Date: 27 Sep 1998 20:22:47 -0400 From: Kevin Street <street@iname.com> To: Peter Wemm <peter@netplex.com.au> Cc: Terry Lambert <tlambert@primenet.com>, bde@zeta.org.au (Bruce Evans), Don.Lewis@tsc.tdk.com, freebsd-current@FreeBSD.ORG, roberto@keltia.freenix.fr Subject: Re: Softupdates panics Message-ID: <8790j584rs.fsf@kstreet.interlog.com> In-Reply-To: Peter Wemm's message of "Mon, 28 Sep 1998 05:46:36 %2B0800" References: <199809272146.FAA14600@spinner.netplex.com.au>
next in thread | previous in thread | raw e-mail | index | archive | help
Peter Wemm <peter@netplex.com.au> writes: > Terry Lambert wrote: > > > >} > added `noatime' to two of my heavily used softupdates slices in > > > >} > > > >} noatime with SU is a known Bad Thing[tm]. > > > > > > > >I hadn't heard this. I don't know why that would be true. > > > > > > It is superstition. Perhaps stamping atimes hides some bugs in > > > softupdates by causing more frequent updates or changing the timing > > > of the updates. > > > > Doing "noatime" results in the loss of an inode write order > > dependency. > > So? what's being written while you're reading a file? > > If you're reading and writing the file together, then the modtime stamps > will cause the inode to be written out. If you're just reading, there is > no metadata dependency at all because nothing is changed at all. > > Had anybody got problems with softupdates that can be specifically > attributed to the use of -noatime ? The dumps I posted earlier in this thread were clearly triggered by noatime. With noatime in /etc/fstab I got 2 dumps in a few minutes doing big port compiles. Without noatime, I'm back to stability. This doesn't prove that noatime *causes* the problem, but it certainly triggers it in my case. -- Kevin Street street@iName.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?8790j584rs.fsf>