Re: Using FAI to install a "root Server"
> On Fri, 11 May 2007 11:05:12 +0200, "Torsten Schlabach" <[EMAIL > PROTECTED]> said: > - I have no access to the hardware, i cannot insert any CDs or the like. > - I have no power over any DHCP infrastructure in the network the box is in. > So I have set up Lilo on the server to boot into this kernel and used the append variable to add the kernel parameters for IP config and mounting the NFS root filesystem. > This seems to have worked, because if I reboot the server (I can do that through a remote control web interface) I can see that it mounts the /srv/fai/nfsroot from my FAI server. That's not bad. >> From that I conclude that I boot the right kernel and that the kernel parameters for root=/dev/nfs and the corresponding IP configuration seem to have worked. yep. > But IIUC I should - in a next step - see that the install client mounts the /srv/fai/config directory, but it does not even attempt to and I wonder why. Read the log files or put them on paste.debian.net Which FAI version are you using? There where changes from FAI 2.x to 3.X. > There is no initrd involved, is there? No. FAI currently does no use an initrd. -- regards Thomas
Re: unknown package removed
> On Thu, 5 Apr 2007 15:35:18 -0700, "Russel Hill" <[EMAIL PROTECTED]> said: > WARNING: These unknown packages are remove from the installation list: > my-package > I can explicitly load the package using apt-get, aptitude, and Did you also check this during installation inside the /tmp/target chroot, where install_packages is called? Maybe the sources.list is not correct inside the chroot. -- regards Thomas
A different cause for "Warning: unable to open an initial console"
> On Wed, 28 Mar 2007 16:48:29 +0300, [EMAIL PROTECTED] (Toomas Tamm) said: > Dear mr. Lange, > I today had an experinece with FAI which appears not to be documented > on the net. You may post my e-mail to the web if you wish, or forward > to a mailing list, but please remove my e-mail addresses to reduce > spam. > I was struggiling with "Warning: unable to open an initial console" > during boot. Exactly the same symptoms as in > http://www.mail-archive.com/linux-fai@rrz.uni-koeln.de/msg03419.html > Just a reboot soon after mounting nfsroot. > The reason was quite different, however: > I had included the "udev" package into NFSROOT_PACKAGES in > /etc/fai/make-fai-nfsroot.conf because I had been struggling with udev > at a point furhter down the installation sequence. This was a "just in > case" change and I had even forgotten to create a new nfsroot, so it > had been sitting there for a while. Today I generated a new nfsroot > for unrelated reasons and suddenly FAI stopped working with the > symptoms described above. > After a lot of trial and error and with the help of subversion (always > keep your fai configuration under revision control! :-) ) I traced it > back to the nearly-empty /dev subdirectory which waits for the devfs > to be mounted... > Probably a workaround would have been to populate the /dev with > required nodes. I tried adding /dev/console and /dev/tty[0-9] but that > was apparently not sufficient. At this point I removed the udev > package from NFSROOT_PACKAGES, re-made the nfsroot, and things went > back to normal. > I hope this report, if published, will save some trouble and > frustration for someone in the future. > PS I used Debian 'stable' (sarge) on i386 throughout.
Re: rdate timeout failure
> On Sun, 18 Mar 2007 13:47:10 +0100, Rainer Fleischhacker <[EMAIL > PROTECTED]> said: > I've configured the variables TIMESRVS_1 and NTPSRVS_1 and NTPSRVS in > config space FAIBASE.var but no changing. When I turned the command off: > is there the same problem: rdate: connect: Connection timed out Are you sure you disabled the rdate called inside the nfsroot? -- regards Thomas
FAI 3.1.8 released
The next minor update of FAI is released. You can get FAI 3.1.8 and the new fai-kernels package version 1.17 by adding this line to your sources.list file. deb http://www.informatik.uni-koeln.de/fai/download etch koeln -- regards Thomas
Re: SAS disks and setup_harddisks.
> On Wed, 07 Mar 2007 12:49:21 +, "James Bromberger" <[EMAIL > PROTECTED]> said: > sfdisk: bad input Hi James, did you try to set sfdisk="--force" in a class/*.var script? or try sfdisk --force dev/cciss/c0d0 < var/log/fai/current/partition.cciss_c0d0 manually and see if this helps. -- regards Thomas
Re: Err http://127.0.0.1 etch Release.gpg
> On Wed, 7 Mar 2007 09:17:06 +, "Javi Legido" <[EMAIL PROTECTED]> said: > 1. On an etch system, today, is not possible to get running fai-cd, > becuase it relays on mkinitrd-cd, wich it's only avalaible on a sarge > system. Then, I have two ways: Yep. > a) Wait until fai-cd uses initramfs-tools instead of mkinitrd-cd Yep. > b) 'Downgrade' (with chroot I think) my Debian system from etch to sarge No. You can also setup a sarge chroot (maybe using fai dirinstall) environment on your etch system and use that for creating the ISO image with fai-cd. -- regards Thomas
Re: Err http://127.0.0.1 etch Release.gpg
> On Tue, 6 Mar 2007 16:22:59 +, "Javi Legido" <[EMAIL PROTECTED]> said: > deb http://127.0.0.1/debian etch main contrib non-free > First trouble, when I call fai-cd, is that when I try to install > my_version_1.iso, it crashed because directory > /var/lib/fai/config/class not found. > I took a look in /srv/fai/nfsroot/var/lib/fai/config and I found that > this directory is empty. Yes this is empty on the server. It's only filled with data on the install client or in the fai-cd. BTW, fai-cd is broken on etch, so I hope you are using FAI 3.1.7 on sarge. > FAI_LOCATION=/srv/fai/config (in /etc/fai.conf) > ...the script finds in /srv/fai/nfsroot/var/lib/fai/config to the config? > Any problem, I made this: > cp -R /srv/fai/config/* /srv/fai/nfsroot/var/lib/fai/config/ That's wrong. /srv/fai/config is corrent. When building the fai-cd this directory will be bind mounted. > Err http://127.0.0.1 etch Release.gpg > Could not connect to 127.0.0.1:80 (127.0.0.1). - connect (111 During an installation from CD, you have no web server running on 127.0.0.1 I'm wondering why the fai-cd uses 127.0.0.1 in sources.list. It should use the local mirror from the CD. -- regards Thomas
Re: Timezone is set to UTC after install?
