Date: Thu, 26 Jul 2007 09:01:13 -0500 From: Craig Boston <craig@xfoil.gank.org> To: Pawel Jakub Dawidek <pjd@FreeBSD.org> Cc: Dmitry Morozovsky <marck@rinet.ru>, current@FreeBSD.org Subject: Re: ZFS on a notebook/512M settings Message-ID: <20070726140113.GA49713@nowhere> In-Reply-To: <20070726134231.GO12473@garage.freebsd.pl> References: <20070725234322.G33266@woozle.rinet.ru> <20070725222035.GA50522@nowhere> <20070726134231.GO12473@garage.freebsd.pl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Jul 26, 2007 at 03:42:31PM +0200, Pawel Jakub Dawidek wrote: > Let me explain what disabling ZIL really means. Once ZIL is disabled, > fsync(2) is a no-op, ie. calling fsync(2) on a descriptor doesn't mean > your data would be safely stored on disk at the time function returns. > There is no data corruption for local use, only this fsync(2) problem. > > "Data corruption" can happen from NFS client point of view, when your > ZFS file system is exported over NFS and your NFS server crashes. Does this increase the window after an fsync is performed that a crash or power loss will lose recent changes? That's what I got from reading the descriptions of ZIL, perhaps "corruption" was a poor choice of words. Craig
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070726140113.GA49713>