Backport requests of cs 23420..23423 for 4.0 and 4.1

Backport requests of cs 23420..23423 for 4.0 and 4.1

[Juergen Gross]

Hi,

we are experiencing significant performance degradation after live migration of
hvm domains in Xen 4.0 (SLES11 SP1): after live migration the performance is
dropping to less than 90%. I did a backport of cs 23420-23423 and the
performance is okay now.

I would like to request to include these changesets in 4.0 and 4.1. The
backport is quite trivial, I can send patches if you are willing to take them.


Juergen

-- 
Juergen Gross                 Principal Developer Operating Systems
PDG ES&S SWE OS6                       Telephone: +49 (0) 89 3222 2967
Fujitsu Technology Solutions              e-mail: juergen.gross@ts.fujitsu.com
Domagkstr. 28                           Internet: ts.fujitsu.com
D-80807 Muenchen                 Company details: ts.fujitsu.com/imprint.html

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Keir Fraser]

On 13/07/2012 07:08, "Juergen Gross" <juergen.gross@ts.fujitsu.com> wrote:

> Hi,
> 
> we are experiencing significant performance degradation after live migration
> of
> hvm domains in Xen 4.0 (SLES11 SP1): after live migration the performance is
> dropping to less than 90%. I did a backport of cs 23420-23423 and the
> performance is okay now.
> 
> I would like to request to include these changesets in 4.0 and 4.1. The
> backport is quite trivial, I can send patches if you are willing to take them.

Will need an Ack from George and then patches applied by (or at least an Ack
from) a tools maintainer.

 -- Keir

> Juergen

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Juergen Gross]

Am 13.07.2012 10:05, schrieb Keir Fraser:
> On 13/07/2012 07:08, "Juergen Gross"<juergen.gross@ts.fujitsu.com>  wrote:
>
>> Hi,
>>
>> we are experiencing significant performance degradation after live migration
>> of
>> hvm domains in Xen 4.0 (SLES11 SP1): after live migration the performance is
>> dropping to less than 90%. I did a backport of cs 23420-23423 and the
>> performance is okay now.
>>
>> I would like to request to include these changesets in 4.0 and 4.1. The
>> backport is quite trivial, I can send patches if you are willing to take them.
>
> Will need an Ack from George and then patches applied by (or at least an Ack
> from) a tools maintainer.

Okay, I'll send the patches for 4.1 and 4.0 and George can Ack or Nack them.


Thanks, Juergen

-- 
Juergen Gross                 Principal Developer Operating Systems
PDG ES&S SWE OS6                       Telephone: +49 (0) 89 3222 2967
Fujitsu Technology Solutions              e-mail: juergen.gross@ts.fujitsu.com
Domagkstr. 28                           Internet: ts.fujitsu.com
D-80807 Muenchen                 Company details: ts.fujitsu.com/imprint.html

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Andrew Cooper]

On 13/07/12 10:26, Juergen Gross wrote:
> Am 13.07.2012 10:05, schrieb Keir Fraser:
>> On 13/07/2012 07:08, "Juergen Gross"<juergen.gross@ts.fujitsu.com>  wrote:
>>
>>> Hi,
>>>
>>> we are experiencing significant performance degradation after live migration
>>> of
>>> hvm domains in Xen 4.0 (SLES11 SP1): after live migration the performance is
>>> dropping to less than 90%. I did a backport of cs 23420-23423 and the
>>> performance is okay now.
>>>
>>> I would like to request to include these changesets in 4.0 and 4.1. The
>>> backport is quite trivial, I can send patches if you are willing to take them.
>> Will need an Ack from George and then patches applied by (or at least an Ack
>> from) a tools maintainer.
> Okay, I'll send the patches for 4.1 and 4.0 and George can Ack or Nack them.
>
>
> Thanks, Juergen
>

For what it is worth, XenServer have had these patches (the
pre-upstreamed versions) in all Xen-4.x releases we have done.  We even
had a customer support escalation and backported them to our last
Xen-3.4 release.

