chris@trantor.umd.edu (Chris Torek) (02/28/88)
In article <1013@cadre.dsl.PITTSBURGH.EDU> sean@cadre.dsl.PITTSBURGH.EDU (Sean McLinden) writes: >A few months ago, someone posted a note stating that the default >number of inodes created by mkfs was inadequate for large filesystems. The problem is not large file systems *per se*; rather, if there are many blocks per cylinder group, mkfs will generate only 2048 inodes per c.g., and you will wind up with fewer than the default 1-inode-per-2048 bytes. In practise, I find that `user' file systems (home directories) here at Maryland CSD work well with about 6144 bytes per inode, and `system' file systems (/usr) work well with about 7168 bytes per inode. The news file system at this moment has about 5000 bytes per inode, but has had fewer than that in the past; -i 2048 is not unreasonable for it. >Does anyone recall what was suggested for large (~350 Mbyte), or greater >filesystems? After running newfs, use `df -i' to see how many inodes were allocated; divide kbytes*1024 by ifree to see how many bytes per inode you got. If the number is too big, re-run newfs with an additional `-c <n>', where <n> is less than 16. With fewer cylinders per group, there will be fewer blocks to divide amongst those maximum-of-2048 inodes. (And who ever said 350 MB was a big file system? :-) ...: % gyre df Filesystem kbytes used avail capacity Mounted on /dev/ra0a 38959 17038 18025 49% / /dev/ra0d 585915 400419 126904 76% /usr /dev/ra1d 583635 476223 49048 91% /x /dev/ra2a 38751 1272 33604 4% /betaroot /dev/ra2d 585915 440761 86562 84% /betausr ) -- In-Real-Life: Chris Torek, Univ of MD Computer Science, +1 301 454 7163 (still on trantor.umd.edu because mimsy is not yet re-news-networked) Domain: chris@mimsy.umd.edu Path: ...!uunet!mimsy!chris