Date: Sun, 1 May 2011 11:32:33 -0500 (CDT) From: Bob Friesenhahn <bfriesen@simple.dallas.tx.us> To: Rick Macklem <rmacklem@uoguelph.ca> Cc: freebsd-fs@freebsd.org Subject: Re: make the experimental NFS subsystem the default one Message-ID: <alpine.GSO.2.01.1105011124260.20825@freddy.simplesystems.org> In-Reply-To: <640208384.682241.1303948694525.JavaMail.root@erie.cs.uoguelph.ca> References: <640208384.682241.1303948694525.JavaMail.root@erie.cs.uoguelph.ca>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 27 Apr 2011, Rick Macklem wrote: > > I don't know anything about ZFS, but I would think that, if you see a > major performance improvement, that ZFS isn't committing stuff to logs > so that data won't be lost. > > Maybe the ZFS folks can comment? (I don't remember seeing the details > of what you change? If you sent a patch, sorry, but I've misplaced it.) Zfs will loose as much as 5 seconds worth of data (and maybe even 10 seconds) if the data is written slowly and/or the server has quite a lot of RAM. It commits data in order so the written data will be completely coherent for that snapshot in time, but the result may still be completely corrupted from the client's perspective. 5 (or 10!) seconds of data could be quite a lot of data, and could represent entire new directory trees, or large directory trees which were removed. Individual file content could be overwritten hundreds of times before the point where the server arbitrarily decides to commit it. If the server bounces, its data won't match what the client thinks it should have. Bob -- Bob Friesenhahn bfriesen@simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?alpine.GSO.2.01.1105011124260.20825>