From: Digestifier To: Linux-Admin@senator-bedfellow.mit.edu Reply-To: Linux-Admin@senator-bedfellow.mit.edu Date: Wed, 24 Nov 93 22:13:16 EST Subject: Linux-Admin Digest #186 Linux-Admin Digest #186, Volume #1 Wed, 24 Nov 93 22:13:16 EST Contents: Re: ext[2]fs or xiafs better for news processing? (Stephen Tweedie) Re: adjtime() for Linux? (Steve Tinney) Re: Any LINUX installations in real life environments? (Jim Morris) Re: adjtime() for Linux? (Frank Kardel) Re: Incompatibilities between ext2fs and mount! (Kai Petzke) Re: Watcher... (Harald Milz) Re: Shadow passwords? (Rob Janssen) Re: File System Corruption Lessons Learned (long) (Serge Robyns) Re: "New" directory structure... (Jim Dodd) Re: Watcher... (Bernd Meyer) ld.so save space? (Aubrey Jaffer) Re: Berkeley Fast Filesystem (Mats Wichmann) Re: Default shell on login (Titusz Feher) Process table filling up with zombies (Michael T. Nygard) Re: Network reliability? ("John E. Davis") Re: Slackware: fsck always check at boot time (Bill C. Riemers) Re: How to send commands to /dev/modem or /dev/cua1 ? (JJWISEMAN@DELPHI.COM) lpd - Printer daemon (Tung Nguyen) ---------------------------------------------------------------------------- From: sct@dcs.ed.ac.uk (Stephen Tweedie) Subject: Re: ext[2]fs or xiafs better for news processing? Date: Wed, 24 Nov 1993 14:57:30 GMT > Hi, > which filesystem is likely to give the best performance when dealing > with news processing (many small files, large directories, etc.)? The ext2fs includes a file allocation heuristic which is quite good at keeping related files close together, so it does offer good performance for news partitions in which operations often have to be repeated over all files in a directory (eg. news expires). However, obtaining objective results is hard, because no matter which filesystem you use, performance will degrade somewhat over time due to fragmentation. From the evidence I have, ext2fs does appear to suffer less from fragmentation than other filesystems. There's another point - filesystem checking. There is a new version of e2fsck under development, and the latest public pre-release is on tsx-11.mit.edu in /pub/linux/ALPHA/ext2fs. Although it is still officially in ALPHA, I don't know of any major problems with it, and it does fix filesystem faults which cannot be fixed by any of the other filesystems' fsck systems. The only buggette I know of in the public ALPHA is that it is occasionally necessary to run e2fsck twice in order to correct some really nasty problems (for example if the root directory inode needs to be reconstructed for some reason). It is also *very* much faster: of the order of 500% faster than the old e2fsck. The new version is much better at lost+found handling than older ones. You can thank Theodore Ts'o for this code. Cheers, Stephen. --- Stephen Tweedie (JANET: sct@uk.ac.ed.dcs) Department of Computer Science, Edinburgh University, Scotland. ------------------------------ From: sjt@enlil.museum.upenn.edu (Steve Tinney) Subject: Re: adjtime() for Linux? Date: 24 Nov 1993 16:01:29 GMT OK, Harald Koenig supplied the straight dope: the routine was included in pl13c-pl13n (or thereabouts). It was temporarily removed because of a library conflict, but will re-emerge in all its glory in the next official release of the kernel/libc. He suggested grabbing it from one of those, which I did (pl13j, linux/include/linux/timex.h, for the record). Thanks everyone, Steve ------------------------------ From: jmorris@darkstar.rastek.com (Jim Morris) Subject: Re: Any LINUX installations in real life environments? Date: 24 Nov 1993 16:24:01 GMT I noted several responses to this thread from educational types, so I thought I would make some input as a Linux user in a true commercial environment. I work for a small software development house, where we develope embedded software for 680x0 and various RISC controllers. We have run PCTCP networking for years, using sevral Unix-based file/mail servers, and DOS based PC's using FTP Software's PC/TCP package for DOS. We first learned of Linux back in Febreuary of this year (2/93), and started off then with several user's PCs coverted over to Linux "workstations". Since then, most of our programmers have begun to use Linux for most of their work, except for those cased where we must boot DOS (typically to run a debugger or emulator tool). But, at least in my case, most of my coding happens under Linux, and my preliminary compilations are performed with GCC. We now have a network server that is running Linux, and is exporting both its hard drive and a CD-ROM drive to the network. In fact, that is its main purpose - to share the CD-ROM drive, which none of our commercial Unices have a driver for. Works great. This server has been up for several months now, with a load varying from 2-15 users. Uptime has been up to 20+ days, and would be longer, except for local power outages that seem to happen at least once