[one-users] path error when PXE-booting VMs

[Michael W. Lucas]

Hi,

Thanks for the patch.

Short answer: still not working, but the failure is different.

My patched one_vmm_kvm.rb looks like the below (I don't know Ruby, so
I could have quite easy made an error and not known it):

--- one_vmm_kvm.rb-dist 2009-11-17 15:15:24.966874194 -0500
+++ one_vmm_kvm.rb      2009-11-17 15:15:56.445072283 -0500
@@ -85,7 +85,12 @@
         domain = tmp.read
         tmp.close()

-        cmd = "cat > #{remote_dfile} && #{LIBVIRT[:create]} #{remote_dfile}"
+#        cmd = "cat > #{remote_dfile} && #{LIBVIRT[:create]} #{remote_dfile}"
+        deployment_filename = File.baseame remote_dfile
+        images_path               =
+remote_dfile.gsub(deployment_filename,'')
+        cmd = "mkdir -p #{images_path} && cat > #{remote_dfile} &&
+#{LIBVIRT[:create]} #{remote_dfile}"

         deploy_exe = SSHCommand.run("'#{cmd}'", host, log_method(id), domain)

oned.log says:

Tue Nov 17 15:32:42 2009 [ReM][D]: VirtualMachineAllocate invoked
Tue Nov 17 15:32:42 2009 [DiM][D]: Allocating a new VM
Tue Nov 17 15:33:04 2009 [ReM][D]: VirtualMachineDeploy invoked
Tue Nov 17 15:33:04 2009 [DiM][D]: Deploying VM 2
Tue Nov 17 15:33:04 2009 [TM][D]: Message received: TRANSFER SUCCESS 2 -

2/vm.log shows:

Tue Nov 17 15:33:04 2009 [DiM][I]: New VM state is ACTIVE.
Tue Nov 17 15:33:04 2009 [LCM][I]: New VM state is PROLOG.
Tue Nov 17 15:33:04 2009 [VM][I]: Virtual Machine has no context
Tue Nov 17 15:33:05 2009 [LCM][I]: New VM state is BOOT
Tue Nov 17 15:33:05 2009 [VMM][I]: Generating deployment file: /usr/local/one/var/2/deployment.0

It appears that we never try to run the command to start the VM,
however.

Tcpdump on the compute node shows that the head node is not even
attempting to SSH to the compute node when deploying the VM.  (I do
see SSH traffic when oned monitors the compute node, so I believe my
packet sniffer is configured correctly.)

Any other suggestions?

