[one-users] cannot migrate virtual machines in one 3.4

Carlos A. caralla at upv.es
Tue Apr 24 08:36:48 PDT 2012


Hi,

I have also checked this option, but I found also a problem.

If I change the system datastore (0) to set the TM_MAD ssh and then I 
create a new VM and try to migrate it, the  vm.log fragment is next:

------------------------------------------------
Tue Apr 24 17:17:07 2012 [LCM][I]: New VM state is SAVE_MIGRATE
Tue Apr 24 17:17:10 2012 [VMM][I]: ExitCode: 0
Tue Apr 24 17:17:10 2012 [VMM][I]: Successfully execute virtualization 
driver operation: save.
Tue Apr 24 17:17:10 2012 [VMM][I]: ExitCode: 0
Tue Apr 24 17:17:10 2012 [VMM][I]: Successfully execute network driver 
operation: clean.
Tue Apr 24 17:17:11 2012 [LCM][I]: New VM state is PROLOG_MIGRATE
Tue Apr 24 17:17:11 2012 [TM][I]: ExitCode: 0
Tue Apr 24 17:17:15 2012 [TM][I]: mv: Moving 
dellblade01:/srv/cloud/one/var//datastores/0/2985 to 
dellblade03:/srv/cloud/one/var//datastores/0/2985
Tue Apr 24 17:17:15 2012 [TM][I]: ExitCode: 0
Tue Apr 24 17:17:15 2012 [LCM][I]: New VM state is BOOT
Tue Apr 24 17:17:16 2012 [VMM][I]: ExitCode: 0
Tue Apr 24 17:17:16 2012 [VMM][I]: Successfully execute network driver 
operation: pre.
Tue Apr 24 17:17:16 2012 [VMM][I]: Command execution fail: 
/var/tmp/one/vmm/kvm/restore 
/srv/cloud/one/var//datastores/0/2985/checkpoint dellblade03 2985 
dellblade03
Tue Apr 24 17:17:16 2012 [VMM][E]: restore: Command "virsh --connect 
qemu:///system restore /srv/cloud/one/var//datastores/0/2985/checkpoint" 
failed: error: Failed to restore domain from 
/srv/cloud/one/var//datastores/0/2985/checkpoint
Tue Apr 24 17:17:16 2012 [VMM][I]: error: cannot close file: Bad file 
descriptor
Tue Apr 24 17:17:16 2012 [VMM][E]: Could not restore from 
/srv/cloud/one/var//datastores/0/2985/checkpoint
Tue Apr 24 17:17:16 2012 [VMM][I]: ExitCode: 1
Tue Apr 24 17:17:16 2012 [VMM][I]: Failed to execute virtualization 
driver operation: restore.
Tue Apr 24 17:17:16 2012 [VMM][E]: Error restoring VM: Could not restore 
from /srv/cloud/one/var//datastores/0/2985/checkpoint
Tue Apr 24 17:17:16 2012 [DiM][I]: New VM state is FAILED
------------------------------------------------

and the transfer.1.migrate file is

------------------------------------------------
MV ssh dellblade01:/srv/cloud/one/var//datastores/0/2985/disk.0 
dellblade03:/srv/cloud/one/var//datastores/0/2985/disk.0
MV ssh dellblade01:/srv/cloud/one/var//datastores/0/2985 
dellblade03:/srv/cloud/one/var//datastores/0/2985
------------------------------------------------

Now I have the "checkpoint" file in dellblade03, but not the disk.0. 
This is strange because the transfer.1.migrate file tries to 
specifically move the disk.0 file but not the checkpoint file. I guess 
that the problem is the order of transferences in this case. I think 
that moving the folder once the disk has been moved is deleting the just 
moved disk. If I manually create a disk.0 file, I am able to manually 
restore the VM using virsh commands. Is there any way to solve this issue?

On the other hand, I cannot see why the system datastore transfer mad 
needs to be set to ssh.

Regards,
Carlos A.

El 24/04/12 17:17, Ruben S. Montero escribió:
> Hi,
>
> Yes this may be the problem. Clould you check the output of
> onedatastore show 0 (and 1). The TM_MAD associated with the datastore
> should be ssh. If not, could you try to update it (onedatastore
> update). There should not be any "shared" keyword as you suggest.
>
> Note that the changes on the datastore (i.e. the actual TM used) are
> only reflected on new VMs. VMs created before the changes will use the
> original TM_MAD values...
>
> Cheers
>
> Ruben
>
> On Tue, Apr 24, 2012 at 5:07 PM, Carlos A.<caralla at upv.es>  wrote:
>> Hello,
>>
>> I am upgrading my ONE 3.2 deployment to ONE 3.4 but I have one problem with
>> migration of VM between nodes (not live migration).
>>
>> When using ONE 3.2 migration was fine, but now migration fails and I cannot
>> find how to solve this problem.
>>
>> I have the default datastore, that it is a "filesystem" based on ssh tm_mad
>> (and the system datastore "0").
>>
>> When I migrate the VM, I find the next vm.log fragment:
>> ------------------------------------------------
>> Tue Apr 24 16:53:51 2012 [LCM][I]: New VM state is SAVE_MIGRATE
>> Tue Apr 24 16:53:54 2012 [VMM][I]: ExitCode: 0
>> Tue Apr 24 16:53:54 2012 [VMM][I]: Successfully execute virtualization
>> driver operation: save.
>> Tue Apr 24 16:53:54 2012 [VMM][I]: ExitCode: 0
>> Tue Apr 24 16:53:54 2012 [VMM][I]: Successfully execute network driver
>> operation: clean.
>> Tue Apr 24 16:53:54 2012 [LCM][I]: New VM state is PROLOG_MIGRATE
>> Tue Apr 24 16:53:55 2012 [TM][I]: ExitCode: 0
>> Tue Apr 24 16:53:55 2012 [TM][I]: ExitCode: 0
>> Tue Apr 24 16:53:55 2012 [LCM][I]: New VM state is BOOT
>> Tue Apr 24 16:53:55 2012 [VMM][I]: ExitCode: 0
>> Tue Apr 24 16:53:55 2012 [VMM][I]: Successfully execute network driver
>> operation: pre.
>> Tue Apr 24 16:53:55 2012 [VMM][I]: Command execution fail:
>> /var/tmp/one/vmm/kvm/restore
>> /srv/cloud/one/var//datastores/0/2984/checkpoint dellblade03 2984
>> dellblade03
>> ------------------------------------------------
>>
>> And the next transfer.1.migrate file appears
>> ------------------------------------------------
>> MV ssh dellblade01:/srv/cloud/one/var//datastores/0/2984/disk.0
>> dellblade03:/srv/cloud/one/var//datastores/0/2984/disk.0
>> MV shared dellblade01:/srv/cloud/one/var//datastores/0/2984
>> dellblade03:/srv/cloud/one/var//datastores/0/2984
>> ------------------------------------------------
>>
>> The problem is that the disk.0 file is not transfered to dellblade03. It
>> seems that the phase of executing the transference of files is omited.
>>
>> Moreover the "shared" keyword appears while there is not any shared file
>> system (but the system one that should not be considered when moving from
>> one host to another). But also the checkpointing file is not moved.
>>
>> Note: migration from one host to the same host works (as expected). So virsh
>> is able to restore the state of a saved VM.
>>
>> Any idea on this?
>>
>> Thank you in advance.
>>
>> _______________________________________________
>> Users mailing list
>> Users at lists.opennebula.org
>> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>




More information about the Users mailing list