<div dir="ltr">Live-snapshot is definitely generic. Mainstream hypervisors (vmware, kvm, hyperv) already support it for a long time. In <a href="https://review.openstack.org/#/c/34036/">https://review.openstack.org/#/c/34036/</a>, I see the last comment made by Russell Bryant is that only libvirt backend request to implement it. I am very curious about that. To my understanding, at least vmware, kvm, hyperv should propose to implement this function. Why did we abandon that last year? Can we consider to continue this work in Juno?<br>
</div><div class="gmail_extra"><br><br><div class="gmail_quote">On Wed, Mar 12, 2014 at 3:15 AM, Jay Pipes <span dir="ltr"><<a href="mailto:jaypipes@gmail.com" target="_blank">jaypipes@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="HOEnZb"><div class="h5">On Tue, 2014-03-11 at 06:35 +0000, Bohai (ricky) wrote:<br>
> > -----Original Message-----<br>
> > From: Jay Pipes [mailto:<a href="mailto:jaypipes@gmail.com">jaypipes@gmail.com</a>]<br>
> > Sent: Tuesday, March 11, 2014 3:20 AM<br>
> > To: <a href="mailto:openstack-dev@lists.openstack.org">openstack-dev@lists.openstack.org</a><br>
> > Subject: Re: [openstack-dev] [nova] a question about instance snapshot<br>
> ><br>
> > On Mon, 2014-03-10 at 12:13 -0400, Shawn Hartsock wrote:<br>
> > > We have very strong interest in pursing this feature in the VMware<br>
> > > driver as well. I would like to see the revert instance feature<br>
> > > implemented at least.<br>
> > ><br>
> > > When I used to work in multi-discipline roles involving operations it<br>
> > > would be common for us to snapshot a vm, run through an upgrade<br>
> > > process, then revert if something did not upgrade smoothly. This<br>
> > > ability alone can be exceedingly valuable in long-lived virtual<br>
> > > machines.<br>
> > ><br>
> > > I also have some comments from parties interested in refactoring how<br>
> > > the VMware drivers handle snapshots but I'm not certain how much that<br>
> > > plays into this "live snapshot" discussion.<br>
> ><br>
> > I think the reason that there isn't much interest in doing this kind of thing is<br>
> > because the worldview that VMs are pets is antithetical to the worldview that<br>
> > VMs are cattle, and Nova tends to favor the latter (where DRS/DPM on<br>
> > vSphere tends to favor the former).<br>
> ><br>
> > There's nothing about your scenario above of being able to "revert" an instance<br>
> > to a particular state that isn't possible with today's Nova.<br>
> > Snapshotting an instance, doing an upgrade of software on the instance, and<br>
> > then restoring from the snapshot if something went wrong (reverting) is<br>
> > already fully possible to do with the regular Nova snapshot and restore<br>
> > operations. The only difference is that the "live-snapshot"<br>
> > stuff would include saving the memory view of a VM in addition to its disk state.<br>
> > And that, at least in my opinion, is only needed when you are treating VMs like<br>
> > pets and not cattle.<br>
> ><br>
><br>
> Hi Jay,<br>
><br>
> I read every words in your reply and respect what you said.<br>
><br>
> But i can't agree with you that memory snapshot is a feature for pat not for cattle.<br>
> I think it's a feature whatever what do you look the instance as.<br>
><br>
> The world doesn't care about what we look the instance as, in fact, currently almost all the<br>
> mainstream hypervisors have supported the memory snapshot.<br>
> If it's just a dispensable feature and no users need it, I can't understand why<br>
> the hypervisors provide it without exception.<br>
><br>
> In the document " OPENSTACK OPERATIONS GUIDE" section " Live snapshots" has the<br>
> below words:<br>
> " To ensure that important services have written their contents to disk (such as, databases),<br>
> we recommend you read the documentation for those applications to determine what commands<br>
> to issue to have them sync their contents to disk. If you are unsure how to do this,<br>
> the safest approach is to simply stop these running services normally.<br>
> "<br>
> This just pushes all the responsibility to guarantee the consistency of the instance to the end user.<br>
> It's absolutely not convenient and I doubt whether it's appropriate.<br>
<br>
</div></div>Hi Ricky,<br>
<br>
I guess we will just have to disagree about the relative usefulness of<br>
this kind of thing for users of the cloud (and not users of traditional<br>
managed hosting) :) Like I said, if it does not affect the performance<br>
of other tenants' instances, I'm fine with adding the functionality in a<br>
way that is generic (not hypervisor-specific).<br>
<br>
Best,<br>
-jay<br>
<div class="HOEnZb"><div class="h5"><br>
<br>
_______________________________________________<br>
OpenStack-dev mailing list<br>
<a href="mailto:OpenStack-dev@lists.openstack.org">OpenStack-dev@lists.openstack.org</a><br>
<a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
</div></div></blockquote></div><br><br clear="all"><br>-- <br><div dir="ltr">Qin Zhao<br></div>
</div>