[one-users] "force" deploy of VM

[Carlos Martín Sánchez]

Hi,

On Wed, Mar 19, 2014 at 1:06 PM, Mario Benincasa <mbeninca at gmail.com> wrote:

> Thanks Carlos for your answer.
>
> About your suggestion, do you mean calling the fault tolerance script
> (host_error.rb) with the "-f" option?
> Or some more tweaking?
> Because the comments in the oned.conf file say "force resubmission of
> suspended VMs".
> In my case VMs are not in suspended state, but in pending state.
> Resubmission means redeploy, i.e. new VM id allocated?
>
> Thanks
>
>     Mario
>

I was suggesting to modify the script, and perform the onevm deploy just
after the vm.delete(true) call.
The hook is a simple ruby script (host_error.rb), it shouldn't be too
difficult. But if are not familiar with the ruby API, you can also create
another bash script to perform the deployment, and then call that one from
the host_error.rb hook.

Let us know if you run into any problems.

Regards
--
Carlos Martín, MSc
Project Engineer
OpenNebula - Flexible Enterprise Cloud Made Simple
www.OpenNebula.org <http://www.opennebula.org/> | cmartin at opennebula.org |
@OpenNebula <http://twitter.com/opennebula> <cmartin at opennebula.org>



>
> On Wed, Mar 19, 2014 at 11:58 AM, Carlos Martín Sánchez <
> cmartin at opennebula.org> wrote:
>
>> Hi,
>>
>>  On Tue, Mar 18, 2014 at 10:18 AM, Mario Benincasa <mbeninca at gmail.com>wrote:
>>
>>> Hi all
>>>
>>> I have a simple question (well I believe it is simple :-) )
>>> I searched through documentation and mailing list archive, but I wasn't
>>> able to find anything.
>>> Apologies if the question has already been answered somewhere.
>>>
>>> After template instantiation, the VM is placed in pending state until
>>> the scheduler decides to run it.
>>> If there are not enough resources, it will remain in pending state.
>>> "onevm deploy" will anyway run it, even in case of not enough resources.
>>>
>>> My question is: is it possible to automatically have the deploy of the
>>> VM, even if the scheduler would leave it in pending state?
>>>
>>> I need this for the fault tolerance script: I have two nodes, each
>>> running a certain number of VMs. In case a node crashes, I want the VMs to
>>> be deployed in the surviving node, even if the scheduler disagrees.
>>> Currently, the VMs are left in pending state, thus they are not
>>> recovering the service. Of course a manual "onevm deploy" would recover,
>>> but I would like to have it automatically :-)
>>> The impacted resource appears to be the RAM: the sum of the VMs' RAM
>>> allocation exceeds the physical RAM, but in real operating conditions the
>>> RAM usage is quite under the allocated RAM, and thus I expect no problem in
>>> such an over provisioning.
>>>
>>> Thanks for any hint.
>>>
>>>    Mario
>>>
>>
>> We don't have any way to tell the scheduler to ignore the VM
>> requirements, and I think it would be a problematic feature. Memory
>> overcommitment (up to a certain amount) may not have a noticeable impact,
>> but a full disk will certainly do.
>>
>> I think your best option is to have the fault tolerance script do the
>> forced deploy.
>>
>> Regards
>> --
>> Carlos Martín, MSc
>> Project Engineer
>> OpenNebula - Flexible Enterprise Cloud Made Simple
>> www.OpenNebula.org <http://www.opennebula.org/> | cmartin at opennebula.org
>>  | @OpenNebula <http://twitter.com/opennebula> <cmartin at opennebula.org>
>>
>>
>>
>>
>
>
> --
> Mario Benincasa
> Via del Conservatorio 55
> 00186 Roma - Italy
> tel.   +39 066872917
> mob.   +39 3346210331
> fax:   +39 0697656510
> email: mbeninca at gmail.com
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20140319/075b9a28/attachment-0002.htm>