-- 
Andrew Cooper - Dom0 Kernel Engineer, Citrix XenServer
T: +44 (0)1223 225 900, http://www.citrix.com

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Keir Fraser]

On 13/07/2012 11:07, "Andrew Cooper" <andrew.cooper3@citrix.com> wrote:

> On 13/07/12 10:26, Juergen Gross wrote:
>> Am 13.07.2012 10:05, schrieb Keir Fraser:
>>> On 13/07/2012 07:08, "Juergen Gross"<juergen.gross@ts.fujitsu.com>  wrote:
>>> 
>>>> Hi,
>>>> 
>>>> we are experiencing significant performance degradation after live
>>>> migration
>>>> of
>>>> hvm domains in Xen 4.0 (SLES11 SP1): after live migration the performance
>>>> is
>>>> dropping to less than 90%. I did a backport of cs 23420-23423 and the
>>>> performance is okay now.
>>>> 
>>>> I would like to request to include these changesets in 4.0 and 4.1. The
>>>> backport is quite trivial, I can send patches if you are willing to take
>>>> them.
>>> Will need an Ack from George and then patches applied by (or at least an Ack
>>> from) a tools maintainer.
>> Okay, I'll send the patches for 4.1 and 4.0 and George can Ack or Nack them.

I haven't seen anything from George or Ian. These are likely to not make it
for the next 4.0/4.1 point releases.

 -- Keir

>> Thanks, Juergen
>> 
> 
> For what it is worth, XenServer have had these patches (the
> pre-upstreamed versions) in all Xen-4.x releases we have done.  We even
> had a customer support escalation and backported them to our last
> Xen-3.4 release.

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Ian Jackson]

Keir Fraser writes ("Re: [Xen-devel] Backport requests of cs 23420..23423 for 4.0 and 4.1"):
> On 13/07/2012 07:08, "Juergen Gross" <juergen.gross@ts.fujitsu.com> wrote:
> > I would like to request to include these changesets in 4.0 and
> > 4.1. The backport is quite trivial, I can send patches if you are
> > willing to take them.
> 
> Will need an Ack from George and then patches applied by (or at least an Ack
> from) a tools maintainer.

Thanks for replying Keir, but I'm rather queasy about this.

These patches have not been in any released version of Xen and are
fairly substantial.  I would say that we should not backport anything
that isn't a critical bugfix which hasn't been sitting in a released
version of Xen for a while; and a new feature ought to be considered
very carefully.

Now maybe the unfortunately extended 4.2 release cycle may mean we
should relax this rule but I'd prefer to see a clear justification for
why this is important to retrofit to 4.1.

And the Xen 4.0 tree is in the deep freeze and I don't think we should
be backporting anything other than critical bugfixes for it.

Ian.

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Keir Fraser]

On 19/07/2012 16:11, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:

> Keir Fraser writes ("Re: [Xen-devel] Backport requests of cs 23420..23423 for
> 4.0 and 4.1"):
>> On 13/07/2012 07:08, "Juergen Gross" <juergen.gross@ts.fujitsu.com> wrote:
>>> I would like to request to include these changesets in 4.0 and
>>> 4.1. The backport is quite trivial, I can send patches if you are
>>> willing to take them.
>> 
>> Will need an Ack from George and then patches applied by (or at least an Ack
>> from) a tools maintainer.
> 
> Thanks for replying Keir, but I'm rather queasy about this.
> 
> These patches have not been in any released version of Xen and are
> fairly substantial.  I would say that we should not backport anything
> that isn't a critical bugfix which hasn't been sitting in a released
> version of Xen for a while; and a new feature ought to be considered
> very carefully.

They have been in use in Citrix XenServer for a while. However, this is
still late in stable point-release cycle to drop these in. It might make
sense to have them in 4.2.0 for a while, and then reconsider for backport
for 4.1.4 (I think you are probably right to nack for the 4.0 branch).

 -- Keir

> Now maybe the unfortunately extended 4.2 release cycle may mean we
> should relax this rule but I'd prefer to see a clear justification for
> why this is important to retrofit to 4.1.
> 
> And the Xen 4.0 tree is in the deep freeze and I don't think we should
> be backporting anything other than critical bugfixes for it.
> 
> Ian.

