rcsh@yamnet.UUCP (/999999999) (03/28/89)
Recently had posted a request for feedback for a printer driver (perceived) problem. My RT then proceeded to die (inode count on a minidisk, more later), I was down for a couple days and lost about two days of news. Fearing I may of missed the any posting concerning the printer problem. I'm asking anyone who posted any comments to repost, I also lost mail. On to a more serious issue, the one that caused my problems. I tried to create a filesystem on a 300 Meg minidisk for news. Having run out Inodes on the previous filesystem I was determined to create a filesystem more apropos to News, that is, one with a excess of Inodes. I tried a straight mkfs first: mkfs /dev/hd7 605469:96000 This proceeded to create the filesystem with the default inode amount (5% of the number of blocks). To beat this I created a proto file of an empty directory and specified the block and Inode count in the proto file. Then made the filesystem using this proto file. With the filesystem built I do a df, AIX tells me that I've already used 68% on my inodes on this empty filesystem. Thinking that df might be brain damage I proceeded to create 40,000 files with no problem. Presuming I was fine, that df was hosed, I ran uucico with this filesystem as /usr/spool and am kindly informed I have no free Inodes! I have tried reducing the number of total blocks from 605469 all the way down to 520000 in an attempt to get this working....no luck. Any Ideas? Here is a long transcript of the series of commands to just create the filesystem. This is on a 6150 model 135, running AIX 2.2.1 and the very latest updates. Comments have been added, enclosed in hard brackets ([]). --Begin Transcript-- [ First create the minidisk on the *Porta-file* ] MINIDISK CUSTOMIZING COMMANDS Minidisks commands are: Command Description add Create a new minidisk change Change mount characteristics of an existing minidisk delete Delete an existing minidisk show Show minidisk information To EXIT the minidisks command, press F3. To USE a minidisks command, type the command and press Enter. > add Processing minidisk information... Will there be an AIX file system on this minidisk ? yes no Type yes or no and press Enter. > no The following block sizes are available for this minidisk: 512 1024 2048 To choose from the list, type the blocksize and press Enter. To SHOW minidisk information in requested blocksize, press F2. > 512 Type the number of 512 blocks for this minidisk and press Enter. To SHOW minidisk information in requested blocksize, press F2. > 605469 The following fixed disks are available: hdisk0 hdisk1 hdisk2 hdisk5 Type the fixed disk on which to create this minidisk and press Enter. To SHOW minidisk information in requested blocksize, press F2. > hdisk5 If you do not choose a location for this minidisk on the disk, the minidisk will be placed in the first available free space on the disk. Do you want to choose a location ? yes no Type yes or no and press Enter. To SHOW minidisk information in requested blocksize, press F2. > n Minidisk Name: hd7 The following information can be changed: Name Description Current Possible Choice Choices bs Block Size 512 512,1024,2048 nob Number of Blocks 605469 fd Fixed Disk hdisk5 hdisk0,hdisk1,... To CHANGE a current choice, type the name followed by your new choice (example: nob 1000), and press Enter. To CANCEL creation of this minidisk, press F3. To CREATE this minidisk with the current choices, press Enter. To SHOW minidisk information, press F2. > Creating the minidisk... The minidisk has been successfully created. The minidisk name is hd7 To CONTINUE, press Enter. > End Minidisk Session. [ Now the proto-file ] # mkdir empty_directory # proto empty_directory >proto_file # vi proto_file [ Add Filesystem Spec Lines ] # cat proto_file : Prototype file for empty_directory /etc/boot 605469 96000 : FS line: boot num_blks num_inodes d--755 0 0 $ $ # mkfs /dev/hd7 proto_file mkfs: destroy /dev/rhd7 (y)?y Bootstrap Length 28148 (0x6e18), seglen 77824 (0x13000) Device /dev/rhd7: From prototype: proto_file fs size: 605469 inodes: 96000 [ <- Note number of Inodes ] interleave: 508:4 cluster: 2048 fsys id: NONAME pack id: VOLXXX boot pgm: /etc/boot # mount /dev/hd7 /usr/empty/directory # df Device Mounted on total free used ifree used /dev/hd0 / 38684 8700 77% 9395 6% /dev/hd2 /usr 96808 40652 58% 23506 6% /dev/hd6 /vrm 3992 1012 74% 32 75% /dev/hd1 /u 2772 2752 0% 732 0% /dev/hd3 /tmp 4776 4644 2% 1275 0% /dev/hd11 /v 137776 117212 14% 31279 2% /dev/hd10 /w 139556 82268 41% 14730 17% /dev/hd7 /usr/empty_directory 593404 593400 0% 30462 68% [ ** Note Inode Free (ifree) count and percentage! ** ] # ls -a emp* . .. # -- End Transcript -- Any input would be greatly appreciated, this is getting to be a bit too much. The remaining question in my mind is if I drive my '88 Honda CRX (standard model) into an 6150 RT, model 135 running AIX 2.2.1, impacting at 55 mph, will it damage my car at all? -- Greg Noel [Gn] Delphi Information Systems, Westlake Vlg, Ca. The food is bad, and the portions are small. -W Allen
