[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Libguestfs] [PATCH 2/6] v2v: rhv-upload-plugin: Use API_VERSION 2



On Sat, Jan 23, 2021 at 5:01 PM Richard W.M. Jones <rjones redhat com> wrote:
>
> On Sat, Jan 23, 2021 at 04:57:20PM +0200, Nir Soffer wrote:
> > On Sat, Jan 23, 2021 at 2:22 PM Richard W.M. Jones <rjones redhat com> wrote:
> > >
> > > On Sat, Jan 23, 2021 at 02:02:16PM +0200, Nir Soffer wrote:
> > > >
> > > >
> > > > On Sat, Jan 23, 2021, 08:28 Richard W.M. Jones <rjones redhat com> wrote:
> > > >
> > > >     On Sat, Jan 23, 2021 at 12:45:20AM +0200, Nir Soffer wrote:
> > > >     >  @failing
> > > >     > -def pwrite(h, buf, offset):
> > > >     > +def pwrite(h, buf, offset, flags):
> > > >     >      http = h['http']
> > > >     >
> > > >     >      count = len(buf)
> > > >     > @@ -234,7 +253,7 @@ def pwrite(h, buf, offset):
> > > >     >
> > > >     >
> > > >     >  @failing
> > > >     > -def zero(h, count, offset, may_trim):
> > > >     > +def zero(h, count, offset, flags):
> > > >
> > > >     Assume there's no way to pass the possible FUA flag through to oVirt here?
> > > >
> > > >
> > > > We have flush=true, which send nbd flush command after the write/zero.
> > > >
> > > > Maybe we could handle flush=true as fua when using nbd backend, I'm not sure
> > > > how important is this.
> > > >
> > > > I think we miss converting fua to flush=true to have full api 2 support.
> > >
> > > Any copy program ought not to set this flag on general writes/zeroes,
> > > so it's not really important for pwrite/zero, but a nice to have.
> >
> > Also we don't report can_fua, so the client is not allowed to send the fua
> > flag. Ignoring the fua flag is correct.
>
> Hmm, very true.

If I understand NBD_CMD_FLAG_FUA correctly, the way we allow to send
flush=y in write and zero requests is equivalent, so in a future virt-v2v
version we can convert:

    write(..., flags=NBD_CMD_FLAG_FUA)

To:

    PUT /path?flush=y

And same for zero(), using flush=true in the json payload.

So we can add can_fua now with the current of imageio, using
h["can_flush"] value.

On the server side we probably can optimize flush=y to
NBD_CMD_FLAG_FUA, I filed a bug for this:
https://bugzilla.redhat.com/1919548

> > > However nbdcopy has a --flush flag which sends a flush command to the
> > > destination when the copy has finished.  We ought to obey this one.\
> >
> > Sure, but this is handled by flush().
>
> Agreed.
>
> Rich.
>
> --
> Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
> Read my programming and virtualization blog: http://rwmj.wordpress.com
> virt-top is 'top' for virtual machines.  Tiny program with many
> powerful monitoring features, net stats, disk stats, logging, etc.
> http://people.redhat.com/~rjones/virt-top
>


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]