[Lxc-users] lxc-create -B zfs

Tamas Papp tompos at martos.bme.hu
Wed Nov 20 16:53:44 UTC 2013


On 11/20/2013 05:24 PM, Serge Hallyn wrote:
> Quoting Tamas Papp (tompos at martos.bme.hu):
>> On 10/01/2013 07:14 AM, Serge Hallyn wrote:
>>> Right now if we say
>>>
>>> 	lxc-create -t ubuntu -n u1 -B zfs
>>>
>>> that means make the rootfs be type zfs.  Where/how should it be
>>> specified that $lxcpath/$lxcname should be a new zfs, not
>>> $lxcpath/$lxcname/rootfs?  'zfs.mounttype = [whole | rootfs]' ?
>> Well, it's good idea.
>> BTW I think it would be the same for lvm, btrfs and so on.
>>
>>> Something like that would be ok for me, my main concern is actually
>>> that the special casing in lxcapi_create() and lxcapi_destroy()
>>> be done cleanly and concisely.  But yeah it should be possible.  Do
>>> you mind sending a patch?
>> It would be great, but I checked and unfortunately it's much more then I
>> can do with my programming skills:(
> Ok, lemme put out a general call - is there anyone who is using or
> wants to use lxc with zfs and who would have time to help straighten
> this out?  If/when I get two days where nothing else is likely to come
> up, I'd like to sit down and address this, but it'll take that time
> to clear my head to think about this right so that I make it better
> not worse...
>
> (I hope that sometime in december I'll have such a block of time,
> but hopefully someone beats me to it)

Seeing how busy you are, I'm doubt, it will ever happen:)
Anyway, if there is no volunteer, just drop the request. I can live with
a workaround (llxc-create container,  zfs create/lvcreate dataset, mv
container/* dataset/).
Though I'm afraid there will be more features in the future, that
require the layout compliance (eg. migration tool) and I'll be in trouble..


tamas




More information about the lxc-users mailing list