> On Sat, 03 Mar 2007 12:22:48 -0500, Ralph Crongeyer <[EMAIL PROTECTED]> > said: >> The variable name as used in fai/config/scripts/FAIBASE/10-misc is >> "TIMEZONE", not "time_zone". Some variables were change in FAI 3.0 Have a look at the file /usr/share/doc/fai-client/NEWS.gz or at http://www.informatik.uni-koeln.de/fai/NEWS -- regards Thomas
Re: Problems setting up a node
> On Tue, 27 Feb 2007 11:58:58 +0100, Steffen Grunewald <[EMAIL PROTECTED]> > said: > And BTW, is there a simple way to check the file integrity of a Debian > mirror repository? Use dscverify which is part of the package devscripts. -- regards Thomas
Re: Problems setting up a node
> On Tue, 27 Feb 2007 11:58:58 +0100, Steffen Grunewald <[EMAIL PROTECTED]> > said: > I keep getting random MD5Sum errors, and sometimes even base.tgz doesn't > unpack cleanly. Mmmm. If something went wrong, check the cable. No kidding! Do you use NFS over UDP on Gbit? Don't do this. With Gbit you should use TCP for NFS, which is default with FAI 3. > And BTW, is there a simple way to check the file integrity of a Debian > mirror repository? If you manage your mirror using debmirror, I would use adding the option --md5sums. -- regards Thomas
Re: Creating image file ... failed during creation with mke2fs !
> On Fri, 23 Feb 2007 14:06:28 +0100, "Javi Legido" <[EMAIL PROTECTED]> > said: > First of all I wanna thanks your great job and your fast answer. > I put 'set -x' just below 'echo y | /sbin/mke2fs -i $bytes_per_inode > -m 0 -q $tmpimage' in /usr/sbin/mkinitrd-cd > Then it appeared this line: > ¿Se continúa de todas formas? (s,n) > In english: it continue anyway? (y,n) > Then the spanish user's we should change: > 'echo y | /sbin/mke2fs -i $bytes_per_inode -m 0 -q $tmpimage' > to > 'echo s | /sbin/mke2fs -i $bytes_per_inode -m 0 -q $tmpimage' What a weird bug. It should work if you call it this way LANG=C fai-cd or maybe LC_ALL=C fai-cd -- regards Thomas
Re: Creating image file ... failed during creation with mke2fs !
> On Thu, 22 Feb 2007 16:07:05 +0100, "Javi Legido" <[EMAIL PROTECTED]> > said: > Hi. > I have fai 3.1.6 installed on a Debian system, but when I try to > create an iso image with fai-cd I get this error: > ... > Creating image file ... failed during creation with mke2fs ! Mmm, looking at /usr/sbin/mkinitrd-cd it seems that this line fails: echo y | /sbin/mke2fs -i $bytes_per_inode -m 0 -q $tmpimage Add "set -x" to debug this script. -- regards Thomas
Re: Sending DHCP requests ....... timed out!
> On Sun, 18 Feb 2007 00:31:05 -0800 (PST), Heba Farouk <[EMAIL PROTECTED]> > said: > Im configuring a small cluster of 7 machines on > debian sarge 3.1 using network booting. Im using > CISCO gigabit switch between the machines. All things > went fine until it reaches the point which it sends > the DHCP requests to the server. I used tcpdump to > check the requests the client sends to the DHCP server > but I receive nothing. Which network cards do you have? There are known problems (and soultions) when using e1000 cards, or it may be the "fast port" (disable spanning tree) option on cisco switches. Have a look at: http://www.uni-koeln.de/bin2/maillist/linux-fai/20061206.172617.2/174229?aehnlich -- regards Thomas
FAI 3.1.7 released
A new minor release of FAI is available. It fixes two small but important bugs. FAI 3.1.7 can be downloaded via this sources.list entry. deb http://www.informatik.uni-koeln.de/fai/download etch koeln -- regards Thomas
Re: installation time needed for gnome and scrollkeeper
> On Tue, 13 Feb 2007 19:17:22 +0100, Oliver Osburg <[EMAIL PROTECTED]> > said: > I just finish a new FAI server based on etch and the installation times > are horrible (>4000 seconds) and as far as I can see the time is consumed > by endless runs of "gconf*" and scrollkeeper. Just did a test using etch on a P4 2.6Ghz with 1GB RAM and a 120GB hard disk using the simple examples (with GNOME and XORG) from FAI 3.1.6. . . The install took 556 seconds. Calling task_chboot So, there must be wrong in your environment. Read the log files and check if your disk has DMA enabled and check DNS. -- regards Thomas
Re: FAI won't install snort?
