Date: Mon, 22 Jan 2007 07:24:41 -0600 From: Eric Anderson <anderson@centtech.com> To: Ivan Voras <ivoras@fer.hr> Cc: freebsd-geom@freebsd.org Subject: Re: A beginner project Message-ID: <45B4BB19.4050604@centtech.com> In-Reply-To: <ep2d8f$i0m$1@sea.gmane.org> References: <20070120073117.GB60202@x12.dk> <45B456B6.6040808@centtech.com> <20070122084852.GA18648@x12.dk> <45B4B406.7060906@centtech.com> <ep2d8f$i0m$1@sea.gmane.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 01/22/07 07:12, Ivan Voras wrote: > Eric Anderson wrote: > >> Not snapshot like a photo, snapshot like 'point in time copy of storage'. >> > > I think the "current" problem with geom-based snapshots is how will they > interact with gjournal - i.e. do you snapshot the journal or not? I think it only matters if you use UFS, if you care about a consistent filesystem on the snapshot. You could probably tie a hook into geom snapshot to tell gjournal to flush the journal, take the snapshot, and then continue. A similar hook (like gjournal has) to flush the buffer down to the disk would be good to have in geom snapshot also. I actually think the hard part is keeping track of the snapshotted blocks, and keeping them in an order that doesn't completely destroy performance, while still allowing an 'instant' snapshot. UFS snapshots take a very very long time on large file systems, so a geom snapshot (to be useful) should be snappy (pun intended). Eric -- ------------------------------------------------------------------------ Eric Anderson Sr. Systems Administrator Centaur Technology An undefined problem has an infinite number of solutions. ------------------------------------------------------------------------
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?45B4BB19.4050604>