[Qemu-devel] Virtio 9p live migration patches

[Qemu-devel] Virtio 9p live migration patches

[Benoît Canet]

Hello Aneesh,

I am interested in working on 9p live migration.

I remember that you had some live migration related patches on github.

I don't find these patches anymore. Do you still have them somewhere ?
What was missing from it ?

Best regards

Benoît

Re: [Qemu-devel] Virtio 9p live migration patches

[Aneesh Kumar K.V]

Benoît Canet <benoit.canet@irqsave.net> writes:

> Hello Aneesh,
>
> I am interested in working on 9p live migration.
>
> I remember that you had some live migration related patches on github.
>
> I don't find these patches anymore. Do you still have them somewhere ?
> What was missing from it ?
>

Had to redo the tree, because of the fork changes. Now pushed as
live-migration branch 

git://github.com/kvaneesh/qemu.git live-migration

-aneesh

Re: [Qemu-devel] Virtio 9p live migration patches

[Benoît Canet]

Thanks a lot,

Do you have an idea of what is left to be done on it ?

Best regards

Benoît

> Le Wednesday 03 Apr 2013 à 00:00:53 (+0530), Aneesh Kumar K.V a écrit :
> Benoît Canet <benoit.canet@irqsave.net> writes:
> 
> > Hello Aneesh,
> >
> > I am interested in working on 9p live migration.
> >
> > I remember that you had some live migration related patches on github.
> >
> > I don't find these patches anymore. Do you still have them somewhere ?
> > What was missing from it ?
> >
> 
> Had to redo the tree, because of the fork changes. Now pushed as
> live-migration branch 
> 
> git://github.com/kvaneesh/qemu.git live-migration
> 
> -aneesh
> 

Re: [Qemu-devel] Virtio 9p live migration patches

[Aneesh Kumar K.V]

Benoît Canet <benoit.canet@irqsave.net> writes:

> Thanks a lot,
>
> Do you have an idea of what is left to be done on it ?
>

It has to rewritten with the new model with respect to migration. Also
these is problem of how much time does it take to migrate the fid
information ?. We also need to ensure new 9p request are blocked.
Another option is to add migration support to protocol and handle
these in the clients. 

-aneesh

Re: [Qemu-devel] Virtio 9p live migration patches

[Benoît Canet]

Thanks for the explanations I'll start working on the patchset.

Best regards

Benoît

> Le Wednesday 03 Apr 2013 à 12:03:13 (+0530), Aneesh Kumar K.V a écrit :
> Benoît Canet <benoit.canet@irqsave.net> writes:
> 
> > Thanks a lot,
> >
> > Do you have an idea of what is left to be done on it ?
> >
> 
> It has to rewritten with the new model with respect to migration. Also
> these is problem of how much time does it take to migrate the fid
> information ?. We also need to ensure new 9p request are blocked.
> Another option is to add migration support to protocol and handle
> these in the clients. 
> 
> -aneesh
> 
> 

Re: [Qemu-devel] Virtio 9p live migration patches

[Paolo Bonzini]

Il 04/04/2013 14:37, Benoît Canet ha scritto:
> We also need to ensure new 9p request are blocked.

Migration runs with the VM paused.  It would be simplest to flush all
the requests before migrating, that's what the block layer does.  (The
migration of requests we have in virtio-blk, scsi-disk etc. is only for
rerror/werror=stop; it is not invoked in the common case).

Paolo

Re: [Qemu-devel] Virtio 9p live migration patches

[Benoît Canet]

There is still the need to serialize the fid linked list.
I saw in the current code base that virtio-blk.c was using the old API to
serialize a linked list.

Does writing support for serializing linked list in vmstate worth it ?
Or is it better to keep the old API to serialize ?

Regards

Benoît

> Le Thursday 04 Apr 2013 à 15:42:06 (+0200), Paolo Bonzini a écrit :
> Il 04/04/2013 14:37, Benoît Canet ha scritto:
> > We also need to ensure new 9p request are blocked.
> 
> Migration runs with the VM paused.  It would be simplest to flush all
> the requests before migrating, that's what the block layer does.  (The
> migration of requests we have in virtio-blk, scsi-disk etc. is only for
> rerror/werror=stop; it is not invoked in the common case).
> 
> Paolo
>