> On Fri, 9 Feb 2007 13:54:07 -0500, Ralph Crongeyer <[EMAIL PROTECTED]> > said: > mkdir: cannot create directory `/var/log/fai/current': No such file or directory strange. I never saw this. > dpkg: error processing /var/cache/apt/archives/snort_2.3.3-11_i386.deb > (--unpack): > subprocess pre-installation script returned error exit status 1 This is the main problem. snort does not install correctly inside the chroot environment. I can reproduce this error, but I do not know a solution. I'm pretty sure this is not a fault of FAI but a bug in snort's pre or postinst scripts. -- regards Thomas
Re: FAI won't install snort?
> On Fri, 9 Feb 2007 11:30:58 -0500, Ralph Crongeyer <[EMAIL PROTECTED]> > said: > I've added snort to /srv/fai/config/package_config/FAIBASE but it doesn't > get installed? > How can I fix this? Read the log files, for e.g. fai.log. -- regards Thomas
Re: Hostname is the IP?
> On Thu, 8 Feb 2007 12:42:38 -0500, Ralph Crongeyer <[EMAIL PROTECTED]> > said: > When I do an install via PXE and DHCP the hostname is the IP address of the > box? > How do I change this? Add this line to your dhcpd.conf: use-host-decl-names on; -- regards Thomas
Re: Define FAI_CONFIG_SRC from dhcp info
> On Tue, 06 Feb 2007 19:52:25 +0100, Nicolas Courtel <[EMAIL PROTECTED]> > said: > My FAI server (version 3.1.6 for Etch) is connected to several networks, > and has a different name on each network. > I have configured the DHCP server so that it sends the appropriate name > on each interface, and would like to do the same thing for the > FAI_CONFIG_SRC variable : I'm looking for an FAI variable which would be > set using the "server-name" dhcp variable, and could be used as the NFS > server in FAI_CONFIG_SRC. Have a look at dhclient-perl. There you see that the DHCP vendor option 170 are mapped to FAI_CONFIG_SRC. option_170 FAI_CONFIG_SRC So you should use this in dhcpd.conf option option-170 "nfs://faiserver/usr/local/share/fai"; # FAI_CONFIG_SRC But I never tested this. > if there's a another way to > dynamically set FAI_CONFIG_SRC? Another trick would be to write some shell code into fai.conf inside the nfsroot. This file is sourced and defines FAI_CONFIG_SRC. So you could write some code that sets this variable depending on your network environment. But if the DHCP thing work, I guess that's the best solution for you. BTW. Maybe you like to update your FAI questionnaire from 2004 and sent it back to me. -- regards Thomas
Re: Problems with FAI installation from USB DVD drive
> On Fri, 02 Feb 2007 19:56:46 +0100, Anders Jackson <[EMAIL PROTECTED]> > said: > Wouldn't it be easier to use a USB memmory stick with write protecktion > for this? I did not managed to boot the fai-cd from USB stick yet. Does anybody succeed in this? Any help is welcome. I hope to replace mkinitrd-cd with a new boot technic for fai-cd this year. -- regards Thomas
Re: Problems with FAI installation from USB DVD drive
> On Fri, 2 Feb 2007 15:36:03 +0100, <[EMAIL PROTECTED]> said: > In our company, we're installing all our linux systems (debian sarge) > with FAI. Nice. How many? Have you filled out the FAI questionnaire? http://www.informatik.uni-koeln.de/fai/questionnaire > Now I wanna reach the FAI support for Dell Latitude D420. It's a > subnotebook. > The problem is now, this model doesn't have a CD or DVD drive build-in. > There's only a DVD drive aviable which can be connectet with USB. > I can boot from my FAI CD Is this a FAI CD created with the fai-cd command or using make-fai-bootfloppy and put that image onto a CD? I think you are using the fai-cd command. But a better approach is to build a boot floppy image and put this onto a CD. Then you can boot the fai kernel from CD, but you will get the NFSROOT and config space via NFS. I hope you are allowed to use NFS in your enironment ;-) -- regards Thomas
Re: Logfiles
> On Wed, 31 Jan 2007 15:55:33 +0100, Adrian von Bidder <[EMAIL PROTECTED]> > said: > FAI_LOGPROTO=ssh and LOGUSER=fai are set. Also LOGSERVER or SERVER must be set. Otherwise the remote copying of the log files will not be executed. -- regards Thomas
Re: ssh key / fai-chboot options?
> On Wed, 31 Jan 2007 15:25:57 +0100, Adrian von Bidder <[EMAIL PROTECTED]> > said: > * to use the initrd trick for all hosts, does fai-chboot has an easy option > so I don't have to modify the pxelinux.cfg file manually each time? (I > don't want to create a default file - I don't want mistakes to happen when > somebody randomly net-boots some machine!) Currently fai-chboot does not has options for adding initrd information. But you can create a template manually, and then use the copy option -c. See man fai-chboot for more details. -- regards Thomas
Re: Problem with e1000 driver
> On Tue, 30 Jan 2007 09:52:51 +0100, Adrian von Bidder <[EMAIL PROTECTED]> > said: > linux-source-2.6.18 2.6.18.dfsg.1-9), I can't boot boxes with e1000 I have the same problem on my thinkpad. > use nfs root. Is there a hack that'd allow me setting up nfs root from > initrd (can I boot through an initrd with PXE?) Have a look at this mail. http://www.uni-koeln.de/bin2/maillist/linux-fai/20070105.220333/174065 and the following mails. Maybe I should put the kernel and the initrd on the fai webpage. -- regards Thomas
FAI 3.1.6 released. Fixes grave bug in 3.1.5
FAI 3.1.6 was just uploaded and is now available in the repository from my URL deb http://www.informatik.uni-koeln.de/fai/download etch koeln It fixes a grave bug in fai-class, that did not set HOSTNAME and so most classes were not defined. So if you are using FAI 3.1.5 you should upgrade to 3.1.6 soon. -- regards Thomas
Re: FAI cancel client install during HD partitioning.