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Pasi Kärkkäinen]

On Thu, Jul 19, 2012 at 04:23:43PM +0100, Keir Fraser wrote:
> On 19/07/2012 16:11, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:
> 
> > Keir Fraser writes ("Re: [Xen-devel] Backport requests of cs 23420..23423 for
> > 4.0 and 4.1"):
> >> On 13/07/2012 07:08, "Juergen Gross" <juergen.gross@ts.fujitsu.com> wrote:
> >>> I would like to request to include these changesets in 4.0 and
> >>> 4.1. The backport is quite trivial, I can send patches if you are
> >>> willing to take them.
> >> 
> >> Will need an Ack from George and then patches applied by (or at least an Ack
> >> from) a tools maintainer.
> > 
> > Thanks for replying Keir, but I'm rather queasy about this.
> > 
> > These patches have not been in any released version of Xen and are
> > fairly substantial.  I would say that we should not backport anything
> > that isn't a critical bugfix which hasn't been sitting in a released
> > version of Xen for a while; and a new feature ought to be considered
> > very carefully.
> 
> They have been in use in Citrix XenServer for a while. However, this is
> still late in stable point-release cycle to drop these in. It might make
> sense to have them in 4.2.0 for a while, and then reconsider for backport
> for 4.1.4 (I think you are probably right to nack for the 4.0 branch).
> 

Hello,

Should these patches now go in to 4.1.4-rc1 ? 


-- Pasi


>  -- Keir
> 
> > Now maybe the unfortunately extended 4.2 release cycle may mean we
> > should relax this rule but I'd prefer to see a clear justification for
> > why this is important to retrofit to 4.1.
> > 
> > And the Xen 4.0 tree is in the deep freeze and I don't think we should
> > be backporting anything other than critical bugfixes for it.
> > 
> > Ian.
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Juergen Gross]

Am 19.07.2012 17:11, schrieb Ian Jackson:
> Keir Fraser writes ("Re: [Xen-devel] Backport requests of cs 23420..23423 for 4.0 and 4.1"):
>> On 13/07/2012 07:08, "Juergen Gross"<juergen.gross@ts.fujitsu.com>  wrote:
>>> I would like to request to include these changesets in 4.0 and
>>> 4.1. The backport is quite trivial, I can send patches if you are
>>> willing to take them.
>>
>> Will need an Ack from George and then patches applied by (or at least an Ack
>> from) a tools maintainer.
>
> Thanks for replying Keir, but I'm rather queasy about this.
>
> These patches have not been in any released version of Xen and are
> fairly substantial.  I would say that we should not backport anything
> that isn't a critical bugfix which hasn't been sitting in a released
> version of Xen for a while; and a new feature ought to be considered
> very carefully.
>
> Now maybe the unfortunately extended 4.2 release cycle may mean we
> should relax this rule but I'd prefer to see a clear justification for
> why this is important to retrofit to 4.1.

Live migration is a main high-availability feature of our next release.

A performance degradation of 10% and more will not be easily accepted for
a system which is expected to be up 24/7.

I assume there is a reason the patches are in XCP right now. :-)


Juergen

-- 
Juergen Gross                 Principal Developer Operating Systems
PDG ES&S SWE OS6                       Telephone: +49 (0) 89 3222 2967
Fujitsu Technology Solutions              e-mail: juergen.gross@ts.fujitsu.com
Domagkstr. 28                           Internet: ts.fujitsu.com
D-80807 Muenchen                 Company details: ts.fujitsu.com/imprint.html

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Andrew Cooper]

On 20/07/12 06:54, Juergen Gross wrote:
> Am 19.07.2012 17:11, schrieb Ian Jackson:
>> Keir Fraser writes ("Re: [Xen-devel] Backport requests of cs 23420..23423 for 4.0 and 4.1"):
>>> On 13/07/2012 07:08, "Juergen Gross"<juergen.gross@ts.fujitsu.com>  wrote:
>>>> I would like to request to include these changesets in 4.0 and
>>>> 4.1. The backport is quite trivial, I can send patches if you are
>>>> willing to take them.
>>> Will need an Ack from George and then patches applied by (or at least an Ack
>>> from) a tools maintainer.
>> Thanks for replying Keir, but I'm rather queasy about this.
>>
>> These patches have not been in any released version of Xen and are
>> fairly substantial.  I would say that we should not backport anything
>> that isn't a critical bugfix which hasn't been sitting in a released
>> version of Xen for a while; and a new feature ought to be considered
>> very carefully.
>>
>> Now maybe the unfortunately extended 4.2 release cycle may mean we
>> should relax this rule but I'd prefer to see a clear justification for
>> why this is important to retrofit to 4.1.
> Live migration is a main high-availability feature of our next release.
>
> A performance degradation of 10% and more will not be easily accepted for
> a system which is expected to be up 24/7.
>
> I assume there is a reason the patches are in XCP right now. :-)
>
>
> Juergen
>

XCP is basically XenServer minus the propriety stuff, so it uses 'our'
Xen amongst other things.

