Return to BSD News archive
Xref: sserve comp.unix.admin:14373 comp.unix.bsd:12780 comp.unix.internals:6366 comp.unix.wizards:31365 Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!agate!dog.ee.lbl.gov!horse.ee.lbl.gov!torek From: torek@horse.ee.lbl.gov (Chris Torek) Newsgroups: comp.unix.admin,comp.unix.bsd,comp.unix.internals,comp.unix.wizards Subject: Re: why is altering MINFREE discouraged ? Date: 21 Oct 1993 18:14:16 GMT Organization: Lawrence Berkeley Laboratory, Berkeley CA Lines: 31 Message-ID: <34896@dog.ee.lbl.gov> References: <1993Oct20.143941.7910@cs.cornell.edu> <34862@dog.ee.lbl.gov> <2a6e3t$17e@newserv.ksu.ksu.edu> NNTP-Posting-Host: 128.3.112.15 In article <2a6e3t$17e@newserv.ksu.ksu.edu> rjq@phys.ksu.edu (Rob Quinn) writes: > So are those of us with 4.3BSD type OS's (like SunOS4.1.2[right?]) safe to >start using 5%? SunOS 4.anything is not really all that much `4.3BSD-like' (though it is certainly more so than SunOS 5.x), but yes, in general. File systems that have a preponderance of small files (news spools?) may need more care than average, but even there 5% ``ought to' work. >> -i 6144 for `average' FSes >> -i 8192 for FSes with lots of large files > >Yes, I've been doing this. Any reason to use powers of 2? Aesthetics. :-) > When I use 'df' and 'df -i' to look for a good balance, it seems like an >average filesystem should have an even greater value than 8192, but I've kept >things low ``just in case.'' (just in case I don't know what I'm doing) It is a major disaster to run out of inodes before the disk is full, because once you are out of inodes, you can no longer create new files (without, at least, first removing old ones). This means it is wise to leave some room for error. Also, each individual inode is only 128 bytes, so it takes 64 inodes to use up one 8K block, and 1 megabyte = 8192 inodes. A few extra is not a large space penalty. -- In-Real-Life: Chris Torek, Lawrence Berkeley Lab CSE/EE (+1 510 486 5427) Berkeley, CA Domain: torek@ee.lbl.gov