Upgrade to latest stable version v1.3.3

[Thiago Martins]

Hey guys,

 Is it possible to upgrade Libvirt to v1.3.2 for Xenial?

 It have many bug fixes an improvements for Xen XL driver!

Cheers!
Thiago

[Serge Hallyn (serge-hallyn)]

It's very late in the cycle for a version update.

Assigning this to Stefan to gauge the importance of the xen improvements.

[Thiago Martins (martinx)]

Mmm... Okay... APT is still receiving cycle updates... ;-)

Also, OpenStack with Open Soruce Xen, recommends Libvirt 1.3.2! I think
that this is something important for an LTS like Xenial.

On 17 March 2016 at 23:06, Serge Hallyn <email address hidden> wrote:

> It's very late in the cycle for a version update.
>
> Assigning this to Stefan to gauge the importance of the xen
> improvements.
>
> ** Changed in: libvirt (Ubuntu)
> Assignee: (unassigned) => Stefan Bader (smb)
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1558857
>
> Title:
> Upgrade to latest stable version v1.3.2
>
> To manage notifications about this bug go to:
>
> https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/1558857/+subscriptions
>

[Serge Hallyn (serge-hallyn)]

Thanks for the input. Specific reasons like that are valuable.

[Stefan Bader (smb)]

Well as it stands right now, at least the (admittedly more basic things) stuff I do run is working with 1.3.1. So I cannot add any exciting arguments. And upstream tends to always recommend "the latest stuff", so next week this could be 1.3.3. ;)

Technically it looks possible but with the risk of being beyond Debian right now. Which has a few challenges on the packaging side (mostly to ensure it is clear where things came from) but also the normal approach for a LTS release is to be a little more conservative about latest and greatest as that also means not that well tested/settled.

So I personally would be hesitant unless there are even more specific reasons like 1.3.2 allows to do x and that will be important because of this reason. The problem is/was that newer not only means fixes but also new breakage. So I rather stick to a reasonable pace and want to see very good reasons to move fast(er) (besides not being the runner type anyways ;)).

[Simon Déziel (sdeziel)]

As mentioned in LP: #1553023, this new release brings ZFS support which would be really nice to get now that everyone wants to jump on the ZFS train :)

IIRC, upstream libvirt releases every 3 months so that would be the last FFe for Xenial.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Thiago Martins (martinx)]

I agree that it is important to be conservative, nevertheless, take a look
here:

--
"Although libvirt 1.2.15 works with Xen, libvirt 1.3.2 or newer is
recommended."

"We do however recommend using libvirt 1.3.2, which is fully supported and
tested as part of the Xen Project CI loop. It addresses live migration
monitoring related issues and adds support for peer-to-peer migration mode,
which nova relies on."

https://github.com/openstack/openstack-manuals/blob/master/doc/config-reference/source/compute/hypervisor-xen-libvirt.rst
--

--
Another example, Mesa is 11.1 but, Xorg team is planning to upgrade it to
11.2:
https://launchpad.net/~canonical-x/+archive/ubuntu/x-staging
--

Cheers!
Thiago

On 18 March 2016 at 10:49, Launchpad Bug Tracker <<email address hidden>
> wrote:

> Status changed to 'Confirmed' because the bug affects multiple users.
>
> ** Changed in: libvirt (Ubuntu)
> Status: New => Confirmed
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1558857
>
> Title:
> Upgrade to latest stable version v1.3.2
>
> To manage notifications about this bug go to:
>
> https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/1558857/+subscriptions
>

[Serge Hallyn (serge-hallyn)]

Quoting Simon Déziel (<email address hidden>):
> As mentioned in LP: #1553023, this new release brings ZFS support which
> would be really nice to get now that everyone wants to jump on the ZFS
> train :)

We already have zfs cherrypicked into 1.3.1 in xenial.

[Serge Hallyn (serge-hallyn)]

Quoting Stefan Bader (<email address hidden>):
> Well as it stands right now, at least the (admittedly more basic things)
> stuff I do run is working with 1.3.1. So I cannot add any exciting
> arguments. And upstream tends to always recommend "the latest stuff", so
> next week this could be 1.3.3. ;)

Right, that's my feeling.

> Technically it looks possible but with the risk of being beyond Debian
> right now. Which has a few challenges on the packaging side (mostly to
> ensure it is clear where things came from) but also the normal approach
> for a LTS release is to be a little more conservative about latest and
> greatest as that also means not that well tested/settled.
>
> So I personally would be hesitant unless there are even more specific

Right, wanted to make sure you didn't know of any current bad xen issues.

> reasons like 1.3.2 allows to do x and that will be important because of
> this reason. The problem is/was that newer not only means fixes but also
> new breakage. So I rather stick to a reasonable pace and want to see
> very good reasons to move fast(er) (besides not being the runner type
> anyways ;)).

Agreed. Thanks, Stefan.