-- 
Andrew Cooper - Dom0 Kernel Engineer, Citrix XenServer
T: +44 (0)1223 225 900, http://www.citrix.com

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[George Dunlap]

On 19/07/12 22:54, Juergen Gross wrote:
> Am 19.07.2012 17:11, schrieb Ian Jackson:
>> Keir Fraser writes ("Re: [Xen-devel] Backport requests of cs 23420..23423 for 4.0 and 4.1"):
>>> On 13/07/2012 07:08, "Juergen Gross"<juergen.gross@ts.fujitsu.com>  wrote:
>>>> I would like to request to include these changesets in 4.0 and
>>>> 4.1. The backport is quite trivial, I can send patches if you are
>>>> willing to take them.
>>> Will need an Ack from George and then patches applied by (or at least an Ack
>>> from) a tools maintainer.
>> Thanks for replying Keir, but I'm rather queasy about this.
>>
>> These patches have not been in any released version of Xen and are
>> fairly substantial.  I would say that we should not backport anything
>> that isn't a critical bugfix which hasn't been sitting in a released
>> version of Xen for a while; and a new feature ought to be considered
>> very carefully.
>>
>> Now maybe the unfortunately extended 4.2 release cycle may mean we
>> should relax this rule but I'd prefer to see a clear justification for
>> why this is important to retrofit to 4.1.
> Live migration is a main high-availability feature of our next release.
>
> A performance degradation of 10% and more will not be easily accepted for
> a system which is expected to be up 24/7.
Is there a reason you can't just do as XenServer and XCP have done, and 
have them in a local patch queue?

Obviously it's better to keep a local patch queue as short as possible, 
but it doesn't seem like you're really going to be that crippled if we 
wait to check them in.

  -George

Re: Backport requests of cs 23420..23423 for 4.0 and 4.1

[Juergen Gross]

Am 20.07.2012 21:36, schrieb George Dunlap:
> On 19/07/12 22:54, Juergen Gross wrote:
>> Am 19.07.2012 17:11, schrieb Ian Jackson:
>>> Keir Fraser writes ("Re: [Xen-devel] Backport requests of cs
>>> 23420..23423 for 4.0 and 4.1"):
>>>> On 13/07/2012 07:08, "Juergen Gross"<juergen.gross@ts.fujitsu.com>
>>>> wrote:
>>>>> I would like to request to include these changesets in 4.0 and
>>>>> 4.1. The backport is quite trivial, I can send patches if you are
>>>>> willing to take them.
>>>> Will need an Ack from George and then patches applied by (or at
>>>> least an Ack
>>>> from) a tools maintainer.
>>> Thanks for replying Keir, but I'm rather queasy about this.
>>>
>>> These patches have not been in any released version of Xen and are
>>> fairly substantial. I would say that we should not backport anything
>>> that isn't a critical bugfix which hasn't been sitting in a released
>>> version of Xen for a while; and a new feature ought to be considered
>>> very carefully.
>>>
>>> Now maybe the unfortunately extended 4.2 release cycle may mean we
>>> should relax this rule but I'd prefer to see a clear justification for
>>> why this is important to retrofit to 4.1.
>> Live migration is a main high-availability feature of our next release.
>>
>> A performance degradation of 10% and more will not be easily accepted for
>> a system which is expected to be up 24/7.
> Is there a reason you can't just do as XenServer and XCP have done, and
> have them in a local patch queue?

That's plan C. :-)

We are using SLES as base, so plan B is asking Suse...

> Obviously it's better to keep a local patch queue as short as possible,
> but it doesn't seem like you're really going to be that crippled if we
> wait to check them in.

Correct.

I had to try. :-)


Juergen

-- 
Juergen Gross                 Principal Developer Operating Systems
PDG ES&S SWE OS6                       Telephone: +49 (0) 89 3222 2967
Fujitsu Technology Solutions              e-mail: juergen.gross@ts.fujitsu.com
Domagkstr. 28                           Internet: ts.fujitsu.com
D-80807 Muenchen                 Company details: ts.fujitsu.com/imprint.html