> On Wed, 17 Jan 2007 16:06:15 -0300, Marcelo Santana <[EMAIL PROTECTED]> > said: > "Disks found:Disk /dev/cciss/c0d0: cannot get geometry" > "Illegal division by zero at /usr/sbin/setup_harddisks line 298." > ... You can force sfdisk to use a certain geometry by defining a varaible in .../class/*.var. sfdisk="--force -C 7752 -H 240 -S 63" I need this when installaing a dual boot machine, where Wind did the partitioning and sfdisk always think the partition was made for a different geometry that linux detects. It's works well for me. So try to determine which geometry you must force for your disks. P.S: Please do not attach the complete log file, if we only need a certain part of it. -- regards Thomas
Re: USB keyboard with ubuntu fai
> On Tue, 16 Jan 2007 22:24:41 -0600, Carl Caum <[EMAIL PROTECTED]> said: > I've never been able to use USB > keyboards during an FAI installation. > USB keyboards don't work. I'm running fai-server version 2.10.5 with > 2.6.15.6-ubuntu1-fai-kernels_1_i386 for running install kernel. My > Any ideas? Yes. Maybe your hardware needs a different kernel module, than those which are already loaded. You can add the kernel module to the script class/20hwdetect.source. This is some detail of this script. kernelmodules="rtc floppy usbkbd ide-disk ide-cd" case $(uname -r) in 2.4*) kernelmodules="$kernelmodules keybdev mousedev hid ide-detect isa-pnp" ;; 2.6*) kernelmodules="$kernelmodules ehci-hcd ohci-hcd uhci-hcd usbhid usbmouse ide-generic mptspi ata_piix" ;; esac -- regards Thomas
Re: FAI cancel client install during HD partitioning.
> On Tue, 16 Jan 2007 11:37:41 +0100, Stephan Hermann <[EMAIL PROTECTED]> > said: > That is one of the problems I had with our HP servers. Some of those > machines weren't ship with msdos labels or any other labels...and > partitioning failes... > how could we come over this problem via fai? Normally, FAI does not have problems with disk that have no disk labels. But a hook for task partition could help, to force a certain disk label type, or to initialize the disk label. -- regards Thomas
Re: FAI cancel client install during HD partitioning.
> On Mon, 15 Jan 2007 16:38:18 -0300, Marcelo Santana <[EMAIL PROTECTED]> > said: > sfdisk: ERROR: sector 0 does not have an msdos signature > /dev/cciss/c0d0: unrecognized partition table type > No partitions found The problem is that your disk contains no (MSDOS) disk label. Most hard disks that you can buy are normally preformatted with an MSDOS disk label. You can create a disk label by running fdisk or cfdisk or parted once. fdisk /dev/cciss/c0d0 then type: o w q > Probing disks: /dev/cciss/c0d0 > Disks found: It's not good that there's no disk name after "Disk found". -- regards Thomas
FAI 3.1.5 released
FAI 3.1.5 was released which will also go into etch. It fixes some bugs which could be fixed with small patches. -- regards Thomas
booting FAI kernel with initrd
I like to answer this question from IRC > lazyb0y> did anybody ever create an initrd for fai that includes features to > get the location of the nfsroot by dhcp and mount the nfsroot > lazyb0y> as opposed to having the kernel features compiled in and do that on > the kernel level? > h01ger> yes,lots of peoole No, not lots of people. I think only oz, h0lger and I did this for FAI. It works this way: On a machine running a stock Debian kernel 2.6.18 (with etch) edit /etc/initramfs-tools/initramfs.conf and change to these lines MODULES=netboot BOOT=nfs then mkinitamfs -o fai-initrd. This creates the initrd for net booting. Copy the the kernel /boot/vmlinuz-2.6.18-3-686 and fai-initrd to /srv/fai/tftp In pxelinux.cfg/ use this file: label fai-generated kernel vmlinuz-2.6.18-3.686 append initrd=fai-initrd ip=dhcp FAI_CONFIG_SRC=nfs://kueppers/usr/local/share/cs-etch FAI_ACTION=install root=/dev/nfs nfsroot=134.95.9.149:/usr/lib/fai/nfsroot-etch,v2,tcp,rsize=32768,wsize=32768 FAI_FLAGS=verbose,sshd,createvt This boots my Thinkpad T60 without problems where the default FAI kernel does not work. Do not forget to change the initramfs.conf back to is previous version, otherwise the next kernel update will create an initrd which will not boot your machine any more (this happend to me once) or use the option -d. -- regards Thomas
FAI 3.1.4 released
FAI 3.1.4 which contains some small but important changes is released and will also go into etch. Since the package xorg changes its dependencies, the class XORG of the examples in FAI did not install some important packages containing xorg video output drivers. Additionally, setup_harddisks has now support for mounting existing NTFS partitions. You can get the new version by adding this line to your sources.list file: deb http://www.informatik.uni-koeln.de/fai/download etch koeln -- regards Thomas
Re: USB stick client for FAI install.
