On 05/07/2012 11:52 PM, Ayal Baron wrote:
>
>
> ----- Original Message -----
>> On 05/07/2012 07:06 PM, Shireesh Anjal wrote:
>>> On Monday 07 May 2012 02:06 AM, Ayal Baron wrote:
>>>>
>>>> ----- Original Message -----
>>>>> i can't see any justification for the 'gluster' prefix,
>>>>> as this is only additional /service/ provided by the project,
>>>>> and Gluster now is a part of the RHT.
>>>> I believe there needs to be an indication which service this is
>>>> about.
>>>> If we will support provisioning other storage types which also
>>>> have
>>>> volumes then we'd want a way to differentiate.
>>>> However, isn't there a way to simply add gluster as the name
>>>> space?
>>>> i.e. somthing like: /api/gluster/.../volumes ? (instead of
>>>> 'cluster'
>>>> as it is redundant imho)
>>>
>>> A gluster volume is a cluster level entity, and hence
>>> "/api/.../clusters/{cluster:id}" seems like the right parent URI
>>> for the
>>> gluster volumes collection resource.
>>
>> that's true for all other root entities as well:
>> - VM is DC/cluster level
>> - template is DC level
>> - disk is storage domain level
>> - network is DC level
>> - hosts are cluster level (for now)
>>
>> yet all of them have their own root collections as well.
>>
>> I think glustervolumes seems safest/most reasonable for now (either
>> at
>> cluster level or root level as well)
>
> does it make sense to also have gluster/bricks ? if so, I would nest it, i.e.
gluster/{volumes|bricks|...}
bricks are host level, afair they are not used like this at all.
gluster/xxx is interesting as well, though not parallel to current virt
mappings (storage_domains, disks, etc., being root collections)
shireesh - any thoughts about this approach:
- do you want volumes as root collection, or only under cluster
- if root, should these be glustervolumes like other root collection, or
the under a gluster collection.
According to Geert, a key question in deciding whether gluster-volumes
should exist in root collection is: are volumes potentially moveable
between clusters?
The answer, according to Shireesh, is no, and therefore
Geert believes that the volumes should be under cluster.
(Geert or Shireesh - if I misquoted you, please correct me).
Bricks logically compose volumes, and thus are modelled under volumes.
_______________________________________________
Engine-devel mailing list
Engine-devel(a)ovirt.org
http://lists.ovirt.org/mailman/listinfo/engine-devel