[one-users] Questions about host & VM failover mechanisms

Tino Vazquez tinova at fdi.ucm.es
Thu Nov 12 03:39:35 PST 2009


Hi Ranga,

OpenNebula currently doesn't foresee such situation, although I agree they
are insteresting one. We will take it into account for future releases. It
is tricky since OpenNebula cannot do anything with a network unreachable
machine, and if it starts again the VMs, as soon as the host is back with
its VMs we will have IP collisions.

About your last question, the algorithm to pick hosts is inside the
scheduler, in a separate process, so this is not offered via the OpenNebula
core API. You can extract the algorithm from the scheduler though, or you
can use the OpenNebula commands to perform the migration so the scheduler
gets used.

Hope it helps,

-Tino

--
Constantino Vázquez, 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 Tue, Nov 10, 2009 at 11:26 PM, Rangababu Chakravarthula <
rbabu at hexagrid.com> wrote:

>  Thank you Tino for your response.
>
> Let me restate.
>
> When a host is down, one moves the host status to ERR, keeps polling the
> host, keeps the VM status as "RUNNING".  Even though VM status is running,
> users will not be able to connect to the VM's until host is back up.
>
> So, I guess the only solution (if we dont know when the host is going to
> come back) is to resubmit the VM and let one create new VM on different
> host. What will happen to the old VM's in that case?
>
> For the last question, we want to migrate before the shutdown happens.  For
> migration one API needs the destination hostId. one has the algorithm to
> pick the best suitable host for a VM. Is that exposed as API? So that we can
> get the hostId and do migration.
>
> Ranga
>
>
> Tino Vazquez wrote:
>
> Hi Ranga,
>
>
> comments inline,
>
>
> On Tue, Oct 27, 2009 at 6:41 PM, Rangababu Chakravarthula <
> rbabu at hexagrid.com> wrote:
>
>> Opennebula team,
>>
>>  (a) What mechanism does Opennebula use to monitor hosts?
>>
>
> It sends probes (scripts) using ssh
>
>
>>
>>  (b) I would like to know how we can migrate VM's in the following two
>> scenarios.
>>
>>  (1) The host might be up but due to some network issues, opennebula is
>> not able to monitor the host.
>>  (a) Does it move the host status to ERR?
>>
>
> Yep, but it keeps trying to monitor them
>
>
>>  (b) What does it do to the VM's. Does it move to a different state?
>>
>
> No, it tries to connect to them.
>
>
>>  (c) Is there a mechanism to automatically migrate those VM's to
>> different host?
>>
>
> If the host is unreachable, this is tough to do.
>
>
>>
>>  (2) When somebody manually shutdown the host. Vm's on that host are not
>> accessible anymore for users.
>> Is there a way during the shutdown sequence we can trigger a script to
>> migrate the VM's ? Because, by the time opennebula
>> finds that host is down users will get effected. How to address this?
>>
>
> This is tricky. If the physical host is shuttted down, there is virtually
> not a lot you can do, appart from maybe starting the VMs again in another
> host.
>
>
>>
>>  Thanks
>> Ranga
>>
>
> Hope it helps,
>
> -Tino
>
>
>>
>> _______________________________________________
>> Users mailing list
>> Users at lists.opennebula.org
>> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>>
>
>
>
> --
> Constantino Vázquez, 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
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20091112/c9c1a255/attachment-0003.htm>


More information about the Users mailing list