Mailing-List: contact dev-help@cloudstack.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@cloudstack.apache.org
Received-SPF: pass (athena.apache.org: domain of kelven.yang@citrix.com
 designates 66.165.176.63 as permitted sender)
From: Kelven Yang <kelven.yang@citrix.com>
To: "dev@cloudstack.apache.org" <dev@cloudstack.apache.org>
CC: "cloudstack-dev@incubator.apache.org"
	<cloudstack-dev@incubator.apache.org>
Date: Tue, 9 Apr 2013 10:54:52 -0700
Subject: Re: [DISCUSS] pause VM
Thread-Topic: [DISCUSS] pause VM
Thread-Index: Ac41S1cJfYqAcSuFS/WaO37JqnkjVw==
Message-ID: <CD89A007.3C58F%kelven.yang@citrix.com>
In-Reply-To: 
 <CALFpzo5bJ0gC_tKuJStBCuDhjdfMKOs86XdYp8YwLHrKRxfRmg@mail.gmail.com>
Accept-Language: en-US
Content-Language: en-US
user-agent: Microsoft-MacOutlook/14.2.3.120616
acceptlanguage: en-US
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0


On 4/8/13 7:11 PM, "Marcus Sorensen" <shadowsor@gmail.com> wrote:

>Perhaps. I think there are separate issues here. There is volume snapshot
>and vm snapshot. If the hypervisor is capable of making volumes consistent
>, but VM snapshot is more about saving running state, then the two may not
>serve the same purpose when leveraging san-based snapshots. Its absolutely
>possible to pause the VM, quiesce the disks, and then snapshot them
>consistently utilizing underlying storage capabilities.
>
>But we are getting off in the weeds here with a single example, when what
>I
>want to discuss is pausing a VM in general.

For a general pause-VM feature, I agree that we probably need it for some
use cases, just not so convinced that we need it in any urgency, and the
use case of using it to cover application level data integrity issue may
mislead users.

Let's see if there are more opinions in the community about this, adding a
new VM state actually involves some work in Sync, resource allocation
areas, that's the reason I'm cautious to ask about the real
mandatory/valid use case for this feature.

-Kelven

>On Apr 8, 2013 6:19 PM, "Kelven Yang" <kelven.yang@citrix.com> wrote:
>
>>
>>
>> On 4/8/13 4:19 PM, "Marcus Sorensen" <shadowsor@gmail.com> wrote:
>>
>> >Oh yes, I know these things, but here's my point. You said:
>> >
>> >"When CloudStack takes a snapshot of volume, it already utilizes
>>snapshot
>> >feature provided by hypervisor to allow taking point of time storage
>>image
>> >in a safe manner from operating system point of view."
>> >
>> >What about five volumes? Can we ensure they are all snapshotted at that
>> >same point in time without also pausing the VM, in addition to using
>>those
>> >hypervisor-provided features?  For example, let's say the VM is using a
>> >volume manager and striping data across five data disks, we absolutely
>> >have
>> >to take all data disk snapshots at the same point in such a scenario.
>>We
>> >of
>> >course have to rely on the user to tell us that they want this, but we
>> >should have the capability.
>> >
>> >If we already have a VM snapshot feature that does every connected disk
>> >simultaneously while the VM is suspended, or the hypervisor can do it
>>for
>> >us by quickly pausing/snapshotting/resuming seamlessly then fine,
>>that's
>> >great. It just didn't seem like that functionality existed now, so I
>>was
>> >wondering if anyone is working on it.
>>
>> I'm not sure using the tactic to pause a VM for the purpose of ensuring
>> full data integrity is the right answer for the use case you pointed. As
>> far as I know for VMware hypervisors, taking a VM snapshot does all the
>> things your mentioned above (quiesce the guest file system, snapshot
>> memory etc). Xen Server may be a little different, for crash
>>consistency,
>> I think we ask user to stop a VM(instead of pausing it) before taking a
>> volume snapshot.
>>
>> So the question is, can "pausing VM" offer what we are looking for or is
>> it a half-completed solution? Forcely-pausing a VM to offer a not
>>ensured
>> solution to customer is what I'm worrying about.
>>
>> -Kelven
>>
>> >
>> >That aside, it seems like all of the hypervisors CS controls provide
>>their
>> >own UI to pause a VM. Is it something that has been discussed?
>> >
>> >
>> >
>> >
>> >On Mon, Apr 8, 2013 at 4:16 PM, Kelven Yang <kelven.yang@citrix.com>
>> >wrote:
>> >
>> >>
>> >>
>> >> On 4/8/13 1:39 PM, "Marcus Sorensen" <shadowsor@gmail.com> wrote:
>> >>
>> >> >Has anyone put together a suggestion for adding a 'paused' state to
>> >> >virtual
>> >> >machines? Or is anyone working on it? This seems fairly
>> >>straightforward,
>> >> >and I think we may need the functionality going forward. Even if
>>people
>> >> >rarely use it themselves, if storage solutions are going to offer
>> >> >snapshots
>> >> >they may want to pause the VM in order to get an atomic snapshot if
>> >>the VM
>> >> >has multiple disks. So at least internally it may be a useful
>>utility.
>> >>
>> >> Pause VM itself can not ensure data integrity. True data integrity
>>can't
>> >> be ensured without participant of guest OS and related application
>>(for
>> >> example, Microsoft offers Volume Shadow Copy Service to orchestrate
>>and
>> >> allow applications to participate the process via a set of tools and
>>API
>> >> interaction to achieve that).
>> >>
>> >> When CloudStack takes a snapshot of volume, it already utilizes
>>snapshot
>> >> feature provided by hypervisor to allow taking point of time storage
>> >>image
>> >> in a safe manner from operating system point of view. The ability of
>> >> pausing VM is not mandatory in this case, although pausing/resuming
>>VM
>> >>may
>> >> be useful in certain use case, it does not seem to be in any urgency
>>or
>> >> fit naturally in general cloud-operations.
>> >>
>> >> -Kelven
>> >>
>> >> >
>> >>
>> >>
>>
>>