[openstack-dev] [cinder][nova] about re-image the volume

Matt Riedemann mriedemos at gmail.com
Mon Apr 9 20:05:53 UTC 2018


On 4/9/2018 1:00 PM, Duncan Thomas wrote:
> Hopefully this flow means we can do rebuild root filesystem from
> snapshot/backup too? It seems rather artificially limiting to only do
> restore-from-image. I'd expect restore-from-snap to be a more common
> use case, personally.

Hmm, now you've got me thinking about image-defined block device 
mappings, which is something you'd have if you snapshot a volume-backed 
instance and then later use that image snapshot, which has metadata 
about the volume snapshot in it, to later create (or rebuild?) a server.

Tempest has a scenario test for the boot from volume case here:

https://review.openstack.org/#/c/555495/

I should note that even if you did snapshot a volume-backed server and 
then used that image to rebuild another non-volume-backed server, nova 
won't even look at the block_device_mapping_v2 metadata in the snapshot 
image during rebuild, it doesn't treat it like boot from volume does 
where nova uses the image-defined BDM to create a new volume-backed 
instance.

And now that I've said that, I wonder if people would expect the same 
semantics for rebuild as boot from volume with those types of 
images...it makes my head hurt. Maybe mdbooth would like to weigh in on 
this given he's present in this thread.

-- 

Thanks,

Matt



More information about the OpenStack-dev mailing list