rhel6 guest fail to do save?

rhel6 guest fail to do save?

[Zhang, Yang Z]

Anyone had tried the save/restore against rhel6 guest? I found it always throw timeout when I tried to save the guest. And only fail with rhel6 guest.(ubuntu, fedora are ok).

best regards
yang

Re: rhel6 guest fail to do save?

[Pasi Kärkkäinen]

On Wed, Jan 12, 2011 at 10:15:59AM +0800, Zhang, Yang Z wrote:
> Anyone had tried the save/restore against rhel6 guest? I found it always throw timeout when I tried to save the guest. And only fail with rhel6 guest.(ubuntu, fedora are ok).
> 

I think people have generally been able to save/restore with rhel6,
but there are some issues with live migration between different MHz CPUs,
for example: https://bugzilla.redhat.com/show_bug.cgi?id=663755

Did you search the Redhat bugzilla for save/restore issues?

btw I assume you're talking about rhel6 PV domU..?

-- Pasi

RE: rhel6 guest fail to do save?

[Zhang, Yang Z]

> -----Original Message-----
> From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> Sent: Wednesday, January 12, 2011 3:37 PM
> To: Zhang, Yang Z
> Cc: xen-devel@lists.xensource.com
> Subject: Re: [Xen-devel] rhel6 guest fail to do save?
> 
> On Wed, Jan 12, 2011 at 10:15:59AM +0800, Zhang, Yang Z wrote:
> > Anyone had tried the save/restore against rhel6 guest? I found it always
> throw timeout when I tried to save the guest. And only fail with rhel6
> guest.(ubuntu, fedora are ok).
> >
> 
> I think people have generally been able to save/restore with rhel6,
> but there are some issues with live migration between different MHz CPUs,
> for example: https://bugzilla.redhat.com/show_bug.cgi?id=663755
> 
> Did you search the Redhat bugzilla for save/restore issues?
> 
> btw I assume you're talking about rhel6 PV domU..?
Sorry, I mean the HVM.


best regards
yang
> -- Pasi

RE: rhel6 guest fail to do save?

[Ian Jackson]

Zhang, Yang Z writes ("RE: [Xen-devel] rhel6 guest fail to do save?"):
> From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> > I think people have generally been able to save/restore with rhel6,
> > but there are some issues with live migration between different MHz CPUs,
> > for example: https://bugzilla.redhat.com/show_bug.cgi?id=663755
...
> Sorry, I mean the HVM.

This is known not to work.  It fails in my automatic tests.

The current RHEL6 kernel has enough pv-on-hvm drivers to get the
toolstack to try to negotiate with the guest during save, but not
enough (or not working enough) drivers to actually complete the
negotiation.  So the toolstack times out and the migrate/save fails.

Ian.

Re: rhel6 guest fail to do save?

[Pasi Kärkkäinen]

On Wed, Jan 12, 2011 at 05:16:12PM +0000, Ian Jackson wrote:
> Zhang, Yang Z writes ("RE: [Xen-devel] rhel6 guest fail to do save?"):
> > From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> > > I think people have generally been able to save/restore with rhel6,
> > > but there are some issues with live migration between different MHz CPUs,
> > > for example: https://bugzilla.redhat.com/show_bug.cgi?id=663755
> ...
> > Sorry, I mean the HVM.
> 
> This is known not to work.  It fails in my automatic tests.
> 
> The current RHEL6 kernel has enough pv-on-hvm drivers to get the
> toolstack to try to negotiate with the guest during save, but not
> enough (or not working enough) drivers to actually complete the
> negotiation.  So the toolstack times out and the migrate/save fails.
> 

Ok. 

Zhang: Would you like to open an entry to Redhat bugzilla 
about the RHEL6 PVHVM save/restore problem? 

-- Pasi

RE: rhel6 guest fail to do save?

