Re: [Qemu-devel] [Qemu-block] [PATCH 15/18] block/mirror: Add active mirroring

[Stefan Hajnoczi <stefanha@redhat.com>]

On Tue, Sep 19, 2017 at 10:57:50AM +0100, Daniel P. Berrange wrote:
> On Tue, Sep 19, 2017 at 10:44:16AM +0100, Stefan Hajnoczi wrote:
> > On Mon, Sep 18, 2017 at 06:26:51PM +0200, Max Reitz wrote:
> > > On 2017-09-18 12:06, Stefan Hajnoczi wrote:
> > > > On Sat, Sep 16, 2017 at 03:58:01PM +0200, Max Reitz wrote:
> > > >> On 2017-09-14 17:57, Stefan Hajnoczi wrote:
> > > >>> On Wed, Sep 13, 2017 at 08:19:07PM +0200, Max Reitz wrote:
> > > >>>> This patch implements active synchronous mirroring.  In active mode, the
> > > >>>> passive mechanism will still be in place and is used to copy all
> > > >>>> initially dirty clusters off the source disk; but every write request
> > > >>>> will write data both to the source and the target disk, so the source
> > > >>>> cannot be dirtied faster than data is mirrored to the target.  Also,
> > > >>>> once the block job has converged (BLOCK_JOB_READY sent), source and
> > > >>>> target are guaranteed to stay in sync (unless an error occurs).
> > > >>>>
> > > >>>> Optionally, dirty data can be copied to the target disk on read
> > > >>>> operations, too.
> > > >>>>
> > > >>>> Active mode is completely optional and currently disabled at runtime.  A
> > > >>>> later patch will add a way for users to enable it.
> > > >>>>
> > > >>>> Signed-off-by: Max Reitz <mreitz@redhat.com>
> > > >>>> ---
> > > >>>>  qapi/block-core.json |  23 +++++++
> > > >>>>  block/mirror.c       | 187 +++++++++++++++++++++++++++++++++++++++++++++++++--
> > > >>>>  2 files changed, 205 insertions(+), 5 deletions(-)
> > > >>>>
> > > >>>> diff --git a/qapi/block-core.json b/qapi/block-core.json
> > > >>>> index bb11815608..e072cfa67c 100644
> > > >>>> --- a/qapi/block-core.json
> > > >>>> +++ b/qapi/block-core.json
> > > >>>> @@ -938,6 +938,29 @@
> > > >>>>    'data': ['top', 'full', 'none', 'incremental'] }
> > > >>>>  
> > > >>>>  ##
> > > >>>> +# @MirrorCopyMode:
> > > >>>> +#
> > > >>>> +# An enumeration whose values tell the mirror block job when to
> > > >>>> +# trigger writes to the target.
> > > >>>> +#
> > > >>>> +# @passive: copy data in background only.
> > > >>>> +#
> > > >>>> +# @active-write: when data is written to the source, write it
> > > >>>> +#                (synchronously) to the target as well.  In addition,
> > > >>>> +#                data is copied in background just like in @passive
> > > >>>> +#                mode.
> > > >>>> +#
> > > >>>> +# @active-read-write: write data to the target (synchronously) both
> > > >>>> +#                     when it is read from and written to the source.
> > > >>>> +#                     In addition, data is copied in background just
> > > >>>> +#                     like in @passive mode.
> > > >>>
> > > >>> I'm not sure the terms "active"/"passive" are helpful.  "Active commit"
> > > >>> means committing the top-most BDS while the guest is accessing it.  The
> > > >>> "passive" mirror block still works on the top-most BDS while the guest
> > > >>> is accessing it.
> > > >>>
> > > >>> Calling it "asynchronous" and "synchronous" is clearer to me.  It's also
> > > >>> the terminology used in disk replication (e.g. DRBD).
> > > >>
> > > >> I'd be OK with that, too, but I think I remember that in the past at
> > > >> least Kevin made a clear distinction between active/passive and
> > > >> sync/async when it comes to mirroring.
> > > >>
> > > >>> Ideally the user wouldn't have to worry about async vs sync because QEMU
> > > >>> would switch modes as appropriate in order to converge.  That way
> > > >>> libvirt also doesn't have to worry about this.
> > > >>
> > > >> So here you mean async/sync in the way I meant it, i.e., whether the
> > > >> mirror operations themselves are async/sync?
> > > > 
> > > > The meaning I had in mind is:
> > > > 
> > > > Sync mirroring means a guest write waits until the target write
> > > > completes.
> > > 
> > > I.e. active-sync, ...
> > > 
> > > > Async mirroring means guest writes completes independently of target
> > > > writes.
> > > 
> > > ... i.e. passive or active-async in the future.
> > > 
> > > So you really want qemu to decide whether to use active or passive mode
> > > depending on what's enough to let the block job converge and not
> > > introduce any switch for the user?
> > > 
> > > I'm not sure whether I like this too much, mostly because "libvirt
> > > doesn't have to worry" doesn't feel quite right to me.  If we don't make
> > > libvirt worry about this, then qemu has to worry.  I'm not sure whether
> > > that's any better.
> > > 
> > > I think this really does get into policy territory.  Just switching to
> > > active mode the instant target writes are slower than source writes may
> > > not be what the user wants: Maybe it's OK for a short duration because
> > > they don't care about hard convergence too much.  Maybe they want to
> > > switch to active mode already when "only" twice as much is written to
> > > the target as to the source.
> > > 
> > > I think this is a decision the management layer (or the user) has to make.
> > 
> > Eric: Does libvirt want to be involved with converging the mirror job
> > (i.e. if the guest is writing to disk faster than QEMU can copy data to
> > the target)?
> 
> Libvirt doesn't really want to set the policy - it will just need to expose
> the mechansim & information to make such decisions upto the application.
> 
> I agree with Max that we don't want QEMU making this policy decision on
> its own. Simply switching from passive to active mode is not an approach
> that will be satisfactory to all scenarios. It might be preferrable to
> apply CPU throttling of the guest, or I/O rate throttling, or temporarily
> pause the guest to allow catch or any number of other policies. Neither
> libvirt or QEMU knows which is best for the scenario at hand.

Okay.

Stefan