njs@scifi.UUCP (Nicholas J. Simicich) (03/29/89)
In article <106@yamnet.UUCP> rcsh@yamnet.UUCP writes: > (A lot of stuff about not being able to create a file system > with a lot of inodes). Well, unfortunately, AIX/RT is uses 16 bits for ino_t, as an unsigned short. So there is no way you can have more than 2**16-1 (65535) inodes. I suspect that poor df is underflowing or something. Unfortunately, however, I've never been successful in convincing mkfs to actually construct a file system with more than 32000 inodes. I suspect it is a hard coded limit, set that way out of paranoia worries regarding programs that might want to treat the inode number as a signed short. Makes it tough to keep a full feed for two weeks, but I found some groups to expire early :-). In any case, the output from df seems bogus. Additionally, 2.2 as distributed had the disappearing inode bug. If you are trying to run news on 2.2, make sure you get a recent version of the fix diskettes from support. -- Nick Simicich --- uunet!bywater!scifi!njs --- njs@ibm.com (Internet)
jim@applix.UUCP (Jim Morton) (03/30/89)
In article <106@yamnet.UUCP>, rcsh@yamnet.UUCP (/999999999) writes: > I tried to create a filesystem on a 300 Meg minidisk for news. Having run > out of Inodes on the previous filesystem I was determined to create a > filesystem more apropos to News, that is, one with a excess of Inodes. > I tried a straight mkfs first: mkfs /dev/hd7 605469:96000 I always find it easier to just do a backup by NAME, boot up the installation and maintenance diskette, and use the "make a file system" option which asks you for how many inodes (files). Then reboot AIX and restore your partition's files. Make sure you do a backup by name, though, or you won't be able to restore your files or it will restore the same old minidisk layout and file system parameters. -- Jim Morton, APPLiX Inc., Westboro, MA UUCP: ...harvard!m2c!applix!jim jim@applix.m2c.org
gn@yamnet.UUCP (/999999999) (04/08/89)
In article <568@scifi.UUCP>, njs@scifi.UUCP (Nicholas J. Simicich) writes: > In article <106@yamnet.UUCP> rcsh@yamnet.UUCP writes: > > (A lot of stuff about not being able to create a file system > > with a lot of inodes). [Including Proto file examples ] > > Well, unfortunately, AIX/RT is uses 16 bits for ino_t, as an unsigned > short. So there is no way you can have more than 2**16-1 (65535) > inodes. I suspect that poor df is underflowing or something. > Unfortunately, however, I've never been successful in convincing mkfs > to actually construct a file system with more than 32000 inodes. I This turns out to be the problem, and I have successfully created a minidisk/filesystem with 64000 inodes. You are correct, mkfs will not except more then 32000 as a parameter, so you have to fool it. The way I accomplished this is with a proto generated file. The problem isn't just with df though, if it was, I could live with it - uucp also checks if there are enough inodes to do business and fails. Thanks to those who helped me out on this. Has anyone had any luck with the problem of the open to the printer that never fails? No help from IBM on this (not surprisingly). From the front line: Level I & II support for AIX 2.2 ends 5/1. IBM policy states that they will support both the current version and the previous version of AIX. Asking about what appears to be a contradiction, I was told that there was no contradiction, they will be supporting 2.1 (previous version) and 2.2.1 (current). The changes between 2.2 and 2.2.1 strike me as much more severe then that of 2.1 to 2.2. Bad news when you have an installed base of +300 machines mostly running 2.2, and none running 2.2.1 (we haven't upgraded our turn-key package yet). This lack of support is discouraging, and would upset me much more if support was actually able to resolve the problems I've been encountering with AIX (gumbyware like snaconfig for one). Anyone had any real luck (other then quickly delivered updates) from IBM support? Isn't there a charge to update to 2.2.1 ? (we have a couple of machines running 2.2.1, but I don't handle the purchasing) -- Greg Noel [Gn] Delphi Information Systems, Westlake Vlg, Ca. Nothing Works, and Nobody Cares -W. Allen -- -- Greg Noel [Gn] Delphi Information Systems, Westlake Vlg, Ca. The food is bad, and the portions are small. -W Allen