> On Wed, 3 Jan 2007 12:38:36 -0500, Ralph Crongeyer <[EMAIL PROTECTED]> > said: > I have several machines comming in that have no floppy and the motherboard > doesn't support PXE boot. Really no PXE? IMO most new boards with onboard NIC's have PXE. But anyway. Currently, we can't create a bootable USB stick for FAI but maybe you can put an image from rom-o-matic.net onto an USB stick. Another option is to put only the kernel (and may a boot loader) onto a USB stick which then mount the rest via NFS. I think it should also be possible to create a boot floppy image using make-fai-bootfloppy and put thisfloppy image via dd onto a USB stick. In the next few month I like to work on booting FAI (completelty) from USB stick. -- regards Thomas
FAI 3.1.3 released
FAI 3.1.3 is now available. It's only a bug fix release, which also fixes one security bug. The log files on the local disk are now only readable by root and the group adm, because they contain the root password hash. You will get the new version when using this line in sources.list: deb http://www.informatik.uni-koeln.de/fai/download etch koeln -- regards Thomas
Re: Solved / new Problem ;-) Was: Re: Sending DHCP requests .... timed out
> On Wed, 06 Dec 2006 18:13:52 +0100, Ingo Wichmann <[EMAIL PROTECTED]> > said: > I solved this by creating a diretory for the syslog pid-file in the line > before syslog is called: > mkdir -p /tmp/var/run > And i've put a symlink > /var/run -> /tmp/var/run > in nfsroot Which FAI version are you using? Newer FAI versions mount a tmpfs onto /var/run AFAIR, so this is not needed IMO. -- regards Thomas
Re: fcopy feature request: report if a file was copied
> On Wed, 6 Dec 2006 15:23:58 +0100, Janning Vygen <[EMAIL PROTECTED]> said: > if [ `fcopy -iM /etc/ntp.conf` -gt "0" ] > it would be nice to have an option like > -n --number report number of files changed > and let fcopy report the number of files changed Good idea. But then fcopy must not write any error or verbose messages when using -n. Or should -n report the number of files changes using the exit/return code? -- regards Thomas
Re: fcopy
> On Tue, 5 Dec 2006 15:46:35 +0100, Janning Vygen <[EMAIL PROTECTED]> said: > If i fcopy a file which is alread there with same content, but the file in > SOURCE has other file access permissions, why it is not copied? I should be copied, but the permissions should be changed. Which fcopy are you using? Maybe try fcopy -vD to get debuggin and verbose output. > Why does fcopy compares files only literally? Maybe a small bug. But first try -D for debugging output. If you join the IRC channel #fai on oftc.net and paste your output to paste.debian.net we may help you very quickly. -- regards Thomas
Re: Sending DHCP requests .... timed out
> On Sun, 03 Dec 2006 14:55:27 +0100, Ingo Wichmann <[EMAIL PROTECTED]> > said: > * the install-client boots and gets a pxelinux bootprompt > * kernel loads > Then the kernel prints out: > Sending DHCP requests . timed out! > But no packets are sent, i've checked this with tcpdump. > After a while the kernel tries again ( and again, and ... ): Are you using a Cisco switch? I had similar problems with Cisco switches when not enabling the fastport feature (this disables spanning tree). I guess it's the following problem. The kernel activates the interface. Then the switch needs some time (too long) for doing the spaning tree operation for a loop free topology for any bridged LAN. It seems that the switch need too much time for this. During this time the kernel sends several DHCP packets but they are not forwarded by the switch because the spanning tree in not yet finished. If the kernel send a numer of DHCP packages (for example) without success, it reopens the network devices which triggers another spanning tree detection on the switch. Solution: Try another switch, which is faster in detecting the loop free topology. Or have a look at the kernel source in net/ipv4/ipconfig.c, where the parameters for the number of retries and timeout parameter for reoping the network devices are definded. -- regards Thomas
Re: PRELOADRM help
> On Fri, 1 Dec 2006 10:54:05 -0500 (EST), "Brian E. Seppanen" <[EMAIL > PROTECTED]> said: > I guess this brings up another question. The package in question is my > own debian cfengine package for 2.1.20 rather than the buggy 2.1.14. > This may not be a strictly FAI question, but I would really appreciate > some pointers on how I can override a debian package with my own. This is called apt pinning. man apt_preferences apt_preferences(5) - Preference control file for APT -- regards Thomas
Re: PRELOADRM help
> On Thu, 30 Nov 2006 14:58:35 -0500 (EST), "Brian E. Seppanen" <[EMAIL > PROTECTED]> said: > I'm using FAI 2.8.4 with debian sarge. I have a customized package that > I think I should use PRELOADRM to download for installing. I don't PRELOADRM should ot be used for packages, it was meant in the past for thing like a flash tarball or something else. If you have a customized packages, create a local package repository and add this to the sources.list file. This is much cleaner than using PRELOADRM. -- regards Thomas
Re: trouble running fai-cd (fai 3.1.1 etch)
> On Thu, 30 Nov 2006 12:50:14 +0100 (MET), "Thomas Memenga" <[EMAIL > PROTECTED]> said: > i grabbed the fai cd (FAI 3.1.1, etch) and installed a faiserver with it. I have added some configs and now i want to create a new install-cd with fai-cd, but i get this error: > ERROR: mkinitrd-cd not found. Please install package. > But mkinitrd-cd is not available in etch. Is fai-cd broken in 3.1.1 for etch ? Yes. You have to copy your etch nfsroot, mirror (created by fai-mirror) and config space to a sarge system where you can use fai-cd to build an ISO image. I'm thinking of creating a sarge chroot tarball so people can use this chroot on etch to build the fai-cd. In the future (but after the etch release) fai-cd will use initramfs-tools instead of mkinitrd-cd. You can also have a look at holgers fai cd using live-package technique. -- regards Thomas
Re: class arithmetic?
