Is there any more docs on this other then the blurb on beta.redhat.com? I got a spare non-critical victim err... test machine I want to try this on and have all the RPMS, but would really like to chew on a little more detail before I leap. Regards, Henri -- "People die." --The Cynic's Book of Wisdom
Hi, On Mon, Nov 20, 2000 at 08:45:15AM -0600, henris@bga.com wrote:> Is there any more docs on this other then the blurb on beta.redhat.com? > I got a spare non-critical victim err... test machine I want to try this > on and have all the RPMS, but would really like to chew on a little > more detail before I leap.There is also an older readme in the ext3 tarball on ftp.*.kernel.org:/pub/linux/kernel/people/sct/ext3/*. What other information are you after? --Stephen
Stephen, sct@redhat.com said:> Yep, the release notes should cover all of that. Use "mke3fs -j" to > set journal size, and the default is 15MB.Care to comment on factors affecting the best size of journal? Are there any gotchas about making one too big? Presumably making one too small is a performance limiter. Nigel. -- [ - Opinions expressed are personal and may not be shared by VData - ] [ Nigel Metheringham Nigel.Metheringham@VData.co.uk ] [ Phone: +44 1423 850000 Fax +44 1423 858866 ]
Hi, On Tue, Nov 21, 2000 at 09:46:05AM +0000, Nigel Metheringham wrote:> > sct@redhat.com said: > > Yep, the release notes should cover all of that. Use "mke3fs -j" to > > set journal size, and the default is 15MB. > > Care to comment on factors affecting the best size of journal?If it is too small, there are two impacts. First, the writeback caching is more limited, so bursty write traffic becomes synchronous sooner. Secondly, you limit the amount of reording that can be done and the ability of the filesystem to merge writebacks together, increasing seeks. The penalties of a large journal are that the journal can occupy more pinned physical memory, and recovery takes longer. The memory consumption is probably the most significant won for most users. Cheers, Stephen