[Zhang, Yang Z]

> -----Original Message-----
> From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> Sent: Thursday, January 13, 2011 3:38 AM
> To: Ian Jackson
> Cc: Zhang, Yang Z; xen-devel@lists.xensource.com
> Subject: Re: [Xen-devel] rhel6 guest fail to do save?
> 
> On Wed, Jan 12, 2011 at 05:16:12PM +0000, Ian Jackson wrote:
> > Zhang, Yang Z writes ("RE: [Xen-devel] rhel6 guest fail to do save?"):
> > > From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> > > > I think people have generally been able to save/restore with rhel6,
> > > > but there are some issues with live migration between different MHz
> CPUs,
> > > > for example: https://bugzilla.redhat.com/show_bug.cgi?id=663755
> > ...
> > > Sorry, I mean the HVM.
> >
> > This is known not to work.  It fails in my automatic tests.
> >
> > The current RHEL6 kernel has enough pv-on-hvm drivers to get the
> > toolstack to try to negotiate with the guest during save, but not
> > enough (or not working enough) drivers to actually complete the
> > negotiation.  So the toolstack times out and the migrate/save fails.
> >
> 
> Ok.
> 
> Zhang: Would you like to open an entry to Redhat bugzilla
> about the RHEL6 PVHVM save/restore problem?
sure
here is the link:
https://bugzilla.redhat.com/show_bug.cgi?id=669252

best regards
yang
> -- Pasi

RE: rhel6 guest fail to do save?

[Ian Campbell]

On Wed, 2011-01-12 at 17:16 +0000, Ian Jackson wrote: 
> Zhang, Yang Z writes ("RE: [Xen-devel] rhel6 guest fail to do save?"):
> > From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> > > I think people have generally been able to save/restore with rhel6,
> > > but there are some issues with live migration between different MHz CPUs,
> > > for example: https://bugzilla.redhat.com/show_bug.cgi?id=663755
> ...
> > Sorry, I mean the HVM.
> 
> This is known not to work.  It fails in my automatic tests.
> 
> The current RHEL6 kernel has enough pv-on-hvm drivers to get the
> toolstack to try to negotiate with the guest during save, but not
> enough (or not working enough) drivers to actually complete the
> negotiation.  So the toolstack times out and the migrate/save fails.

Does any of the pv-on-hvm functionality work at all in RHEL6?

Perhaps setting "platform_pci = 0" in the guest configuration would work
around this issue? In particular if RHEL6 has no useful pv-on-hvm
functionality in the first place then nothing is lost.

Ian.

Re: rhel6 guest fail to do save?

[Pasi Kärkkäinen]

On Thu, Jan 13, 2011 at 07:36:02AM +0000, Ian Campbell wrote:
> On Wed, 2011-01-12 at 17:16 +0000, Ian Jackson wrote: 
> > Zhang, Yang Z writes ("RE: [Xen-devel] rhel6 guest fail to do save?"):
> > > From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> > > > I think people have generally been able to save/restore with rhel6,
> > > > but there are some issues with live migration between different MHz CPUs,
> > > > for example: https://bugzilla.redhat.com/show_bug.cgi?id=663755
> > ...
> > > Sorry, I mean the HVM.
> > 
> > This is known not to work.  It fails in my automatic tests.
> > 
> > The current RHEL6 kernel has enough pv-on-hvm drivers to get the
> > toolstack to try to negotiate with the guest during save, but not
> > enough (or not working enough) drivers to actually complete the
> > negotiation.  So the toolstack times out and the migrate/save fails.
> 
> Does any of the pv-on-hvm functionality work at all in RHEL6?
> 
> Perhaps setting "platform_pci = 0" in the guest configuration would work
> around this issue? In particular if RHEL6 has no useful pv-on-hvm
> functionality in the first place then nothing is lost.
> 

RHEL6 is supposed to have working PV-on-HVM drivers,
based on Stefano's patches.

Redhat has tested them with RHEL5 dom0, afaik.

-- Pasi

Re: rhel6 guest fail to do save?