> On Wed, 29 Nov 2006 09:08:21 +0100, Steffen Grunewald <[EMAIL PROTECTED]> > said: > is it possible to *remove* classes, or define new ones depending on > existing ones? No and Yes. No, you can't remove a class. Yes, you could define a class depending on another class. We do this in 50-host-classes: (ifclass I386 || ifclass AMD64) && echo GRUB But I always recommend to let classes be a simple ordered list. If you start a hierachy of classes with a sort of inheritance, you will lost the overview and will not understand which classes will influence which part of the intsallation. KISS: Keep it simple and stupid! -- regards Thomas
FAI 3.1.2 released
FAI version 3.1.2 is released. It now uses the 2.6.18 kernel by default and has some minor bug fixes. I've also build a new fai-cd ISO image. The packages and the ISO iamge are available at the FAI homepage http://www.informatik.uni-koeln.de/fai/ -- regards Thomas
Re: Extra partition being created
> On Mon, 27 Nov 2006 11:14:05 -0800, "Carl J. Van Arsdall" <[EMAIL > PROTECTED]> said: > disk_config disk1 > primary / 3000- rw,errors=remount-ro ; -c -j ext3 > logical swap 1000 rw > # logical /homepreserve9 rw,nosuid; -m 1 -j ext3 If you remove the -c flag, your installation will be much faster. This disables the badblock scan of mke2fs. -- regards Thomas
Re: setting up nfsroot on server whose ubuntu release differs from server's release doesn't work...
> On Thu, 23 Nov 2006 15:06:57 +0100, Holger Levsen <[EMAIL PROTECTED]> > said: > maybe you should look into cdebootstrap, it doesnt require a static list of > packages... Sure it needs a static list of packages. /usr/share/cdebootstrap/sid/packages Also debootstrap calculates parts of its package list dynamically, so both are using a static list and calculate other packages by using package dependencies. cdebootstrap does not support the --exclude flag :-( debootstrap => Architecture: all cdebootstrap is a compiled programm :-( which depends on several libdebian-installer.. packages. I do not like this. -- regards Thomas
Interview in German about FAI with Thomas Lange
I gave an interview (in German) for Radiotux.de about the FAI project at the LinuxWorld Expo in cologne. You can hear it here http://www.linuxpodcast.de/index.php?id=77 -- regards Thomas
Re: Fai server can be a different architecture from clients?
> On Wed, 22 Nov 2006 13:13:21 -0500, wtautz <[EMAIL PROTECTED]> said: > Ok. So inorder to create an nfsroot of arch type X one needs to be > on a machine of arch type X? Yep. > 1. I have an FAI server on arch X with various nfsroots of arch types > Y_1, Y_2 (these could be mounted via nfs from hosts of type Y_1, Y_2... > where these Nfsroot would have been created by installing fai-server > and running fai-setup). That's fine. But AFAIK your FAI server cannot export a nfsroot via NFS which was mounted by NFS from another server. So maybe you must copy those nfsroots to your server and then export them via NFS. > 2. I have clients that are of type Y_1, Y_2... and I want to use a single > FAI server to install them. It would be good if a menu would come up > on the client that would allow me to choose the correct FAI kernel image for > that arch. I think this should be feasible. In fact, we have an image > server that already has this ability. Anyone doing this? Attached is a pxelinux.cfg config file (maybe you want to name it default) which let you choose between different boot options. IT should be easy to extend it to support different nfsroots. default harddisk DISPLAY f1.msg F1 f1.msg F2 f2.msg F2 f3.msg PROMPT 1 TIMEOUT 900 label harddisk localboot 0 label memtest kernel memtest86+ label sysinfo kernel vmlinuz-install append ip=dhcp FAI_CONFIG_SRC=nfs://suenner/srv/fai/config FAI_ACTION=sysinfo root=/dev/nfs nfsroot=134.95.9.136:/srv/fai/nfsroot,v3,tcp,rsize=32768,wsize=32768 FAI_FLAGS=verbose,sshd,createvt label demo kernel vmlinuz-install append ip=dhcp FAI_CONFIG_SRC=nfs://suenner/srv/fai/config FAI_ACTION=install root=/dev/nfs nfsroot=134.95.9.136:/srv/fai/nfsroot,v3,tcp,rsize=32768,wsize=32768 FAI_FLAGS=verbose,sshd,createvt HOST=demohost label gnome kernel vmlinuz-install append ip=dhcp FAI_CONFIG_SRC=nfs://suenner/srv/fai/config FAI_ACTION=install root=/dev/nfs nfsroot=134.95.9.136:/srv/fai/nfsroot,v3,tcp,rsize=32768,wsize=32768 FAI_FLAGS=verbose,sshd,createvt HOST=gnomehost P.S: Please do not cite the whole posting, if only a small part is sufficient. -- regards Thomas
Re: Problem with setup_harddisks
> On Wed, 22 Nov 2006 11:16:20 +0100, [EMAIL PROTECTED] said: > /dev/hda: no such file or directory > [...] > setup_harddisks did not create /tmp/fai/disk_var.sh > In my nfsroot don't exist /dev/hda, but I don't know how to do.. Which version of FAI are you using? Are you installing sarge or etch? Search thin mailing list. There are some mails concerning the udev setup using different FAI versions. -- regards Thomas
Re: fai-cd grub menu...
