From owner-cvs-all Sat Oct 13 21:34:38 2001 Delivered-To: cvs-all@freebsd.org Received: from earth.backplane.com (earth-nat-cw.backplane.com [208.161.114.67]) by hub.freebsd.org (Postfix) with ESMTP id DD25B37B40D; Sat, 13 Oct 2001 21:34:31 -0700 (PDT) Received: (from dillon@localhost) by earth.backplane.com (8.11.6/8.11.2) id f9E4YTj46344; Sat, 13 Oct 2001 21:34:29 -0700 (PDT) (envelope-from dillon) Date: Sat, 13 Oct 2001 21:34:29 -0700 (PDT) From: Matt Dillon Message-Id: <200110140434.f9E4YTj46344@earth.backplane.com> To: Peter Wemm Cc: Alfred Perlstein , Mikhail Teterin , des@ofug.org, silby@silby.com, cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG Subject: Re: noatimedir? (Re: cvs commit: src/sys/vm vnode_pager.c) References: <20011013235734.D6B113810@overcee.netplex.com.au> Sender: owner-cvs-all@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG :At yahoo, we have a modification to the noatime semantics, that seems quite :a bit more useful.. : :We have a sysctl (ugh) that sets the maximum atime age of a file at access. : :For example, if we set it to 86400 (1 day) then: : : open/read a file that has atime < 1 hour ago: : no atime update : open/read a file that has atime < 23 hours ago: : no atime update : open/read a file that has atime > 24 hours ago: : atime gets updated once (and subsequent access will see the atime reset) : :This means that we can guarantee a maximum of one atime update per file. It :also means that the atime on a file is always roughly accurate to within a :day. This helps us a lot for archiving files that have not been read for :(say) 3 months. : :Ideally this should be a mount option, but that then means that ufs needs :to start taking mount *arguments* (like nfs, with all the nfsargs version :incompatabilities), not just binary flags. : :Cheers, :-Peter :-- :Peter Wemm - peter@FreeBSD.org; peter@yahoo-inc.com; peter@netplex.com.au It does seem like it ought to be a mount option, but then again a sysctl is very simple and won't require us to mess around with the mount structures. again. If you want to part with the patch, I'd like to see what it entails. I suspect it's fairly simple, right? If so I think it's worth throwing into the system and then adding a section to tuning(7) regarding it. -Matt To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message