openstack team mailing list archive

[Mohammed Amine SAYA <asaya.openstack@xxxxxxxxx>]

Hi Mark,

Thanks for your answer.

What I noticed is that this issue happens only when I try to start a VMfrom a
bootable volume. I created the colume with
cinder create --image-id __image_id__ volume_name.

I don't see it when I boot a VM from an image in glance (nova boot --image
__image_id__ --flavor __flavor_id__ vm_name).
Which runs just fine. I can ping and ssh into the VM. But this is ephemeral
boot and I want to be able to boot on a
persistent volume so that the application won't lose data from previous
session. And you are right, I can see nova-compute
getting around nbd module in this case.

I followed the instructions in this paragraph (http://docs.openstack
.org/grizzly/openstack-compute/admin/content/instance-creation.html#boot_from_volume)
but without success until now.

Did you manage to boot from a volume? If nbd is not the right module to use
then what else should I use to boot a VM from
a persistent volume?

Thanks for your help,
Amine.


On Fri, May 3, 2013 at 3:49 PM, Mark Lehrer <mark@xxxxxxxxx> wrote:

>
>  After loading nbd module in compute node, I continued trying to boot from
>> a
>> bootable cinder volume but this is what I see in nova-compute.log.
>>
>
> I am not using Cinder so maybe this isn't an option for you, but my
> recommendation is to blacklist the "nbd" module altogether, and remove
> the nbd option from the nova-compute upstart init file.  It is a neat
> protocol but it has not been nearly reliable enough to use on my
> system.
>
> nova-compute is able to work around the missing nbd module (with
> glance at least), and it cut my image create and snapshot times
> drastically in addition to reducing the launch & snapshot failures.
>
> Mark
>