* migration
@ 2004-10-01 1:01 James Harper
2004-10-01 10:43 ` migration Ian Pratt
0 siblings, 1 reply; 5+ messages in thread
From: James Harper @ 2004-10-01 1:01 UTC (permalink / raw)
To: xen-devel
I just tested live migration, and after migrating I could ping the
domain in its new location but couldn't connect to it. Shortly after I
found that the iscsi target (another linux box) had crashed, so possibly
the migration actually worked and the iscsi failure was why I couldn't
connect.
I tried it a second time and 'xm migrate' is just sitting there.
Is live migration considered working at this point? Is there anything
special I should know about it? What are the steps to do a non-live
migration?
Thanks
James
-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: migration
2004-10-01 1:01 migration James Harper
@ 2004-10-01 10:43 ` Ian Pratt
0 siblings, 0 replies; 5+ messages in thread
From: Ian Pratt @ 2004-10-01 10:43 UTC (permalink / raw)
To: James Harper; +Cc: xen-devel, Ian.Pratt
> I just tested live migration, and after migrating I could ping the
> domain in its new location but couldn't connect to it. Shortly after I
> found that the iscsi target (another linux box) had crashed, so possibly
> the migration actually worked and the iscsi failure was why I couldn't
> connect.
>
> I tried it a second time and 'xm migrate' is just sitting there.
>
> Is live migration considered working at this point? Is there anything
> special I should know about it? What are the steps to do a non-live
> migration?
It's certainly believed to work, but it's not been widely tested.
I've done hundreds of migrations of domains running loaded
databases and web servers, and it works for me.
'xm migrate' on its own does a stop-and-copy migrate. Adding the
'-l' flag makes it a live migrate where the OS isn't stopped.
Live migration is a lot more "exciting" from an implementation
point of view than stop-and-copy, so if you do have bug reports
it would be very useful to know whether you can repeat them in
stop-and-copy mode.
NB there are a few migration fixes that are in unstable that
haven't been pushed through to 2.0 yet.
Ian
-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
^ permalink raw reply [flat|nested] 5+ messages in thread
* RE: migration
@ 2004-10-01 11:14 James Harper
2004-10-01 14:14 ` migration Ian Pratt
0 siblings, 1 reply; 5+ messages in thread
From: James Harper @ 2004-10-01 11:14 UTC (permalink / raw)
To: Ian.Pratt; +Cc: xen-devel
>
> It's certainly believed to work, but it's not been widely tested.
>
> I've done hundreds of migrations of domains running loaded
> databases and web servers, and it works for me.
Me too, I followed up to this email saying that the cause of both issues
I had was most likely because my eth0 barfed.
> 'xm migrate' on its own does a stop-and-copy migrate. Adding
> the '-l' flag makes it a live migrate where the OS isn't stopped.
Thought it might be something like that.
> Live migration is a lot more "exciting" from an
> implementation point of view than stop-and-copy, so if you do
> have bug reports it would be very useful to know whether you
> can repeat them in stop-and-copy mode.
I was logged in the whole time and there was about 1/2 a second where it
seemed to freeze, but no bugs to report.
> NB there are a few migration fixes that are in unstable that
> haven't been pushed through to 2.0 yet.
I'm running unstable built today.
James
-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: migration
2004-10-01 11:14 migration James Harper
@ 2004-10-01 14:14 ` Ian Pratt
2004-10-01 15:21 ` migration Jacob Gorm Hansen
0 siblings, 1 reply; 5+ messages in thread
From: Ian Pratt @ 2004-10-01 14:14 UTC (permalink / raw)
To: James Harper; +Cc: Ian.Pratt, xen-devel
> > Live migration is a lot more "exciting" from an
> > implementation point of view than stop-and-copy, so if you do
> > have bug reports it would be very useful to know whether you
> > can repeat them in stop-and-copy mode.
>
> I was logged in the whole time and there was about 1/2 a second where it
> seemed to freeze, but no bugs to report.
With the old xend and old IO we used to get live migration times
downtimes in the order of <50ms for quiescent domains.
I'm confident we can get back to that order of performance with a
few tweaks Mike and I are currently working on.
Ian
-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: migration
2004-10-01 14:14 ` migration Ian Pratt
@ 2004-10-01 15:21 ` Jacob Gorm Hansen
0 siblings, 0 replies; 5+ messages in thread
From: Jacob Gorm Hansen @ 2004-10-01 15:21 UTC (permalink / raw)
To: Ian Pratt; +Cc: James Harper, xen-devel
Ian Pratt wrote:
>>>Live migration is a lot more "exciting" from an
>>>implementation point of view than stop-and-copy, so if you do
>>>have bug reports it would be very useful to know whether you
>>>can repeat them in stop-and-copy mode.
>>
>>I was logged in the whole time and there was about 1/2 a second where it
>>seemed to freeze, but no bugs to report.
>
>
> With the old xend and old IO we used to get live migration times
> downtimes in the order of <50ms for quiescent domains.
>
> I'm confident we can get back to that order of performance with a
> few tweaks Mike and I are currently working on.
hi,
I have an alternative implementation, where live-migration takes place
entirely within the guest domain. It is currently only implemented for
Linux 2.4, and in the old 1.3 I/O model. It has (in my opinion) a few
benefits, because Linux is able to make better informed decisions during
migration than Xen is.
My hope is to someday be compatible with the latest and greatest
Xen-version, so I have not released any source for this yet. However, if
you are interested, let me know and I can send you my patches for
XenoLinux (no changes to Xen needed), or some binaries to play with.
There is a short paper describing my approach at
http://www.diku.dk/~jacobg/nomadxen.pdf
best,
Jacob
-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2004-10-01 15:21 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-10-01 11:14 migration James Harper
2004-10-01 14:14 ` migration Ian Pratt
2004-10-01 15:21 ` migration Jacob Gorm Hansen
-- strict thread matches above, loose matches on Subject: below --
2004-10-01 1:01 migration James Harper
2004-10-01 10:43 ` migration Ian Pratt
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.