Path: utzoo!utgpu!cunews!bnrgate!brchh104!brchs1!bnr.ca!rice.edu!sun-spots-request From: codonics!codonics.com!bret@uunet.uu.net (Bret Orsburn) Newsgroups: comp.sys.sun Subject: Inode density problem Keywords: No Digest Subjects in Unmoderated Mode Message-ID: <3434@brchh104.bnr.ca> Date: 4 Jun 91 18:40:00 GMT Sender: news@brchh104.bnr.ca Organization: Sunspots, Psuedo-Unmoderated Lines: 39 Approved: sun-spots@rice.edu X-Original-Date: 22 May 91 02:49:28 GMT I'm having trouble achieving an acceptable inode density on a 1GB+ external SCSI disk drive. The system is a 3/80 running SunOS 4.0.3. Here are the facts as I understand them: The OS has a compiled-in maximum number of inodes per cylinder group, called MAXIPG. This is given in as 2048. The disk drive is a Fujitsu M2266 with 15 heads and 46,635[*] bytes per track. 15 * 46,635 = 699,525 bytes per cylinder. The default number of cylinders per group is 16, which gives 16 * 699,525 = 11,192,400 bytes per group. 11,192,400/MAXIPG = (approx.) 5465, so the default inode density comes out to around one inode per 5465 bytes. That is OK for some things, but is unacceptable for our news partition. The documentation leads me to believe that the correct thing to do is to change the number of cylinders per cylinder group. The man page for mkfs says this parameter (ncpg) must be in the range 1-32. When I try to specify ncpg to be 8, mkfs quits with an error message to the effect that ncpg must be a multiple of 16! (I have tried to change this parameter through newfs and directly through mkfs, with the same result.) So: am I confused or is this a mkfs bug? What should I do to increase my inode density? Any ( well, most :-) suggestions appreciated. bret@codonics.com uunet!codonics!bret Bret Orsburn [*] The number of bytes per track given may not be precisely correct, but it is close enough for the purposes of this exposition.