-----BEGIN PGP SIGNED MESSAGE----- Hi! > -----Ursprüngliche Nachricht----- > Von: owner-netboot@baghira.han.de > [mailto:owner-netboot@baghira.han.de]Im Auftrag von > werewitch@whoever.com > Gesendet: Freitag, 05. Jänner 2001 22:19 > An: netboot@baghira.han.de > Betreff: startup services > > > > Hi > > I am setting up a diskless client . This setup is my first > experience with linux. So please keep that in mind while > reading the following. I am using RH 6.2. For the client I am > mounting only /clients/node1 and /home as read-write. First thing that may lead to an error (see you error-message "httpd/error_log permission failure") System log is written to files in /var/log/ so this directory has to be writeable. You can do without write permissions to the log-files if you a) do not log the system messages (do not start syslogd and klogd) b) send all log infos to another machine, your server. (see man syslogd for the apropriate command line switch) But IMHO you can not do without write permissions to the partitions where the files "/var/run/utmp" "/var/log/wtmp" reside. If you do not log your system messages nothing worse happens. You just have no hint what happened in case of failure :) > All files needed for booting are not > linked to those needed by the server. > I have checked archives concerning the following problems but > to no avail. These are the error messages on the client console: > > 1) chgrp: 'utmp' is an invalid group name > why am I getting this when my /etc/group contains utmp. > (this occurs when chgrp tmp /var/run/utmp /var/log/wtmp is > executed in rc.sysinit) > No idea yet. > 2) /etc/rc.sysinit: /boot/kernel.h file or directory not > found ( I am getting this because I am using a floppy to boot > and there is no boot directory in the client root filesystem > being mounted from the server. > How crucial is this can I comment out the part that needs > this file in rc.sysinit? ) > Which section leads to this error-message? I have a feeling that this is not crucial but only needed in case of a kernel-panic when the "System.map" file is helping to track down the module that caused the failure. I might be wrong. > 3)when rc runs the startscripts in rc3.d it gives the > following errors: > Starting atd: lockd: cannot monitor 192.168.1.1 > lockd: failed to monitor 192.168.1.1 [OK] > (but then it stops as shown by the status command which i > used in the rc script for debugging purpose) > There are two messages mixed together. Prior to the daemon "atd", the "nfs" daemon seems to be started, immediately switching to the background. "lockd" is part of the NFS-server package and used for accepting locking requests for files. "atd" is a daemon used to start programs at a special time and then logout. Unlike cron, which can be used to manage automatic, time-driven and repeated execution over a period of one year, atd is used for single execution of one task within the next 24 hours. So you really do not need it. "lockd" is only needed if you run an NFS server (silly thing on the client!). Perhaps you have copied the whole "/etc" directory of your server to "/clients/node1/etc" the "etc" directory for the client. Maybe you forgot to change the configuratition to match the different IP address of your client. Nevertheless you definitely do not need the NFS daemon! > .......................... > ............................ > > Starting sendmail: > lockd: cannot monitor 192.168.1.1 > lockd: failed to monitor 192.168.1.1 > (btw sendmail gives status running ) Again "lockd" tries to start in the background, failing and printing an error-message just as sendmail is being started!! You definitly do not need sendmail. If you really need a mail-transfer-agent (MTA) then install something like exim-mailer or another small mailer-dameon that just forwards every mail to your Server (used as mail-hub) > httpd runs but then gives a stream of the above lockd > messages then i get > NFS: cannot silly-delete .nfs[lots of digits] after xfs > script is executed and at the same time I get the following > on the server: > fh_verify: httpd/error_log permission failure, acc=2, error=13 > fh_verify: httpd/error_log permission failure, acc=2, error=13 > fh_verify: httpd/error_log permission failure, acc=2, error=13 > fh_verify: httpd/error_log permission failure, acc=2, error=13 > fh_verify: httpd/error_log permission failure, acc=2, error=13 > fh_verify: httpd/error_log permission failure, acc=2, error=13 > fh_verify: httpd/error_log permission failure, acc=2, error=13 > fh_verify: httpd/error_log permission failure, acc=2, error=13 > fh_verify: httpd/error_log permission failure, acc=2, error=13 > fh_verify: var/lock permission failure, acc=13, error=13 The httpd daemon can not write to the log-files because /var is mounted write-protected. I am certain yo do not need a http-daemon!! You can not redirect the log-output of apache to another system because it uses its own mechanism an does not rely on syslogd!! But you can redirect the output to /dev/null > ( my kernel is 2.2.14-5.0 and I have read a couple of posts > about this bug related to silly-delete but I am not sure if > this is because of the bug or because of some lock. If it is > a bug can you please tell me where > to get the patch and how to install it keeping > in mind that /clients/node1 is the root for the client. ) > I do not think that this is a bug. You are starting an NFS server-daemon on the client but you do not need the