Date: Thu, 07 Dec 2006 14:59:30 +0100 From: Ivan Voras <ivoras@fer.hr> To: Eric Anderson <anderson@centtech.com> Cc: freebsd-geom@freebsd.org Subject: Re: gsnapshot around ? Message-ID: <45781E42.8080905@fer.hr> In-Reply-To: <45781663.3060008@centtech.com> References: <20061201104955.GG9880@obiwan.tataz.chchile.org> <el1395$9lt$1@s ea.gmane.org> <4575E597.1030306@centtech.com> <4575EDEA.7020206@fer.hr> <45781663.3060008@centtech.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Eric Anderson wrote: > I've been thinking about a layered approach, so the gsnapshot would > create a 'layer', and all new writes would go to the layer. Then, there > would be another class, to stack layers (kind of like unionfs, except > for block devices) - so when you wanted to view the data on a snapshot, > you would stack the layers how you wanted, and then see the data. Kind > of like journaling, kind of like regular COW, etc, but a little more > Unix-style-ish in the manner of a few building blocks that can be used > in many ways, and stacked together to make something powerful. So instead of messing with GEOM, you'd create a class that does the stacking internally? I don'e know how pjd and phk are going to react to such hackery but I think it has potential. Then the user would only have to specify the "hotplug" class once and leave it in production (without noticable speed penalty) until he needs it, then various things could be stacked into it (like logging, tracing, even snaphosts) with minimal effort.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?45781E42.8080905>