Date: Fri, 26 May 2006 14:44:26 -0500 From: Eric Anderson <anderson@centtech.com> To: Kris Kennaway <kris@obsecurity.org> Cc: freebsd-fs@freebsd.org Subject: Re: Access to UFS stuck during snapshot. Message-ID: <44775A9A.5080907@centtech.com> In-Reply-To: <20060525215314.GA31540@xor.obsecurity.org> References: <4473FA8F.6050102@outblaze.com> <20060524061848.GA58522@xor.obsecurity.org> <44744EF8.2000000@centtech.com> <20060525215314.GA31540@xor.obsecurity.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Kris Kennaway wrote: > On Wed, May 24, 2006 at 07:18:00AM -0500, Eric Anderson wrote: >> Kris Kennaway wrote: >>> On Wed, May 24, 2006 at 02:17:51PM +0800, Kent Ho wrote: >>>> Hi, >>>> >>>> I having problems accessing file system during snapshot runs. snapshot >>>> takes approximately 10 to 20 mins to run. >>>> >>>> During that time access to files in the file system are blocked until >>>> the snapshot has completed. Lot of processes was stuck and piles up on >>>> the fs. >>>> >>>> This fs stores mailboxes and it's accessed 24/7. This is similar to >>>> what described in this thread. >>>> >>>> http://www.freebsd.org/cgi/getmsg.cgi?fetch=14968+0+/usr/local/www/db/text/2006/freebsd-fs/20060312.freebsd-fs >>>> >>>> Has this been improved or fixed in recent releases, I see a bunch of >>>> fixes in march? any possible work around? any alternatives or >>>> recommendations? Looking for high availability fs with snapshots. >>> It's by design. >>> >>> Kris >> I found that making the .snap directories permissions 0700, and then >> making a subdirectory under the .snap that stores the actual snapshots >> helped, as long as I did not make any stat() calls to the snapshot file >> that was in snaplk state. >> >> Kris - you've had your head deep in the snapshot spaghetti recently - is >> this blocking just a matter of the locking on the snapshot file and a >> stat() call blocking on it, which causes an upward locking to the root >> dir of the filesystem? Could this be avoided by 'hiding' the snapshot >> file while a snapshot is in progress? I have been tinkering with hiding >> the snapshot directory during snapshot creation, but I don't have any >> results of anything yet, still too early. I'm a wannabe hacker, so my >> progress is slow. > > No, I/O is suspended while the snapshot is being created. As I said, > it's a design limitation. > > Kris The suspension is only during the second read of the cylinder group changes and such, right? I was under the impression that the suspension was actually quite quick, it's the initial copying of the cylinder groups that takes the most time, which the fs is not suspended until after that initial copy. My tests just now on my laptop seem to indicate as such (running 6-STABLE as of a few days ago). Is -current different now? Eric -- ------------------------------------------------------------------------ Eric Anderson Sr. Systems Administrator Centaur Technology Anything that works is better than anything that doesn't. ------------------------------------------------------------------------
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?44775A9A.5080907>