daemon to mount NFS shares. Mounting NFS shares is completly done transparently in the kernel. So just compile in the NFS support and do not bother with NFS userland programs!! > 4) X Font Server is started displays [OK] but then the status > says xfs dead but subsys locked . You do not need a font server on your client. Indeed this is really head-bangig!! Just start one font-server on your central machine (the NFS-server) and configure your client xwindows to use this one. Now you can centralized manage (add/remove) fonts which is the purpose of a font-server. > After startscript for linuxconf is executed i get an endless stream > of I looked at the start-script of linuxconf and to mee it seems that you need linuxconf just to configure your linux-box. So You do not need it anymore after you have successfully configured your box! > lockd: cannot monitor ip > lockd: failed to monitor ip > and thats it no login prompt either. > Look at your "/etc/inittab" file. Do you have any entry like: "1:2345:respawn:/sbin/getty 38400 tty1". If so be sure that "/sbin/getty" can be found and executed. This is the console-frontend to login. To use a console be sure that you included support in the kernel-configuration. If you want an xwindows login instead, install and make "gdm" or "xdm" start at boot-time! > 5) I'd like to know which of the startup scripts do I > absolutely have to run. > First think about what you really need!! a) Do you need to write log-files? No -> disable syslogd and klogd! b) Do you need to execute tasks in certain time-intervalls? No -> disable crond c) You definitely do not need "atd" or have you ever used it before? d) You do not need the NFS daemon for you do not have any harddisk with partitions to share! e) Do you really need any MTA? Yes -> do not use sendmail!! Use program that just forwards everything to the server So deactivate sendmail f) You definitely do not need "gpm" for you run xwindows with its own mouse-driver g) You may need "lpd" if you locally run programs and want to use the remote-printer attached to your server. If you run programs remotely with just a local visualization then you do not need "lpd" for printing, printing is done on the remote machine then. h) You do not need "routed". i) Depending on your network-configuration you may need some service like "identd" for authentication. These services are rarely needed so use "inetd" to run them. But comment out everything you do not really need. No service listed in "inetd" standard-configuration is needed to run xwindows! j) start "xdm" or "gdm" (gdm is for gnome!) an boot-time k) Do not start a x-font server! l) One HTTP server is enough, so do not start one on the client. To deactivate a daemon use "chkconfig --del <script-name>". > 6) Most of these scripts invoke services for which binaries > are in /usr/sbin or /usr/bin . A lot of howtos on the net do > not mount /usr at all. How then are these services invoked. > Perhaps they mount just the root-filesystem "/" with a whole directory tree, which is much easier, especially if you do not have to be too paranoid. I am using two diskless machines. I have copied all I need, a whole directory tree, to /home/diskless/nodeX and just mount this NFS share as a root-filesystem. This is a good way to start with. If you get more experienced you can play around tightening the system. > I have been working on the above set of problems for more > than a week with zero progress and am quite desperate. Please > do reply. > > Thanx > Did you compile your kernel with direct NFS support (not as a module)? Did you tell the compiled kernel that the root-filesystem is on a NFS share? (changing the setting directly in the image with "rdev image /dev/nfsroot" or using "root=/dev/nfsroot" as a kernel parameter) In your "/etc/exports" file you need you use a parameter (no_root_squash) to allow root (of your client) to act as "root" to your exported files. So be sure you use something like /home/diskless/node1 192.168.4.3(rw,no_root_squash) I had a lot of troubles before I used this parameter because without it the server does not consider the remote (client) root to be root on the server and denies access to some files owned by root. (like log-files) I did not have the time to mess around with changing file permissions in a try-and-error manner to do without this parameter! Hope this helps a bit. Niki - - ------------------------------------------------ Nikolaus Rosenmayr, n.rosenmayr@advance.at ICQ: 71968479 -----BEGIN PGP SIGNATURE----- Version: PGPfreeware 6.5.8 for non-commercial use <http://www.pgp.com> iQEVAwUBOlZlY6dhpQ1utBvrAQGqewgAgEu3KqOGBirX813vmgIGuKqRsCzIaB0w qrJVeAczobNobloug9rq86UKC+r/coNsvwwVmKHz70eM7hj6eNvAaCtMzEwVw+Y5 JM24dsbB+zcffjWi8d86n2OT/je1zvvEndRppU4sQIA3FwllHTIan9NS5QK4suYU oi+25VA7Bqzf2Frw+tddjL43sQLltBYSTf4Cq+kAHnwZV+ZJ1IVIRuOQUcVWvPi+ I6o8+3IQdXpzCty4C/KZGErxzb2FlICMMfIc09aM8ahLly9wq9hLa6knPj0D2W0Z /+6bef8a506RWwdfQkLPdO4j1hA0haOg0uexFhq9X4DE19+hrYXFGg== =PJlz -----END PGP SIGNATURE----- =========================================================================== This Mail was sent to netboot mailing list by: "Niki Rosenmayr" <n.rosenmayr@advance.at> To get help about this list, send a mail with 'help' as the only string in it's body to majordomo@baghira.han.de. If you have problems with this list, send a mail to netboot-owner@baghira.han.de.
For requests or suggestions regarding this mailing list archive please write to netboot@gkminix.han.de.