or twice a month. Proably the biggest reason I keep DOS around at all is for intensive word processing, and for a few DOS-based development tools that I need from time to time. Other than that, Linux is very viable in a commercial development environment, in my opinion. At the very least, even if Linux doesn't meet the needs of end users on your network, it provides a VERY economical alternative for a network server OS. It is at least as stable as Intel-based Unices that we have paid 1000's for! Jim Morris (jmorris@rastek.com) ------------------------------ From: kardel@nessy.informatik.uni-erlangen.de (Frank Kardel) Subject: Re: adjtime() for Linux? Date: Wed, 24 Nov 1993 16:50:27 GMT sjt@enlil.museum.upenn.edu (Steve Tinney) writes: >I'd like to bring up nistime on my Linux box and ftp'd the source >but it requires adjtime(), which I have been unable to find >anywhere. Anyone done this already? > Steve PL 14 will have it. (And will support NTP) Frank Kardel ------------------------------ From: wpp@marie.physik.tu-berlin.de (Kai Petzke) Subject: Re: Incompatibilities between ext2fs and mount! Date: 24 Nov 1993 13:53:59 GMT In <2cvdrn$ejn@sun8.ruf.uni-freiburg.de> schwunn@sun1.ruf.uni-freiburg.de (Erwin Schwunn) writes: >I have changed to the new "ext2fs" (V0.4) programs. But when mounting >the mount message says that ext2fs 0.3c ist mounted. Likely it doesn't >hurd but it not only takes time to proove that yourself have not made >a mistake. The mistake is not with the mount programme - it is with different versions you may have. With a 99p13 kernel, you have some of ext2fs 0.3c, or 0.3d, depending on the ALPHA level you use. Changing the binaries (mke2fs, e2fsck, etc.) does not change the kernel code, too. Anyway, as far as I understand things, the functionality was changed, but the data format kept the same. This was possible, because ext2 was carefully designed to be extensible. You should not discover incompatibilities between 0.3 and 0.4 (of course, you will discover incompatibilities, when using the new features like synchronous files, etc.) Kai -- Kai wpp@marie.physik.tu-berlin.de Advertisement by Microsoft in a well-known German magazine: If you don't like our programmes, than make your own ones. However, they expect you to use Microsoft products for this -:) ------------------------------ From: hm@seneca.ix.de (Harald Milz) Subject: Re: Watcher... Date: Wed, 24 Nov 1993 07:24:40 GMT Reply-To: hm@seneca.ix.de Rob Shady (crt@tiamat.umd.umich.edu) wrote: : > Actually... What I think the original poster meant, was something that you : > can actually whatch what a user on a tty is typing, and perhaps even : > intervine and be able to type commands for the user while they watch. : > Ie: Like a PC BBS system... BTW- I have been looking for something like : > this for quite a long time, if anybody else knows anything, let me know : > as well.. Mail me, since I don't read this group often.. No. Again, this appears to be a question of major interest, so if you have an answer, please _at least_ post. Thank you. Ciao, hm -- Harald Milz (hm@seneca.ix.de) ------------------------------ Crossposted-To: comp.os.linux.help From: rob@pe1chl.ampr.org (Rob Janssen) Subject: Re: Shadow passwords? Date: Wed, 24 Nov 1993 16:14:34 GMT In <2cue0n$eu2@organpipe.uug.arizona.edu> ron@argus.lpl.Arizona.EDU (Ron Watkins) writes: >What are shadow passwords. I haven't run across them before. How do I know >what in my Linux distribution is using them? Is it possible to avoid using >them? I would rather have an old-fashon UNIX setup, small and simple. I >don't want excess stuff in my directories. Please mail any info to >ron@argus.lpl.arizona.edu. Shadow passwords move the password information from the publicly readable /etc/passwd to a more restricted /etc/shadow file. This file also has some extra fields to carry info like password expiration date. It is possible to disable it by recompiling some programs with different options, but you will certainly have more trouble with that (and the effects on future installation of upgrades, networking etc) than you now save by keeping things like you are accustomed to! Rob -- ========================================================================= | Rob Janssen | AMPRnet: rob@pe1chl.ampr.org | | | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU | | e-mail: pe1chl@rabo.nl | Tel. BBS: +31-30715610 (23:00-07:30 LT) | ------------------------------ From: sr@tyana.denkart.be (Serge Robyns) Subject: Re: File System Corruption Lessons Learned (long) Date: 24 Nov 1993 17:10:23 +0100 INN NEWS (news@apollo.west.oic.com) wrote: :>In article <1993Nov22.153621.25661@spdc.ti.com> andy@piziali.lonestar.org (Andrew J. Piziali) writes: :>: After finally getting my Linux system back up and running after a week of :>:unsuccessful attempts to restore the root (and only) file system, I feel :>:... :> I have modified my e2fsck to NOT automatically attach lost