home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.unix.sysv386
- Path: sparky!uunet!cs.utexas.edu!sun-barr!ames!elroy.jpl.nasa.gov!orchard.la.locus.com!hatch!mike
- From: mike@hatch.socal.com (Mike Bernadett)
- Subject: Re: UFS , SVR4, and inodes limits
- Message-ID: <BsxCro.Fwx@hatch.socal.com>
- Organization: Hatch Usenet and E-mail. Playa del Rey, CA
- References: <1992Aug6.014526.13249@isus.UUCP> <MJN.92Aug9042633@pseudo.uucp>
- Date: Thu, 13 Aug 1992 13:42:59 GMT
- Lines: 34
-
- >hoyt@isus.UUCP (Hoyt A. Stearns jr.) writes:
- >
- >>In UHC SVR4, an artificial limit of 64000 inodes is placed on UFS file
- >>systems (mkfs -C). I re-created the filesystem without the "-C",
- >>but the df command reports (accurately) the limit is still there.
- >>the mtune file shows an upper limit of 1300 inode blocks, which I set.
- >
- Murray writes:
- >Actually, the limit is 64K, not 64000.
- >
- >On a System V filesystem, directory entries contain only 2 bytes with
- >which to represent the inode of a file, thus the limit of 64K inodes
- >per filesystem. The actual number of available inodes is one less
- >than this, as inode 0 is reserved to indicate an empty directory
- >entry.
- >
- >This limit is the reason many System V sites must spool comp.lang.c
- >and comp.unix.questions onto separate filesystems. :-)
- >
-
- Pre-SVR4 this was the case, but with the SVR4 ufs filesystem, you should
- be able to get more than 64K inodes (according the the AT&T SVR4 ufs docs)
- if you wail on the mkfs command enough.
-
- Try decreasing the number of cylinders per cylinder group, thereby increasing
- the number of cgs. You may be limited by an inodes-per-cg calculation in mkfs.
-
- decreasing nbpi (block-to-inode ratio) on my AT&T SVR4 gives me a warning that
- I'm going above 64K, but doesn't do anything -- I still get ~20,000 inodes
- on a 125MB filesystem until I increase the number of cylinder groups.
-
- --
- Mike Bernadett mike@hatch.socal.com
- 8635 Falmouth Ave. #105, Playa del Rey, CA 90293 (310) 305-8758
-