNebula xml file and took this line out, restarted the libvirt daemon and rebooted the virtual machine in virt-manager. And it worked! 

I traced the setting to the /etc/one/vmm_exec/vmm_exec_kvm.conf file in the front end. I commented the KERNEL line out and restarted the opennebula daemon. From that point onward every virtual machine I create in OpenNebula works! 

My theory is that by specifying the kernel the VM boots into a kernel that does not have virtio support. So once the bootloader finishes it does not see any bootable device and halts. By omitting the KERNEL parameter I let KVM automatically pick the right kernel. 



Kind regards,
Mark Farragher 


On Mar 28, 2013, at 1:32 PM, Jaime Melis <jmelis at opennebula.org> wrote:

> Hi Mark,
> 
> let's try again with a different BUS. Remove the BUS tag altogether and in the template write this instead:
> ...
> DISK = [ IMAGE_ID = 3, DEV_PREFIX = sd ]
> ...
> 
> cheers,
> Jaime
> 
> 
> On Wed, Mar 27, 2013 at 1:29 PM, Mark Farragher <mark at firstfocus.eu> wrote:
> Hi Javier,
> 
> Interesting! That would explain why I can never get any VM to boot. But the images seem to be okay. This is my front end:
> 
>         oneadmin at metroplex:/etc/one$ ls  /var/lib/one/datastores/101/8231dd33c68c873afa94f1deff1688d2 -al
>         -rw-rw---- 1 oneadmin oneadmin 11811160064 Mar 27 14:41 /var/lib/one/datastores/101/8231dd33c68c873afa94f1deff1688d2
> 
> And this is the deploy folder on the host:
> 
>         root at Pinky:/opt/one/datastores/0/21# ls -al
>         total 11534352
>         drwxr-xr-x 2 oneadmin cloud        4096 Mar 27 17:04 .
>         drwxr-xr-x 3 oneadmin cloud        4096 Mar 27 17:02 ..
>         -rw-r--r-- 1 oneadmin cloud         976 Mar 27 17:04 deployment.0
>         -rw-r----- 1 oneadmin cloud 11811160064 Mar 27 17:04 disk.0
> 
> Both files are 11GB in size. In the PROLOG phase I can see 11 GB being transferred from the front end to the host.
> 
> I also tested with another virtual machine, like this:
> 
> 1. I manually created a new VM on the host in virt-manager
> 2. I booted the VM, connected over the network using SSH, installed some software in it and verified that everything is ok.
> 3. I then copied the image file of this working vm (30GB) to the front end and created an OpenNebula image with it
> 4. I tried to boot a new VM in OpenNebula using this image.
> 
> This gives the exact same error. Isn't it strange that two unrelated disk images both get corrupted in the same way as soon as I put them in OpenNebula?
> 
> My OpenNebula version is 3.6.0 and my host OS is:
> 
>         oneadmin at Pinky:/opt/one/datastores/0$ cat /etc/issue
>         Debian GNU/Linux 6.0 \n \l
> 
>         oneadmin at Pinky:/opt/one/datastores/0$ uname -a
>         Linux Pinky 2.6.32-5-amd64 #1 SMP Mon Feb 25 00:26:11 UTC 2013 x86_64 GNU/Linux
> 
> My front end OS is the same :
> 
>         oneadmin at metroplex:/etc/one$ cat /etc/issue
>         Debian GNU/Linux 6.0 \n \l
> 
>         oneadmin at metroplex:/etc/one$ uname -a
>         Linux metroplex 2.6.32-5-amd64 #1 SMP Mon Feb 25 00:26:11 UTC 2013 x86_64 GNU/Linux
> 
> 
> Is it possible the images are okay and the full 11GB is transferred, but the booting VM is somehow unable to deal with lvm partitions and therefore only sees the boot partition? Do I need to install or upgrade something to get lvm support in kvm working?
> 
> 
> Kind regards,
> Mark Farragher
> 
> 
> 
> On Mar 27, 2013, at 6:02 PM, Javier Fontan <jfontan at opennebula.org> wrote:
> 
> > I think I've found the problem. The Image is not correctly downloaded,
> > its size is 200 Mb instead of 11 Gb. I've also been digging on what
> > could be block device major 254 and it may be an lvm partition. 200 Mb
> > is enough to hold the boot loader and the boot partition but not the
> > root partition, so it crashed.
> >
> > It is strange that you had no error downloading that image. Can you
> > tell me the OpenNebula version and frontend distribution to test this
> > here?
> >
> > Try to download the image again and check the size.
> >
> 
> 
> _______________________________________________
> Users mailing list
> Users at lists.opennebula.org
> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
> 
> 
> 
> -- 
> Jaime Melis
> Project Engineer
> OpenNebula - The Open Source Toolkit for Cloud Computing
> www.OpenNebula.org | jmelis at opennebula.org


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20130329/da4170c9/attachment-0002.htm>