Date: Thu, 27 Apr 2006 09:23:41 -0600 From: Scott Long <scottl@samsco.org> To: Eric Anderson <anderson@centtech.com> Cc: freebsd-fs@freebsd.org, Robert <robert@mhi-tx.com> Subject: Re: suspend writes for external snapshot Message-ID: <4450E1FD.7040008@samsco.org> In-Reply-To: <4450D278.5000802@centtech.com> References: <444FE114.7010106@mhi-tx.com> <20060426211059.GA85780@xor.obsecurity.org> <4450B3B1.8070704@centtech.com> <4450D18D.9060101@mhi-tx.com> <4450D278.5000802@centtech.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Eric Anderson wrote: > Robert wrote: > >> Eric Anderson wrote: >> >>> Kris Kennaway wrote: >>> >>>> On Wed, Apr 26, 2006 at 04:07:32PM -0500, Robert wrote: >>>> >>>>> Greetings, >>>>> I am trying to determine if there is a way to suspend write >>>>> activity on a FreeBSD filesystem (currently 6.0R) in order to take >>>>> a snapshot with external SAN software (Falconstor IPStor)? >>>> >>>> >>>> >>>> Use a FreeBSD snapshot and then image that? >>> >>> >>> >>> He can't do it that way because the Falconstor is a raw block device >>> and has no idea the difference between a snapshot and other >>> filesystem bits. >>> >>> >>> Robert - we have some very alpha patches in house that allow >>> suspension of UFS filesystems for things like this (also working on >>> online UFS growing). I'm not sure how close we really are to letting >>> it loose in the wild, but we will definitely be making all the >>> patches available to whomever wants it. >>> >>> >>> Eric >>> >>> >>> >> >> Sounds great thank you, I will keep an eye out for the patches. >> >> I am still unclear as to the necessity of suspending writes based on >> the following information from http://www.mckusick.com/softdep/ >> >> "By ensuring that the only inconsistencies are unclaimed blocks or >> inodes, soft updates can eliminate the need to run a filesystem check >> program after every system crash. Instead, the system is brought up >> immediately. >> >> When it is convenient, a snapshot is taken and a background task can >> be run on on that snapshot to reclaim any lost blocks and inodes. The >> use of a snapshot allows normal filesystem activity to continue >> concurrently. " >> >> Does this mean that I can take a snapshot of an in use filesystem and >> the only ill effects will be the lost blocks and inodes which can be >> reclaimed with background fsck once the snapshot is booted? > > > I believe so, yes. Someone correct me here if I am wrong. > > Ideally, yes. But for the purposes of snapshots, you want the filesystem to be flushed. Otherwise you might catch an in-progress operation like a file add/delete/truncate that would be better off finished before taking the snapshot Scott
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4450E1FD.7040008>