[one-users] the synchronous problem between the cancel operation and tm_delete.sh for lxc driver (Users Digest, Vol 58, Issue 20)

[Rolandas Naujikas]

On 2012-12-06 20:48, Ruben S. Montero wrote:
> Hi
>
> It seems that this behavior is problematic. Let me elaborate on Simon
> comment. As pointed out by Simon there are two similar VM operations:

You make more confusion here, because in the past I saw similar problem 
when developing my own LVM driver. We are talking, that in delete action 
on VM opennebula executes tm/delete and vmm/cancel actions in parallel 
what causes problems, because lvm volumes could not be destroyed when in 
use by xen VM. As a workaround my LVM driver waits in background until 
it can delete LVM volume.

> * delete, this is equivalent to kill -9 It is a hard kill of the VM.
> It does not check anything and trigger cancellation/delete actions to
> free the associated resources. It is supposed to be used when a VM is
> stuck in any state.
>
> * cancel, it cancels the VM and cleans the resources but checks and
> synchro is in place.
>
> Note that this synchronization problem is only for VM's in running.
>
> This seems to be confusing, we are thinking of moving the delete
> operation to the ADMIN set (so you need admin permissions to delete)
> and so force users to use the safer cancel operation... Any thoughts?

At least it should be configurable.

Regards, Rolandas Naujikas

> Ruben
>
>
>
>
> On Thu, Dec 6, 2012 at 2:44 PM, Simon Boulet <simon at nostalgeek.com> wrote:
>> Hi,
>>
>> On Wed, Dec 5, 2012 at 11:39 PM, cmcc.dylan <dx10years at 126.com> wrote:
>>>
>>> Hi, everyone!
>>>      I'm develpment lxc driver for opennebula and i complete the basic
>>> functions now, but i meet a problem as follows.
>>>      from the log, we can see opennbula execute "cancel" fistrly  and then
>>> execute "tm_delete.sh", But the result is  cancel: Executed "virsh --connect
>>> lxc:/// destroy lxc-5". after "tm_delete.sh".
>>>
>>
>>
>>
>> I can confirm this behaviour too when sending the delete action to a VM. It
>> seems the VMM cancel and the TM delete actions are executed simultaneously
>> by OpenNebula. When sending the cancel action OpenNebula seems to properly
>> wait for the VMM cancel action to complete before calling TM delete.
>>
>> Simon
>>
>> _______________________________________________
>> Users mailing list
>> Users at lists.opennebula.org
>> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>>
>
>
>