volume management

[Pawel Jakub Dawidek]

On Tue, Apr 10, 2007 at 01:03:18PM +0200, Ivan Voras wrote:
> Gergely CZUCZY wrote:
> 
> >>I really think gvirstor is a good fit for you.  Search this list for some info on it, or just play with it a bit.  The author is active on this list and so he'll probably 
> >>pipe up.
> 
> Yes, as soon as I decide what to do when physical space runs out, it will be done. (the current problem is: if I return ENOSPC or EIO to UFS, UFS will panic or get 
> hopelessly confused).

Ivan, as you probably already find out this is not an easy task and it
gets more complex when SU comes into play or any async operations,
because there is noone waiting for the error to return.

Maybe you for now allow to set two modes of handling ENOSPC
(configurable by the user):
1. Panic if there is no physical storage. This way you protect
   consistency. You already printed a warning that gvirstor is running
   out of physical storage, so administrator has a chance to do the job.
2. Hang until the storage is available. Just don't return from I/O until
   new provider is attached to gvirstor.
You may want to leave 3rd option to just return ENOSPC, because besides
UFS there can be other gvirstor consumers that handle errors more
properly.

You may also want to consider sending warnings to devd, which I do in
ZFS (see zfs_fm.c and devd.conf) and should do in others GEOM classes.
This way administrator can configure sending an e-mail when gvristor is
running out of storage.

-- 
Pawel Jakub Dawidek                       http://www.wheel.pl
pjd at FreeBSD.org                           http://www.FreeBSD.org
FreeBSD committer                         Am I Evil? Yes, I Am!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-geom/attachments/20070410/40a676f5/attachment.pgp