> On Tue, 14 Nov 2006 14:23:19 -0800, "Russel Hill" <[EMAIL PROTECTED]> > said: > support. We can now flatline install on either IDE or SATA hard drives > as long as the operator has a keyboard & display connected to choose > the proper grub entry. It will also do unattended reboots on systems > with SATA drives. However, with IDE drives, it requires a > display/keyboard and operator intervention. Why do you need a keyboard when using an IDE drive? Maybe you are using an old FAI version. Newer FAI versions support a disk_config configuration using the disk name disk1 instead of hda or sda. Then the FAI installation will choose the correct device name. Maybe this solves your problems. > entries) IDE drives. But not with the same CD. Should be possible using a newer setup_harddisks version. -- regards Thomas
Re: RAID 5 with FAI
> On Wed, 08 Nov 2006 10:16:17 +0100, Jean-Paul BALOCHE <[EMAIL PROTECTED]> > said: > I would like to use RAID5 with FAI. > How can I do this with FAI ? Is there any documentation regarding raid5 > for fai v2.10.5 (11 july 2006) Have a look at the faiwiki and search for RAID. You need to write a hook or use the modified setup_harddisks from Sam Vilain. -- regards Thomas
FAI 3.1.1 released
The new version 3.1.1 of FAI was released. It only contains some bug fixes, no major user visible changes were made. You can get the Debian package by adding this line to your sources.list deb http://www.informatik.uni-koeln.de/fai/download etch koeln -- regards Thomas
Re: Building VServers with FAI
> On Wed, 1 Nov 2006 17:29:57 +0100, Holger Levsen <[EMAIL PROTECTED]> said: > where in etch did you fid this script, in which package? (i doubt its in > etch..) If you grep in Contents-i386.gz for etch you will find this: usr/lib/util-vserver/vserver-build.fai net/util-vserver So it's in the package util-vserver -- regards Thomas
Re: FAI 3.1 released
> On Mon, 30 Oct 2006 15:31:50 -0500, Kenneth Jacker <[EMAIL PROTECTED]> > said: tl> FAI 3.1 is now released. > I'm considering using 'fai' on a bunch of lab machines next semester. > I'll probably use Ubuntu/dapper which only has v2.10.1. > Can I use v3.1?Would v2.10.1 be "OK"? Ubuntu edgy includes FAI 3.1ubuntu1. But FAI 3.1 may also work for dapper. -- regards Thomas
Re: Using fai-chboot
> On Mon, 23 Oct 2006 11:31:03 -0700, "Carl J. Van Arsdall" <[EMAIL > PROTECTED]> said: > Is it possible to set the fai action as well as specify the nfs root on > the same command line? Yes. > fai-chboot -Iv vmlinuz-install /srv/fai/nfsroot_i386 node-30 Try: fai-chboot -v -k root=/dev/nfs vmlinuz-install /srv/fai/nfsroot_i386 node-30 > Is this isn't already done, has anyone considering changing fai-chboot > to take nfsroot and kernel using options? But the best thing is to create a template for it, for e.g myinst.tmpl. This can be copied using fai-chboot -vc myinst.tmpl hostname > If there are no objections, I'd be happy to work on a patch for this > kind of functionality, it could be useful (at least it seems to be in my > case). Thoughts? Ideas? Patches are always welcome, but maybe the template works for you. -- regards Thomas
Re: Is this a bug?
> On Thu, 19 Oct 2006 10:28:48 -0700, "Carl J. Van Arsdall" <[EMAIL > PROTECTED]> said: > Extracting templates from packages: 100% > /var/lib/dpkg/tmp.ci/preinst: line 63: [: -eq: unary operator expected > /var/lib/dpkg/tmp.ci/preinst: line 63: [: -eq: unary operator expected This is a bug of the postinst script of console-common. IMO you can just ignore it. Have a look at #392777. -- regards Thomas
FAI 3.1 released
FAI 3.1 is now released. It has some bug fixes and the major change that the action dirinst was renamed to dirinstall. The new option -u can be used for installations into a directory. It's available from deb http://www.informatik.uni-koeln.de/fai/download etch koeln A new fai-cd was also built and can be downloaded from http://www.informatik.uni-koeln.de/fai/fai-cd/ NEWS fai (3.1) unstable; urgency=medium - rename action dirinst to dirinstall - fai: add option -u,--hostname -- regards Thomas
Re: FAI MultiDistro
> On Thu, 19 Oct 2006 14:09:29 +1000, Edward Murphy <[EMAIL PROTECTED]> > said: > Primarily i would like to install Ubuntu Server, however I am also > interested in using fai to install Centos, or better yet RedHat AS/ES > Linux. Ok. First Ubuntu. Reinhard is preparing FAI 3.X version for ubuntu edgy. We hope that this version will be included into edgy. You can see the sources here: http://svn.debian.org/wsvn/fai/people/siretart/ For other distribution (rpm based) this is how it's done. Use FAI 3.1 (will be released in a few days) or newer. You will still use a Debian (or Ubuntu) nfsroot that is easily create using the make-fai-nfsroot command. The difficult part is to create the base.tgz file for the other distributions. Have a look at this url http://faiwiki.informatik.uni-koeln.de/index.php/FAI_multi-distribution It's a little bit outdated, because since FAI 3.0 there's a new subdirectory in the config space called basefiles/ where you should put your Centos or SLES or RHES base.tgz file. Call this file SLES10.tar.gz for e.g. and define a class for your install client called SLES10, the it will extract this base tar file and the the default Debian one. The tool for installing packages (called install_packages) also nows how to install pacakges via rpm, urpm, yum and other methods. Have a look at the code. And sure you have to adjust the package_config files, since packages may have different names. And check all scripts in scripts/... directories which are sometimes Debian specific. BTW, I happy about any feedback how this works and maybe you like to update the wiki pages concerning this topic. -- regards Thomas
Re: Installing kernel's for local boot