Thanks,
==ml

On Mon, Nov 16, 2009 at 04:38:23PM +0100, Tino Vazquez wrote:
> Hi Michael,
> 
> The problem is that a diskless VM will cause that the transfer driver
> doesn't get called. This transfer driver is responsible for creating the
> 'images' directory. If you want t a network-booted machine you can try out
> the following: in src/vmm_mad/kvm/one_vmm_kvm.rb change line 88 from:
> 
>         cmd = "cat > #{remote_dfile} && #{LIBVIRT[:create]} #{remote_dfile}"
> 
> to
> 
>         deployment_filename = File.baseame remote_dfile
>         images_path               =
> remote_dfile.gsub(deployment_filename,'')
>         cmd = "mkdir -p #{images_path} && cat > #{remote_dfile} &&
> #{LIBVIRT[:create]} #{remote_dfile}"
> 
> 
> Please try this out, and if you find it useful please come back to us with
> some feedback to we can include it in future releases.
> 
> About your second questions, there are two deployment.0 images due to
> OpenNebula design and its ability to use ssh or nfs. Because the VMM driver
> doesn't know wihch storage it is being used, OpenNebula always uses a
> scratch 'images' directory in the remote node were the deployment.0 file and
> disks are copied from the front-end to there. If the storage is NFS based
> you end up with two deployment.0 files locally. The 'images' directory is
> used because doing a cat of a file over the same file will result on a null
> file. Hope this makes sense :o
> 
> Regards,
> 
> -Tino
> 
> --
> Constantino V?zquez, Grid Technology Engineer/Researcher:
> http://www.dsa-research.org/tinova
> DSA Research Group: http://dsa-research.org
> Globus GridWay Metascheduler: http://www.GridWay.org
> OpenNebula Virtual Infrastructure Engine: http://www.OpenNebula.org
> 
> 
> On Wed, Nov 11, 2009 at 9:21 PM, Michael W. Lucas <
> mwlucas at blackhelicopters.org> wrote:
> 
> > Hi,
> >
> > My ON 1.4beta2/KVM environment is working fine with disk images.  ON
> > cannot start a network-booted machine, however.  The problem appears
> > to be with the path used for the deployment file.  If I use this template:
> >
> > NAME   = onvm1
> > CPU    = 0.5
> > MEMORY = 512
> > OS      = [ BOOT   = network ]
> > NIC     = [ BRIDGE = "br0",
> >           MAC = "00:16:8B:AB:C7:14" ]
> > GRAPHICS = [type="vnc",listen="127.0.0.1",port="-1"]
> >
> > the error log shows:
> >
> > Wed Nov 11 15:14:34 2009 [VMM][D]: Message received: LOG - 2 Command
> > execution fail: 'cat > /one_dir/2/images/deployment.0 && virsh --connect
> > qemu:///system create /one_dir/2/images/deployment.0'
> >
> > /one_dir/2/images/deployment.0 doesn't exist.  But
> > /one_dir/2/deployment.0 does exist, along with vm.log and the transfer
> > files.  There's no "images" subdirectory.
> >
> > On a simple VM with a hard drive, such as this, we do have an images
> > subdirectory.  Oddly, I have separate but identical deployment.0 files
> > in multiple locations.
> >
> > clusterhead1/one_dir/1;ls -lai
> > total 10K
> > 156066 drwxrwxrwx 3 oneadmin oneadmin    6 2009-11-11 15:06 .
> > 155649 drwxr-xr-x 6 oneadmin oneadmin   13 2009-11-11 15:17 ..
> > 156092 -rw-r--r-- 1 oneadmin oneadmin  368 2009-11-11 15:06 deployment.0
> > 156088 drwxrwxrwx 2 oneadmin oneadmin    4 2009-11-11 15:06 images
> > 156087 -rw-r--r-- 1 oneadmin oneadmin   83 2009-11-11 15:06
> > transfer.0.prolog
> > 156067 -rw-r--r-- 1 oneadmin oneadmin 2173 2009-11-11 15:17 vm.log
> >
> > clusterhead1/one_dir/1;ls -lai images/
> > total 5K
> > 156088 drwxrwxrwx 2 oneadmin oneadmin   4 2009-11-11 15:06 .
> > 156066 drwxrwxrwx 3 oneadmin oneadmin   6 2009-11-11 15:06 ..
> > 156093 -rw-r--r-- 1 oneadmin oneadmin 368 2009-11-11 15:06 deployment.0
> > 156089 lrwxrwxrwx 1 oneadmin oneadmin  28 2009-11-11 15:06 disk.0 ->
> > /one_dir/one_images/dns1.raw
> > clusterhead1/one_dir/1;
> >
> > Any suggestions, folks?
> >
> > Thanks,
> > ==ml
> >
> > --
> > Michael W. Lucas        mwlucas at BlackHelicopters.org
> > http://www.MichaelWLucas.com/
> > Latest book:  Cisco Routers for the Desperate, 2nd Edition
> > http://www.CiscoRoutersForTheDesperate.com/
> > _______________________________________________
> > Users mailing list
> > Users at lists.opennebula.org
> > http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
> >

-- 
Michael W. Lucas 	mwlucas at BlackHelicopters.org
http://www.MichaelWLucas.com/
Latest book:  Cisco Routers for the Desperate, 2nd Edition
http://www.CiscoRoutersForTheDesperate.com/