[Don Dutile]

Pasi Kärkkäinen wrote:
> On Thu, Jan 13, 2011 at 07:36:02AM +0000, Ian Campbell wrote:
>> On Wed, 2011-01-12 at 17:16 +0000, Ian Jackson wrote: 
>>> Zhang, Yang Z writes ("RE: [Xen-devel] rhel6 guest fail to do save?"):
>>>> From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
>>>>> I think people have generally been able to save/restore with rhel6,
>>>>> but there are some issues with live migration between different MHz CPUs,
>>>>> for example: https://bugzilla.redhat.com/show_bug.cgi?id=663755
>>> ...
>>>> Sorry, I mean the HVM.
>>> This is known not to work.  It fails in my automatic tests.
>>>
>>> The current RHEL6 kernel has enough pv-on-hvm drivers to get the
>>> toolstack to try to negotiate with the guest during save, but not
>>> enough (or not working enough) drivers to actually complete the
>>> negotiation.  So the toolstack times out and the migrate/save fails.
>> Does any of the pv-on-hvm functionality work at all in RHEL6?
>>
>> Perhaps setting "platform_pci = 0" in the guest configuration would work
>> around this issue? In particular if RHEL6 has no useful pv-on-hvm
>> functionality in the first place then nothing is lost.
>>
> 
> RHEL6 is supposed to have working PV-on-HVM drivers,
> based on Stefano's patches.
> 
> Redhat has tested them with RHEL5 dom0, afaik.
> 
Correct. xen-emul-unplug doesn't work, though, due to late
patch that broke that bit of it.
A fair number of upstream patches for pv-hvm & save/restore
have been backported to the next rhel6 update, so hoping this
state of missing pieces in rhel6 pv-on-hvm will be short-lived.
oh, added benefit: the patches should improve rhel6 (pv-on)hvm
on non-RHEL5 dom0's as well.


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

Re: rhel6 guest fail to do save?

[Pasi Kärkkäinen]

On Thu, Jan 13, 2011 at 04:22:36PM -0500, Don Dutile wrote:
> Pasi Kärkkäinen wrote:
> > On Thu, Jan 13, 2011 at 07:36:02AM +0000, Ian Campbell wrote:
> >> On Wed, 2011-01-12 at 17:16 +0000, Ian Jackson wrote: 
> >>> Zhang, Yang Z writes ("RE: [Xen-devel] rhel6 guest fail to do save?"):
> >>>> From: Pasi Kärkkäinen [mailto:pasik@iki.fi]
> >>>>> I think people have generally been able to save/restore with rhel6,
> >>>>> but there are some issues with live migration between different MHz CPUs,
> >>>>> for example: https://bugzilla.redhat.com/show_bug.cgi?id=663755
> >>> ...
> >>>> Sorry, I mean the HVM.
> >>> This is known not to work.  It fails in my automatic tests.
> >>>
> >>> The current RHEL6 kernel has enough pv-on-hvm drivers to get the
> >>> toolstack to try to negotiate with the guest during save, but not
> >>> enough (or not working enough) drivers to actually complete the
> >>> negotiation.  So the toolstack times out and the migrate/save fails.
> >> Does any of the pv-on-hvm functionality work at all in RHEL6?
> >>
> >> Perhaps setting "platform_pci = 0" in the guest configuration would work
> >> around this issue? In particular if RHEL6 has no useful pv-on-hvm
> >> functionality in the first place then nothing is lost.
> >>
> > 
> > RHEL6 is supposed to have working PV-on-HVM drivers,
> > based on Stefano's patches.
> > 
> > Redhat has tested them with RHEL5 dom0, afaik.
> > 
> Correct. xen-emul-unplug doesn't work, though, due to late
> patch that broke that bit of it.
> A fair number of upstream patches for pv-hvm & save/restore
> have been backported to the next rhel6 update, so hoping this
> state of missing pieces in rhel6 pv-on-hvm will be short-lived.
> oh, added benefit: the patches should improve rhel6 (pv-on)hvm
> on non-RHEL5 dom0's as well.
> 

Great! Thanks for the reply!

-- Pasi