Re: [Nbd] Some errors in the FOSDEM nbd talk
- To: "Richard W.M. Jones" <rjones@...696...>
- Cc: nbd-general@lists.sourceforge.net
- Subject: Re: [Nbd] Some errors in the FOSDEM nbd talk
- From: Wouter Verhelst <w@...112...>
- Date: Mon, 6 Feb 2017 16:33:32 +0100
- Message-id: <20170206153332.ermzgi2sdgbhhf2j@...3...>
- In-reply-to: <20170206120642.GD3597@...696...>
- References: <20170206112204.GB3597@...696...> <20170206120642.GD3597@...696...>
On Mon, Feb 06, 2017 at 12:06:42PM +0000, Richard W.M. Jones wrote:
> On Mon, Feb 06, 2017 at 11:22:04AM +0000, Richard W.M. Jones wrote:
> > You write "Multiple client connections for a single export", and not
> > supported by qemu or nbdkit. Assuming that means the common case
> > where multiple clients connect to a server and access the same export,
> > and not something else, nbdkit certainly supports that, and has done
> > since the beginning. qemu-nbd supports it too, but you have to use
> > the -t option.
>
> Now I see from the talk that this means the NBD client supports
> multiple connections,
Yeah, sortof. I meant to clarify that but I forgot, sorry.
> so not applicable to nbdkit / qemu-nbd.
Actually it is.
You'll need to add the NBD_FLAG_CAN_MULTI_CONN transmission flag if it's
safe to connect multiple times from the same client (i.e., cache
coherency in the face of NBD_CMD_FLUSH and/or NBD_CMD_FLAG_FUA can be
guaranteed across connections), as explained in the spec for that flag.
I'm guessing that in the specific case of nbdkit this should be a
decision you leave to the plugin, as it might depend on how things are
written. In the case of nbd-server, I only set that flag if the
copy-on-write feature is not enabled, since when it is enabled,
nbd-server writes changes to a per-connection "difffile", so then these
changes will not (ever) be shared across connections and the client
would get very confused.
--
< ron> I mean, the main *practical* problem with C++, is there's like a dozen
people in the world who think they really understand all of its rules,
and pretty much all of them are just lying to themselves too.
-- #debian-devel, OFTC, 2016-02-12
Reply to: