On Mon, Feb 01, 2016 at 03:52:12PM -0500, John Ferlan wrote:
>> <volume type='block'>
>> ...
>> <source>
>> <name>%s</name>
>> <format type='thin_pool'>
>> </source>
>> ...
>> </volume>
>>
>> Trying to forward think if/when a vol.xml was used as input...
>
> Maybe we should stop right here and handle the thin pool/volumes in a separate
> patch series. It would be best to create a new libvirt pool type that will be
> dedicated for lvm thin pools and add a proper support for thin pools.
>
>
Without patches 4 & 5 we get the following for virsh vol-dumpxml:
<volume ...>
...
<source>
</source>
...
</volume>
Actually I have a guest with all different types of logical volume and I wasn't
able to list those thin lv via libvirt, so I assumed, that we are somehow
ignoring them so far. I'll try it again together with your patches.
Which is OK - although it looks a bit strange to me - it isn't the only
output that doesn't have <source> defined.
Since a thin lv and thin-pool can be placed into any vg and a vg is
essentially a libvirt pool, I'm not so sure having a dedicated pool type
is necessary. Sure it'd make some things easier, but can we assume every
thin volume instance is configured that way?
Yes, it could be placed in current logical pool, but IMO it would be good to
separate those thin pools and lvs from "classic" lv. If someone would like to
use thin volumes in libvirt, one will have to configure new pool where source
will be desired thin pool.
Pavel