On Tue, Jan 05, 2010 at 06:18:53AM -0600, Marco Peereboom wrote: > You should make that mem patch available to reduce fsck surprises.
yes, but I have to tweak it first. Currently it does not take into account the (expected) number of dirs. It just takes ndirs, which is pretty low for a new fs ;-) -Otto > > On Jan 5, 2010, at 0:14, Otto Moerbeek <o...@drijf.net> wrote: > > >On Mon, Jan 04, 2010 at 10:28:28PM -0500, nixlists wrote: > > > >>Hi. > >> > >>What are the recommended newfs tweaks for an FS that will store > >>mostly > >>large or very large files? Are defaults sufficient for optimum > >>performance, or are they mostly a general case for typical OS small > >>program/text files? Also my guess tweaking with tunefs is useless, > >>since it's a very old tool? I tried tunefs with larger values than > >>default, but that makes the kernel either freeze or panic :D > >> > >>Thanks. > > > >It will work with defaults, but you can use -f and -b to increase > >fragment and block sizes to a max of 65536. That will save space on > >metadata and make fsck_ffs faster and use less memory. > > > > -Otto