Re: [Gluster-devel] AFR documentation

[Chris Johnson]

On Fri, 19 Oct 2007, Thierry EXCOFFIER wrote:

     My $0.02.  First rule of both science and engineering, never
assume.  It almost always causes big problems.

> Krishna Srinivas wrote:
> > >    - If the version numbers are identicals, do you detect that
> > >      the file content is not the same ?
> >
> > No. But we can have improve the safety level to handle such cases
> > by further checking if the copies differ by mtime (by more than 30 secs, 
> > say)
> > or if they have different file sizes etc in which case we can return EIO.
> > Again suggestions are welcome.
>
> Can you store the IP of the last client who have modified the file ?
> If the version is the same but not the client IP then the file
> content is in an inpredictable state.
>
> May be you could assume that if the file is wrote by more than one
> client then the *user application* has locked the file before.
> In this case you need an option to indicate to AFR that
> the file locking succeed only if all the clients are up.
>
> Hope my 2 cents may help.
>
> Thierry.
> --
> Thierry.EXCOFFIER@@liris.univ-lyon1.fr//   GO: Oriental Strategy Game
> LIRIS/UFR Informatique, bat710       //DYLAN: Programming Language
> Universite Claude Bernard LYON 1  __// UNIX: Standard Operating System
> 43 Bd du 11 novembre 1918       //
> 69622 VILLEURBANNE, FRANCE     //WWW: http://www710.univ-lyon1.fr/~exco
>
>
> _______________________________________________
> Gluster-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/gluster-devel

------------------------------------------------------------------------------- 
Chris Johnson               |Internet: address@hidden
Systems Administrator       |Web:      http://www.nmr.mgh.harvard.edu/~johnson
NMR Center                  |Voice:    617.726.0949
Mass. General Hospital      |FAX:      617.726.7422
149 (2301) 13th Street      |If God has a plan for my life then I want to have
Charlestown, MA., 02129 USA |a little chat with the architect.  Me
-------------------------------------------------------------------------------