Initial filesystem design synopsis.

Skip Ford at
Thu Feb 22 05:07:39 PST 2007

Victor Balada Diaz wrote:
> On Wed, Feb 21, 2007 at 02:22:16PM -0800, Matthew Dillon wrote:
> >     Here is my initial outline of the filesystem design.  It is open
> >     for discussion.  Please feel to ask questions for anything you do
> >     not understand.  I do not intend to start coding anything for at
> >     least two weeks.
> > 
> >     There are currently two rough spots in the design.  First, how to
> >     handle segment overflows in a multi-master environment.  Such overflows
> >     can occur when the individual masters or slaves have different historical
> >     data retention policies.  Second, where to store the regeneratable
> >     indexes.
> Maybe i missed it but you don't talk at all about security of the FS
> and i guess that this is a very important topic for a FS that
> it's going to work over the internet and be connected to insecure
> hosts.
> Eg: What if i want to share a file with you, but i don't
> want anyone else on the cluster to be able to read or modify it?

I was wondering as well if there would be NFS functionality
included in some way, which is more or less what you're asking I

An example could be an NFS-exported /var/mail.  It'd be nice if
rather than maintaining a synchronized clustered fs with
everyone's mail in /var/mail/ on every master, then exporting
specific directories to specific addresses via NFS, we could
allow specific addresses to become a part of the fs for specific
directories only.

Short of that, I suppose a FS per consumer would achieve the
same thing if circumstances allow.


More information about the Kernel mailing list