Re: [Gluster-devel] Questions about BDB storage translator

[Anand Avati]

> Yeah, I mentioned I thought it would be ideal for this situation.  Can I
> put my request in here to have the switch translator operate on most (if not
> all) the information that would be available in a stat(2) call?  I
> understand it might not be as simple as actually calling stat as the file
> might not exist yet, but some really useful storage location determinations
> can be made from that information.  E.g. size, block size, UID, GID, mode.


This definitely sounds 'nice'. I cannot say off the hand of how
easy/difficult/reliable it would be to depend on all stat(2) members, but I
too would surely like to schedule based on all of them. It would also be
nice if files could be rescheduled dynamically based on the the stat members
(move file when permission changes, when size changes). This would make
GlusterFS really powerful (in terms of acheiving your dream configuration).
We will consider your suggestion when we work on the switch feature.

avati