Xen devel - Jul 2011 - Issues about domU suspending/resuming

Hi, I have looked into the MAINTAINERS file on the xen source, but I
didn''t find the specific maintainer for responsible for issues about
domU suspending/resuming procedure.

Can you point me to the right person?



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 28/07/2011 17:13, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com> wrote:
> Hi, I have looked into the MAINTAINERS file on the xen source, but I didn¹t
> find the specific maintainer for responsible for issues about domU
> suspending/resuming procedure.
> Can you point me to the right person?
It crosses multiple subsystems. It could be a guest kernel bug for example,
or a toolstack bug. Post a bug report to xen-devel and we can try to triage
it.

 -- Keir





_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

It''s not really a bug, but a performance analysis.
I''m using remus for a HA system and after a system analysis using remus
log, I notice that suspending a the guest can oscillate between 0.299ms to
812.909ms. And the resuming of the same guest oscillates between 0.387ms to
1745,579ms. The guest it''s a virtual machine with 64MB of RAM, without
CPU load.

It''s very strange this large range of values of the suspend/resuming of
a guest.
> -----Original Message-----
> From: Keir Fraser [mailto:keir.xen@gmail.com]
> Sent: quinta-feira, 28 de Julho de 2011 18:00
> To: Gustavo Pimentel
> Cc: xen-devel@lists.xensource.com
> Subject: Re: Issues about domU suspending/resuming
> 
> On 28/07/2011 17:13, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com> wrote:
> 
> > Hi, I have looked into the MAINTAINERS file on the xen source, but I
didn¹t
> > find the specific maintainer for responsible for issues about domU
> > suspending/resuming procedure.
> > Can you point me to the right person?
> 
> It crosses multiple subsystems. It could be a guest kernel bug for example,
> or a toolstack bug. Post a bug report to xen-devel and we can try to triage
> it.
> 
>  -- Keir
> 
> 
> 
> 


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On 28/07/2011 18:16, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com> wrote:
> It''s not really a bug, but a performance analysis.
> I''m using remus for a HA system and after a system analysis using
remus log, I
> notice that suspending a the guest can oscillate between 0.299ms to
812.909ms.
> And the resuming of the same guest oscillates between 0.387ms to
1745,579ms.
> The guest it''s a virtual machine with 64MB of RAM, without CPU
load.
> 
> It''s very strange this large range of values of the
suspend/resuming of a
> guest.
First point of contact would be the listed Remus maintainer. More data would
be useful I expect -- e.g., does the variation occur with basic live
migration (no Remus)?

 -- Keir
>> -----Original Message-----
>> From: Keir Fraser [mailto:keir.xen@gmail.com]
>> Sent: quinta-feira, 28 de Julho de 2011 18:00
>> To: Gustavo Pimentel
>> Cc: xen-devel@lists.xensource.com
>> Subject: Re: Issues about domU suspending/resuming
>> 
>> On 28/07/2011 17:13, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com> wrote:
>> 
>>> Hi, I have looked into the MAINTAINERS file on the xen source, but
I didn¹t
>>> find the specific maintainer for responsible for issues about domU
>>> suspending/resuming procedure.
>>> Can you point me to the right person?
>> 
>> It crosses multiple subsystems. It could be a guest kernel bug for
example,
>> or a toolstack bug. Post a bug report to xen-devel and we can try to
triage
>> it.
>> 
>>  -- Keir
>> 
>> 
>> 
>> 
> 
> 