[Serge Hallyn (serge-hallyn)]

Quoting Thiago Martins (<email address hidden>):
> I agree that it is important to be conservative, nevertheless, take a look
> here:
>
> --
> "Although libvirt 1.2.15 works with Xen, libvirt 1.3.2 or newer is
> recommended."

xenial is on 1.3.1, not 1.2.15

I'm afraid we'll need pointers to specific missing feature to do an update
this late. (for instance, "nova really needs the xl cmdline= support")

I'm just not seeing that much in the changelog.

[Thiago Martins (martinx)]

On 21 March 2016 at 05:22, Serge Hallyn <email address hidden> wrote:

> Quoting Thiago Martins (<email address hidden>):
> > I agree that it is important to be conservative, nevertheless, take a
> look
> > here:
> >
> > --
> > "Although libvirt 1.2.15 works with Xen, libvirt 1.3.2 or newer is
> > recommended."
>
> xenial is on 1.3.1, not 1.2.15
>
> I'm afraid we'll need pointers to specific missing feature to do an update
> this late. (for instance, "nova really needs the xl cmdline= support")
>
> I'm just not seeing that much in the changelog.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1558857
>
> Title:
> Upgrade to latest stable version v1.3.2
>

Sure, I know Xenial have 1.3.1... Up to you guys to decide this! If you say
that Libvirt's changelog doesn't have too much changes between 1.3.1 and
1.3.2, then, I'm good to close this bug has "won't fix" or, maybe, keep it
for Xenial 16.04.1...

Cheers!

[Serge Hallyn (serge-hallyn)]

Hi Thiago,

yeah we may as well keep it open for first thing in 16.10.

[Thiago Martins (martinx)]

Sounds awesome! Maybe we'll have it on Xenial via next Ubuntu Cloud
Archive... :-D

On 22 March 2016 at 14:39, Serge Hallyn <email address hidden> wrote:

> Hi Thiago,
>
> yeah we may as well keep it open for first thing in 16.10.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1558857
>
> Title:
> Upgrade to latest stable version v1.3.2
>
> To manage notifications about this bug go to:
>
> https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/1558857/+subscriptions
>

[Thiago Martins (martinx)]

Guys,

 I'm curious about this...

 * Mesa will be upgraded from 11.1, to 11.2... Big changes...

 * MySQL just got upgraded from 5.6, to 5.7!

 * APT was upgraded from 1.2.5, to 1.2.7...

 All very recent upgrades that just happened on Xenial...

 So, why this big resistance to upgrade Libvirt from 1.3.1, to 1.3.2? Most recent Libvirt 1.3.3 is already on Debian!

 https://tracker.debian.org/pkg/libvirt

 Anyway, I'm just curious about why Libvirt can not receive a last minute update, while many other packages can! Why is that?

Cheers!
Thiago

[Serge Hallyn (serge-hallyn)]

Quoting Thiago Martins (<email address hidden>):
> Guys,
>
> I'm curious about this...
>
> * Mesa will be upgraded from 11.1, to 11.2... Big changes...
>
> * MySQL just got upgraded from 5.6, to 5.7!
>
> * APT was upgraded from 1.2.5, to 1.2.7...
>
> All very recent upgrades that just happened on Xenial...
>
> So, why this big resistance to upgrade Libvirt from 1.3.1, to 1.3.2?
> Most recent Libvirt 1.3.3 is already on Debian!

And not building.

> https://tracker.debian.org/pkg/libvirt
>
> Anyway, I'm just curious about why Libvirt can not receive a last
> minute update, while many other packages can! Why is that?

Because there is no good reason to do so, and the current version
seems to be working well. We were open to it and considered it.
If debian had 1.3.2 packaged we probably would have merged it,
but we didn't want to risk having different .orig tarballs.
If 1.3.3 (was stable and) had great benefits, we would consider
merging it. But finalfreeze is now 3 days away.

[Thiago Martins (martinx)]

On 11 April 2016 at 12:04, Serge Hallyn <email address hidden> wrote:

> Quoting Thiago Martins (<email address hidden>):
> > Guys,
> >
> > I'm curious about this...
> >
> > * Mesa will be upgraded from 11.1, to 11.2... Big changes...
> >
> > * MySQL just got upgraded from 5.6, to 5.7!
> >
> > * APT was upgraded from 1.2.5, to 1.2.7...
> >
> > All very recent upgrades that just happened on Xenial...
> >
> > So, why this big resistance to upgrade Libvirt from 1.3.1, to 1.3.2?
> > Most recent Libvirt 1.3.3 is already on Debian!
>
> And not building.
>
> > https://tracker.debian.org/pkg/libvirt
> >
> > Anyway, I'm just curious about why Libvirt can not receive a last
> > minute update, while many other packages can! Why is that?
>
> Because there is no good reason to do so, and the current version
> seems to be working well. We were open to it and considered it.
> If debian had 1.3.2 packaged we probably would have merged it,
> but we didn't want to risk having different .orig tarballs.
> If 1.3.3 (was stable and) had great benefits, we would consider
> merging it. But finalfreeze is now 3 days away.

Thanks for your reply Serge!