Linux Checkpoint-Restart - v19

Oren Laadan orenl at cs.columbia.edu
Mon Mar 1 14:56:38 PST 2010



Andrew Morton wrote:
> On Mon, 22 Feb 2010 18:17:00 -0500
> Oren Laadan <orenl at cs.columbia.edu> wrote:
> 
>> Hi Andrew,
>>
>> We've put a stake in the ground for our next set of checkpoint/restart
>> patches, v19. It has some great new stuff, and we put extra effort to
>> address your concerns. We would like to have the code included in -mm
>> for wider feedback and testing.
>>
>> This one is able to checkpoint/restart screen and vnc sessions, and
>> live-migrate network servers between hosts. It also adds support for
>> x86-64 (in addition to x86-32, s390x and powerpc). It is rebased to
>> kernel 2.6.33-rc8.
>>
>> Since one of your main concerns was about what is not yet implemented
>> and how complicated or ugly it will be to support that, we've put up
>> a wiki page to address that. In it there is a simple table that lists
>> what is not implemented and the anticipated solution impact, and for
>> some entries a link to more details.
>>
>> The page is here:   http://ckpt.wiki.kernel.org/index.php/Checklist
> 
> Does "Refuses to Checkpoint" mean that an attempt to checkpoint will
> fail, return the failure to userspace and the system continues as
> before?

Yes.

Also, there is a mechanism in place to report the reason for the
failure (e.g. the offending resource) to the user.

> 
>> We want to stress that the patchset is already very useful as-is. We
>> will keep working to implement more features cleanly. Some features we
>> are working on include network namespaces and device configurations,
>> mounts and mounts namespaces, and file locks. Should a complicated
>> feature prove hard to implement, users have alternatives systems like
>> kvm, until we manage to come up with a clean solution.
>>
>> We believe that maintenance is best addressed through testing. We now
>> have a comprehensive test-suite to automatically find regressions.
>> In addition, we ran LTP and the results are the same with CHECKPOINT=n
>> and =y.
>>
>> If desired we'll send the whole patchset to lkml, but the git trees
>> can be seen at:
>>
>>    kernel:       http://www.linux-cr.org/git/?p=linux-cr.git;a=summary
>>    user tools:   http://www.linux-cr.org/git/?p=user-cr.git;a=summary
>>    tests suite:  http://www.linux-cr.org/git/?p=tests-cr.git;a=summary
>>
> 
> I'd suggest waiting until very shortly after 2.6.34-rc1 then please
> send all the patches onto the list and let's get to work.
> 

Sounds good -- will do.

Thanks,

Oren.


More information about the Containers mailing list