On Sun, Jul 14, 2013 at 6:42 PM, Guido Trotter <[email protected]> wrote:
>
> On 13 Jul 2013 10:44, "harryxiyou" <[email protected]> wrote:
>>
>> On Sat, Jul 13, 2013 at 4:08 PM, Guido Trotter <[email protected]>
>> wrote:
>> [...]
>> >> Summary here. Actually, I think we should realize "GlusterFS Ganeti
>> >> Support" by two parts as follows.
>> >>
>> >> After run "gnt-instance -t gluster xxxx", if the VM is QEMU (KVM), it
>> >> would
>> >> run as KVM+GlusterFS way. However, if the VM is XEN, it would run
>> >> as glusterfs file way (Like gnt-instance add -t file xxx). For me, I
>> >> would
>> >> finish the glusterfs file way firstly because I have installed Ganeti
>> >> with XEN
>> >> VMs. After this part finished, I would complete KVM+Glusterfs way for
>> >> QEMU (KVM) VM.
>> >>
>> >> Any comments?
>> >>
>> >
>> > Looks like a good plan. Please consider the option, if the user wants,
>> > to do qemu+kernel backend too.
>>
>> Yeah, I would consider the option and expand my
>> design-glusterfs-ganeti-support doc with these ideas.
>>
>> > This should be encoded as a disk parameter.
>> >
>>
>> Yeah, these two parts (glusterfs for XEN and QEMU (KVM)) have
>> the same disk type called "gluster". They would do different actions
>> after they check their VM type.
>>
>
> Or simply on the parameter, with the addition that setting the parameter to
> userspace would fail for Xen, and that the default would be different
> depending on the hypervisor type (which is the harder change, but we can
> figure it out)
>

I think "-H Hypervisor_name:xxx" option of gnt-instance command could tell us
the hypervisor type and we could do different actions by "Hypervisor_name"
option, right?

Thanks very much ;-)

--
Thanks
Weiwei  Jia (Harry Wei)

Reply via email to