[openstack-dev] [nova][cinder] About rebuilding volume-backed instances.

John Griffith john.griffith8 at gmail.com
Wed Nov 11 04:46:50 UTC 2015


On Mon, Nov 9, 2015 at 8:15 AM, Zhenyu Zheng <zhengzhenyulixi at gmail.com>
wrote:

> Hi, thanks all for replying, sorry I might be a bit unclear.
>
> We have user demands that we only change the root device of an
> volume-backed instance for upper layer services. It's not cloudy but it is
> quite common. And changing OS is another demand that sort of related to
> this.
>

​I'm not sure I agree that it's "common", not to mention "common" in what
context/perspective?  Regardless I don't think the complexity and risk that
something like this introduces is worth the questionable value that it
gives.  No offense perhaps I'm still not quite getting the use case.​  It's
possible I don't know what you mean by "upper layer services".

By the way, when you say things like "it's not cloudy" I unfortunately
automatically think "isn't OpenStack a Cloud Platform"?


> Cinder supports live-backup volume, but not support live-restore a volume.
>

​Nope, no plans at all.  This would be a pretty involved task given that
the backup in cinder is just dd of blocks at the raw device level.  It is
however certainly possible to restore to a "new" volume and then swap
attachments, but then that goes against your desire of keeping UUID's, and
it doesn't work for Cinder backed Instances, so I guess that suggestion is
out as well. ​


>
> Are we planning to support this kind of action?
>

​Nope, no plans at all.  I'd also argue that we rather than compromising
and doing things that are "not cloudy" we should be pushing harder on being
cloudy and the strengths that come from it.​  We should be good at one
paradigm, not bad at all of them.
​

>
> Yours,
> Zheng
>
> On Mon, Nov 9, 2015 at 8:24 PM, Duncan Thomas <duncan.thomas at gmail.com>
> wrote:
>
>> On 9 November 2015 at 09:04, Zhenyu Zheng <zhengzhenyulixi at gmail.com>
>> wrote:
>>
>>>  And Nova side also doesn't support detaching root device, that means we
>>> cannot performing volume backup/restore from cinder side, because those
>>> actions needs the volume in "available" status.
>>>
>>>
>>
>> It might be of interest to note that volume snapshots have always worked
>> on attached volumes, and as of liberty, the backup operation now supports a
>> --force=True option that does a backup of a live volume (via an internal
>> snapshot, so it should be crash consistent)
>>
>>
>> --
>> --
>> Duncan Thomas
>>
>> __________________________________________________________________________
>> OpenStack Development Mailing List (not for usage questions)
>> Unsubscribe:
>> OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20151110/1ce9415b/attachment.html>


More information about the OpenStack-dev mailing list