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

I didn''t test the suspend/resuming with basic live migration.
Do you know how can I invoke some xen api in order to suspend/resuming the
guest? I could make a simple program to be able to log times of
suspending/resuming and on the future to analysis the basic live migration.
Meanwhile I leave here some information about the my test system:
I using xen4.2-unstable on a Intel(R) Pentium(R) 4 CPU 3.20GHz processor with
hyperthread, using credit scheduler. I have pinning dom0 to vcpu0 and domU to
vcpu1.
I have on dom0 the kernel version 2.6.32.40 and on domU the kernel version
2.6.18 kernel which has fast suspend support.
> -----Original Message-----
> From: Keir Fraser [mailto:keir.xen@gmail.com]
> Sent: quinta-feira, 28 de Julho de 2011 18:33
> To: Gustavo Pimentel
> Cc: xen-devel@lists.xensource.com
> Subject: Re: Issues about domU suspending/resuming
> 
> On 28/07/2011 18:16, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com> wrote:
> 
> > It''s not really a bug, but a performance analysis.
> > I''m using remus for a HA system and after a system analysis
using remus log, I
> > notice that suspending a the guest can oscillate between 0.299ms to
812.909ms.
> > And the resuming of the same guest oscillates between 0.387ms to
1745,579ms.
> > The guest it''s a virtual machine with 64MB of RAM, without
CPU load.
> >
> > It''s very strange this large range of values of the
suspend/resuming of a
> > guest.
> 
> First point of contact would be the listed Remus maintainer. More data
would
> be useful I expect -- e.g., does the variation occur with basic live
> migration (no Remus)?
> 
>  -- Keir
> 
> >> -----Original Message-----
> >> From: Keir Fraser [mailto:keir.xen@gmail.com]
> >> Sent: quinta-feira, 28 de Julho de 2011 18:00
> >> To: Gustavo Pimentel
> >> Cc: xen-devel@lists.xensource.com
> >> Subject: Re: Issues about domU suspending/resuming
> >>
> >> On 28/07/2011 17:13, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com>
> wrote:
> >>
> >>> Hi, I have looked into the MAINTAINERS file on the xen source,
but I didn¹t
> >>> find the specific maintainer for responsible for issues about
domU
> >>> suspending/resuming procedure.
> >>> Can you point me to the right person?
> >>
> >> It crosses multiple subsystems. It could be a guest kernel bug for
example,
> >> or a toolstack bug. Post a bug report to xen-devel and we can try
to triage
> >> it.
> >>
> >>  -- Keir
> >>
> >>
> >>
> >>
> >
> >
> 
> 


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Perhaps run a program in domU which calls gettimedofday() in a loop and logs
big deltas between successive calls. If the domain is otherwise idle, you
can probbaly set the threshold to only trigger during the suspend phase of
live migration. Then you don''t need to hook into the suspend/resume
parts of
the toolstack.

 -- Keir


On 28/07/2011 19:22, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com> wrote:
> I didn''t test the suspend/resuming with basic live migration.
> Do you know how can I invoke some xen api in order to suspend/resuming the
> guest? I could make a simple program to be able to log times of
> suspending/resuming and on the future to analysis the basic live migration.
> Meanwhile I leave here some information about the my test system:
> I using xen4.2-unstable on a Intel(R) Pentium(R) 4 CPU 3.20GHz processor
with
> hyperthread, using credit scheduler. I have pinning dom0 to vcpu0 and domU
to
> vcpu1.
> I have on dom0 the kernel version 2.6.32.40 and on domU the kernel version
> 2.6.18 kernel which has fast suspend support.
> 
>> -----Original Message-----
>> From: Keir Fraser [mailto:keir.xen@gmail.com]
>> Sent: quinta-feira, 28 de Julho de 2011 18:33
>> To: Gustavo Pimentel
>> Cc: xen-devel@lists.xensource.com
>> Subject: Re: Issues about domU suspending/resuming
>> 
>> On 28/07/2011 18:16, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com> wrote:
>> 
>>> It''s not really a bug, but a performance analysis.
>>> I''m using remus for a HA system and after a system
analysis using remus log,
>>> I
>>> notice that suspending a the guest can oscillate between 0.299ms to
>>> 812.909ms.
>>> And the resuming of the same guest oscillates between 0.387ms to
1745,579ms.
>>> The guest it''s a virtual machine with 64MB of RAM, without
CPU load.
>>> 
>>> It''s very strange this large range of values of the
suspend/resuming of a
>>> guest.
>> 
>> First point of contact would be the listed Remus maintainer. More data
would
>> be useful I expect -- e.g., does the variation occur with basic live
>> migration (no Remus)?
>> 
>>  -- Keir
>> 
>>>> -----Original Message-----
>>>> From: Keir Fraser [mailto:keir.xen@gmail.com]
>>>> Sent: quinta-feira, 28 de Julho de 2011 18:00
>>>> To: Gustavo Pimentel
>>>> Cc: xen-devel@lists.xensource.com
>>>> Subject: Re: Issues about domU suspending/resuming
>>>> 
>>>> On 28/07/2011 17:13, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com>
>> wrote:
>>>> 
>>>>> Hi, I have looked into the MAINTAINERS file on the xen
source, but I
>>>>> didn¹t
>>>>> find the specific maintainer for responsible for issues
about domU
>>>>> suspending/resuming procedure.
>>>>> Can you point me to the right person?
>>>> 
>>>> It crosses multiple subsystems. It could be a guest kernel bug
for example,
>>>> or a toolstack bug. Post a bug report to xen-devel and we can
try to triage
>>>> it.
>>>> 
>>>>  -- Keir
>>>> 
>>>> 
>>>> 
>>>> 
>>> 
>>> 
>> 
>> 
> 
> 


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

I used the attached file (suspend_resume.c) in order to generate a log
interpreted by a "homemade" program wrote by me to create a graph
analysis of the suspend/resume of a guest during 30 minutes with suspend/resume
cycles of 100ms.
If you observe the ChartAnalysis.png you can see during the 30 minutes of the
suspend/resume test of the guest is very consistent, but it happen about 6
spikes (3 of them reaching about 400ms, about 600 to 700 times more than the
normal average). This test as perform with no CPU load on the dom0 and domU. I
leave you here some stats:
 
Minimum=0.468 ms			
Arithmetic Mean=0.652 ms
Geometric Mean=0.552 ms
Maximum=401.251 ms
> -----Original Message-----
> From: Keir Fraser [mailto:keir.xen@gmail.com]
> Sent: quinta-feira, 28 de Julho de 2011 20:07
> To: Gustavo Pimentel
> Cc: xen-devel@lists.xensource.com; rshriram@cs.ubc.ca
> Subject: Re: Issues about domU suspending/resuming
> 
> Perhaps run a program in domU which calls gettimedofday() in a loop and
logs
> big deltas between successive calls. If the domain is otherwise idle, you
> can probbaly set the threshold to only trigger during the suspend phase of
> live migration. Then you don''t need to hook into the
suspend/resume parts of
> the toolstack.
> 
>  -- Keir
> 
> 
> On 28/07/2011 19:22, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com> wrote:
> 
> > I didn''t test the suspend/resuming with basic live migration.
> > Do you know how can I invoke some xen api in order to suspend/resuming
the
> > guest? I could make a simple program to be able to log times of
> > suspending/resuming and on the future to analysis the basic live
migration.
> > Meanwhile I leave here some information about the my test system:
> > I using xen4.2-unstable on a Intel(R) Pentium(R) 4 CPU 3.20GHz
processor with
> > hyperthread, using credit scheduler. I have pinning dom0 to vcpu0 and
domU to
> > vcpu1.
> > I have on dom0 the kernel version 2.6.32.40 and on domU the kernel
version
> > 2.6.18 kernel which has fast suspend support.
> >
> >> -----Original Message-----
> >> From: Keir Fraser [mailto:keir.xen@gmail.com]
> >> Sent: quinta-feira, 28 de Julho de 2011 18:33
> >> To: Gustavo Pimentel
> >> Cc: xen-devel@lists.xensource.com
> >> Subject: Re: Issues about domU suspending/resuming
> >>
> >> On 28/07/2011 18:16, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com>
> wrote:
> >>
> >>> It''s not really a bug, but a performance analysis.
> >>> I''m using remus for a HA system and after a system
analysis using remus log,
> >>> I
> >>> notice that suspending a the guest can oscillate between
0.299ms to
> >>> 812.909ms.
> >>> And the resuming of the same guest oscillates between 0.387ms
to
> 1745,579ms.
> >>> The guest it''s a virtual machine with 64MB of RAM,
without CPU load.
> >>>
> >>> It''s very strange this large range of values of the
suspend/resuming of a
> >>> guest.
> >>
> >> First point of contact would be the listed Remus maintainer. More
data would
> >> be useful I expect -- e.g., does the variation occur with basic
live
> >> migration (no Remus)?
> >>
> >>  -- Keir
> >>
> >>>> -----Original Message-----
> >>>> From: Keir Fraser [mailto:keir.xen@gmail.com]
> >>>> Sent: quinta-feira, 28 de Julho de 2011 18:00
> >>>> To: Gustavo Pimentel
> >>>> Cc: xen-devel@lists.xensource.com
> >>>> Subject: Re: Issues about domU suspending/resuming
> >>>>
> >>>> On 28/07/2011 17:13, "Gustavo Pimentel"
<gustavo.pimentel@efacec.com>
> >> wrote:
> >>>>
> >>>>> Hi, I have looked into the MAINTAINERS file on the xen
source, but I
> >>>>> didn¹t
> >>>>> find the specific maintainer for responsible for
issues about domU
> >>>>> suspending/resuming procedure.
> >>>>> Can you point me to the right person?
> >>>>
> >>>> It crosses multiple subsystems. It could be a guest kernel
bug for example,
> >>>> or a toolstack bug. Post a bug report to xen-devel and we
can try to triage
> >>>> it.
> >>>>
> >>>>  -- Keir
> >>>>
> >>>>
> >>>>
> >>>>
> >>>
> >>>
> >>
> >>
> >
> >
> 
> 



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel