Re: [SCM] Samba Shared Repository - branch v3-2-test updated -release-3-2-0pre2-2478-g79bda44#

Jeremy,

On Mon, May 26, 2008 at 11:18:27PM -0500, Jeremy Allison wrote:[color=blue]
> The branch, v3-2-test has been updated
> via 79bda4467f399325773211bb4b660363024c66b4 (commit)
> from 26f46d3037d7920df0768e7a10e8a92ffb4ce0cd (commit)
>
> [url]http://gitweb.samba.org/?p=samba.git;a=shortlog;h=v3-2-test[/url]
>
>
> - Log -----------------------------------------------------------------
> commit 79bda4467f399325773211bb4b660363024c66b4
> Author: Jeremy Allison <jra@samba.org>
> Date: Mon May 26 21:17:28 2008 -0700
>
> Re-enable the evil "aio write behind" parameter.
> Jeremy.
>
> -----------------------------------------------------------------------[/color]

Would you please explain why this parameter is evil an why it should be
re-enabled between rc1 an the final release (or maybe rc2)?

Thanks!

Karolin

--
SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen
phone: +49-551-370000-0, fax: +49-551-370000-9
AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
[url]http://www.SerNet.DE[/url], mailto: Info @ SerNet.DE


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4-svn0 (GNU/Linux)

iD8DBQFIO6byKGi9fisXk1ERAr9LAJwLpEwJaxWA7dJP8ZDwaupXM5E2zwCgg+uP
oqVRHy+45fqLw2cZ8lPx+rk=
=56bG
-----END PGP SIGNATURE-----

On Tue, May 27, 2008 at 08:15:15AM +0200, Karolin Seeger wrote:[color=blue]
> Would you please explain why this parameter is evil an why it should be
> re-enabled between rc1 an the final release (or maybe rc2)?[/color]

Oh, the evilness is easy to explain :-)

With this parameter we acknowledge a write request by the
client before we even know if we have enough space on disk.
This makes writes very fast, because we can defer the actual
write to disk, but it prevents any error reporting like disk
full etc.

Volker

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)

iD8DBQFIO6tQUzqjrWwMRl0RArghAJ9lGE3aGcRmUyrPrZgB7thQTtrA0ACfRidV
Ak3ITdu3sDm/68/SvDdgOQw=
=lTFH
-----END PGP SIGNATURE-----

On Tue, May 27, 2008 at 08:15:15AM +0200, Karolin Seeger wrote:[color=blue]
>
> Would you please explain why this parameter is evil an why it should be
> re-enabled between rc1 an the final release (or maybe rc2)?[/color]

Volker explained why it is evil. :-). This one is definately
your call. It's low-to-no risk (which is why I added the
patch), but definately not required. There are people
out there (remaining nameless :-) who are investigating
Samba performance for large-scale systems, who need to
be able to turn this parameter on and off for comparison.
The rest of the supporting code for it is already in the
tree, and this was the one missing piece to complete it.

Had none of the other code been present I would definately
not have added it, but as most of it was already there it
seemed churlish not to complete it by allowing a parameter
to be read :-).

Jeremy.