inodes to :> lost+found. I made this change after experimenting a bit, and found that :> e2fsck could get into a state where multiple runs would only corrupt :> the filesystem MORE rather then less, and traced the problem to it's :> attempt to connect lost inodes to lost+found. I agree, on a HP-UX troubleshooting course we learned to use fsdb (File System DeBugger). We took the output of a fsck and then look with fsdb what's really wrong and fix it like it has to be. We did it during lab exercises and it work fine. Even just looking with fsdb and answering yes and the correct places in fsck helps a lot. We just miss fsdb. ^ ^ o o =======oOO=(_)=OOo============================================================= Serge Robyns - System Engineer and Postmaster Denkart n.v. Molenweg 107 B-2830 Willebroek - Belgium Voice: +32 (3) 866-0022 Fax: +32 (3) 866-0301 e-mail: sr@denkart.be (postmaster@denkart.be) ------------------------------ From: jimd@netcom.com (Jim Dodd) Subject: Re: "New" directory structure... Date: Wed, 24 Nov 1993 18:23:21 GMT Itai Zukerman (zukerman@rosarita.berkeley.edu) wrote: : i've seen mentioned in several places that "linux" is moving to a : new directory arrangement, specifically involving the creation of : /sbin, some shifts in /etc and possibly other changes. (i realize : this is also a matter of taste, hence the quotes). for example, the : lilo documentation comes with QuickInst.new (or somesuch) supporting : the so-called "new" structure. : has there been discussion of this? have i missed it? are there : references? where can i read about the pros and cons of different : directory arrangements? Another reson to go with the shifting of executable files from /etc to /sbin is that it more closely follows the System V standard. Most commercial vendors these days use /sbin for utilities needed during booting like fsck and mount. In fact, these utilities are usually statically linked. /usr/sbin is used for sys admin utilities like mkfs, route, reboot, and ypset. Often, you will find dynamically linked versions of the same utilities found in /sbin. By being more conforming to the current standard will make using other Unix systems more "comfortable". -- Jim Dodd email: jimd@netcom.com ------------------------------ From: root@umibox.hanse.de (Bernd Meyer) Subject: Re: Watcher... Date: Wed, 24 Nov 1993 05:04:42 GMT crt@tiamat.umd.umich.edu (Rob Shady) writes: >webster@kaiwan.com (Tom Webster) writes: >[Description of various definitions of 'WATCH' deleted] >Actually... What I think the original poster meant, was something that you >can actually whatch what a user on a tty is typing, and perhaps even >intervine and be able to type commands for the user while they watch. >Ie: Like a PC BBS system... BTW- I have been looking for something like >this for quite a long time, if anybody else knows anything, let me know >as well.. Mail me, since I don't read this group often.. I did this many months ago to use on guest logins. A version that did just the watch-part was posted on a German linux group (it was hacked from script), and I then added the other part. But it has mahor drawbacks - the main one being that the controlling tty of the user's shell isn't the serial port anymore, but a pseudo-tty. Thus the user has no control whatsoever over the serial port, and this messed up Z-modem downloads for me, so I decided that sneaking wasn't a nice thing from the start, and deleted everything. Sorry. It can be done, but to do it right might be very hard. Bernie -- We both know that the earth is round | Bernd Meyer, EE-student So we can't see the way before us to its end | "Nobody is a failure who has We walk on this way, hand in hand, | friends" (from: "It's a And I hope you are still with me behind the horizon| wonderful life") ------------------------------ From: jaffer@zurich.ai.mit.edu (Aubrey Jaffer) Subject: ld.so save space? Date: 24 Nov 93 15:09:28 I am still running Linux SLS1.01 and Xfree86 V1.2 and have been considering upgrading. Space is tight on my disk and I would like to know if shared executables now result in a smaller distribution for Linux. Also, I would like to mount my root partition read-only, or at least have nothing in it which should be changing so that my system would be more fs-damage resistant and could run e2fsck on the rest of the partitions. On my old pre-linux Unix system this was done by having /usr be a separate partition (which had all the user and spool directories). So my question is: Which directories should be in the root partition (/etc and /bin?) and how large should my root partition be to hold the Slackware 1.10 versions of these directories? Is this information in Slackware 1.10? It wasn't in SLS1.01. ------------------------------ From: mats@netcom.com (Mats Wichmann) Subject: Re: Berkeley Fast Filesystem Date: Wed, 24 Nov 1993 19:54:13 GMT a09878@giant.rsoft.bc.ca (Curt Sampson) writes: >Generally, the performance of the FFS should stay constant over time. >The performance of more traditional Unix filesystems goes down as the >file system is used and it fragments. The allocation policy in FFS gets drastically worse if it doesn't have a fair amount of free space to pick from. This is where the magical "10% free" number comes from. The need to leave 10% free in order to keep reasonable (write) performance is one of the drawbacks of FFS, although some feel it's a reasonable tradeoff. -- Mats Wichmann Software Consultant mats@netcom.com ------------------------------ Subject: Re: Default shell on login From: tif@hercules.elte.hu (Titusz Feher) Date: 24 Nov 93 20:18:20 +0100 koch@rtsg.mot.com (Clifton Koch) writes: : It appears that bash is not getting run at login because my .bashrc is : not being executed. If I manually type 'bash' after login, the .bashrc : is executed and everything is fine. Bash is in /bin, and /bin/bash is in : /etc/shells. The file executed by bash at login is ~/.bash_profile (see man bash). Best, Titusz ------------------------------ From: nygardm@cco.caltech.edu (Michael T. Nygard) Crossposted-To: comp.os.linux.help Subject: Process table filling up with zombies Date: 24 Nov 1993 20:36:06 GMT Hello, world! I'm having some difficulties with process tables filling up. If my box sits idle for more than about 2 hours, I find that the process table gets filled with zombies spawned by rpc.portmap. I can get rid of the zombies by killing rpc.portmap, but I'd like to know if there is a solution. The only common denominator I've been able to find is idle time. Load, network connections, number of processes running, none of these matter. It only happens when no users are logged in. For details, I'm running pl13r, 16 megs RAM, NET-2-debugged. Please email me with any leads. Mike Nygard -- +==================================================+=========================+ |Infinity welcomes careful drivers | Mike Nygard | |Better Than Life - I can't wait for the next one! | nygardm@cco.caltech.edu | +========================================+===================================+ ------------------------------ From: davis@pacific.mps.ohio-state.edu ("John E. Davis") Subject: Re: Network reliability? Reply-To: davis@pacific.mps.ohio-state.edu (John E. Davis) Date: Wed, 24 Nov 1993 21:45:58 GMT In article <1993Nov21.001634.6812@msus1.msus.edu> volkerdi@mhd1.moorhead.msus.edu (Patrick J. Volkerding) writes: You don't have the "latest net stuff" :^) - in.telnetd and a couple other things have been fixed since then. You should get the fixed net utilities from tsx-11.mit.edu, I think they are in /pub/linux/packages/net/new-net2. Are these already in the sources for the 0.99.13 kernel which is also found at tsx-11? -- _____________ #___/John E. Davis\_________________________________________________________ # # internet: davis@amy.tch.harvard.edu # bitnet: davis@ohstpy # office: 617-735-6746 # ------------------------------ From: bcr@bohr.physics.purdue.edu (Bill C. Riemers) Subject: Re: Slackware: fsck always check at boot time Date: 24 Nov 93 19:47:18 GMT I think the point being missed here, is that only the latest bootutils umount command cleanly unmounts root. For those who prefere the usermount version of mount and umount, the simple solution is to rename the bootutils umount to root_umount. Then in brc use "root_umount -a". All methods I've tried to use usermount's umount in brc have always resulted in a failed to unmount error, and the root partition being left marked unclean. Bill ------------------------------ From: jjwisema@news.delphi.com (JJWISEMAN@DELPHI.COM) Subject: Re: How to send commands to /dev/modem or /dev/cua1 ? Date: 24 Nov 1993 19:19:46 -0500 Hmm, for me while : do echo "ath0" sleep 60 done > /dev/modem works. If I just redirect the echo, it seems to happen too fast for the modem. ------------------------------ From: tnguyen@ucssun1.sdsu.edu (Tung Nguyen) Subject: lpd - Printer daemon Date: 25 Nov 1993 02:45:52 GMT I have an HP-III laser jet printer and I'm having a problem getting the printer daemon (lpd) to run. I created the directory and files specify in the printcap file (/usr/spool/lp1/hplj/*) but when I start the lpd daemon it won't run. Does anyone have a working printcap file for an HP-III printer or similiar help me with this? I would appreciate any kind of help on this. If possible please email me directly. Thanks ahead. TN ------------------------------ ** FOR YOUR REFERENCE ** The service address, to which questions about the list itself and requests to be added to or deleted from it should be directed, is: Internet: Linux-Admin-Request@NEWS-DIGESTS.MIT.EDU You can send mail to the entire list (and comp.os.linux.admin) via: Internet: Linux-Admin@NEWS-DIGESTS.MIT.EDU Linux may be obtained via one of these FTP sites: nic.funet.fi pub/OS/Linux tsx-11.mit.edu pub/linux sunsite.unc.edu pub/Linux End of Linux-Admin Digest ******************************