> On Mon, 09 Oct 2006 15:14:35 -0700, "Carl J. Van Arsdall" <[EMAIL > PROTECTED]> said: > I've been playing with getting FAI setup for a couple weeks now. I've Which version are you using? Things have changed with FAI 3.0. > custom kernels. The thing is, the howto is pretty dated. It comments > putting files in files/packages which was deprecated in 2.54 from what I FAI Guide version 2.5.4, 20 april 2006 for FAI package version 2.10.1 . . `files/packages/' THE USE OF THIS DIRECTORY IS NOW OBSOLETE. Maybe you are using an older version of the fai guide. But you are right we have to update the fai guide. > So it seems to me that kernel's are specified in two places: >1. /etc/fai/make-fai-nfsroot.conf >2. specify a kernel for PXE that lives /srv/tftp/fai/ (or whereever > your tftp stuff is housed I guess). yep. > So, my question, the boot kernel is always going to be vmlinuz install > or whatever you point to using fai-chboot right? In FAI 3.0, make-fai-nfsroot has now the option -V to specify the kernel name. See man make-fai-nfsroot for more info. You can also specify the kernel name with fai-chboot. > From the comments in > the conf file, the kernel from make-fai-nfsroot.conf is the kernel that > will boot via nfs. Correct. > I'm a bit confused as to how this works. This is the normal way how diskless clients boots. Have a look at nfsroot.txt in the kernel sources documentation. > Anyhow, following that stuff.. what's the best way to configure fai to > install a specific kernel for class? Build your own kernel using make-kpkg. This will make a Debian package out of this kernel. The create a local Debian package repository, add a line to the soures.lsit file and add the name of your kernel package to the class in package_config. -- regards Thomas
Re: fai-cd using wrong sources.list
> On Thu, 05 Oct 2006 13:45:08 -0700, Russel Hill <[EMAIL PROTECTED]> said: > We're in the process of stepping up to fai 3.0. I had things working a > month or so ago (pre 3.0) and I've since installed the 3.0 packages. I > can build an ISO using fai-cd but when it boots it attempts to install > using /etc/apt/sources.list, which doesn't work from the CD. The > sources.list that would work from the CD is found at > /etc/fai/apt/sources.list. > I haven't been able to find anything in the docs or on the mailing list. > /etc/apt/sources.list has the correct entries to construct nfsroot. I > suppose I could overwrite it between calling fai-mirror and fai-cd. This > would smell because fai-cd already does the work to construct the > correct sources.list. > Am I missing something simple and obvious? I just checked the fai-cd for 3.0. The correct sources.list is in etc/fai/apt suenner[/fai-cd3]# cat etc/fai/apt/sources.list # mirror location for fai CD, file generated by fai-cd deb file:/media/mirror testing main On the old fai-cd (pre 3) another location was used. IT was /etc/apt/sources.list which had the correct entry for a fai-cd. suenner[/old-fai-cd]# cat etc/apt/sources.list # mirror location for fai CD, file generated by fai-cd deb file:/media/mirror stable main Maybe the problem is that the location for the special sources.list (for fai-cd) has changed. -- regards Thomas
fai-cd 3.0 image for i386 available
The new fai-cd ISO image for etch containing FAI 3.0 is available at http://www.informatik.uni-koeln.de/fai/fai-cd/ An image for amd64 is not yet available. -- regards Thomas
Re: script failed but no error message
> On Thu, 28 Sep 2006 10:11:46 +0200, Holger Levsen <[EMAIL PROTECTED]> > said: >> + chroot /tmp/target hwclock --utc --systohc >> error=1 >> > Maybe introducing another class, REALHW, would be better than not executing > tihs when XENU or VSERVER is defined. To make it more generic and without the need of another new class I would do this chroot /tmp/target hwclock --utc --systohc || true It will always succeed and return no error code. -- regards Thomas
Re: FAI 3.0 released
> On Wed, 27 Sep 2006 18:33:20 +0200, Toni Mueller <[EMAIL PROTECTED]> said: > Maybe you have an idea about what might break, eg because you rely on > new Perl modules that only arrived after Sarge? Methinks whatever you > did to make it run on Etch should imho have given you an idea about > what was different on Sarge. No special Perl modules are used in FAI, so this will not be an problem when using FAI 3.0 with sarge, but there may be other issues. You can install FAI 3.0 on a sarge install server without any problems (I do this too). If you want to create the nfsroot, make sure you have debootstrap >= 0.3 installed, not version 0.2.45 from sarge. This is the only package you need for building and etch nfsroot. Creating a sarge nfsroot with FAI 3.0 might not work. I recommend FAI 2.10.5 if you like to install sarge on your install clients using a sarge nfsroot. I hope that now all sarge/etch compability questions are solved. -- regards Thomas
Re: FAI 3.0
> On Wed, 27 Sep 2006 16:25:27 +0200, Toni Mueller <[EMAIL PROTECTED]> said: > while I disagree, to I see it correctly that the new version of FAI is > being rolled again all into one package? No, FAI still has several packages like fai-server, fai-client,... -- regards Thomas
FAI 3.0 released
After - more than three months of very hard work - more than 380 commits - many days with more than 12 hours of work on FAI - more than 9000 lines of diffs but fewer hours of sleep - more than 20 bugs closed and still having back pain, I'm very glad to announce FAI 3.0. Currently my major wishlist bug is "A couple of days without FAI". I hope this bug will not be tagged "wontfix" but I fear it is tagged "etch-ignore" :-) This new major release changes a lot, even from the user's point of view. So please read the NEWS.Debian file if you upgrade from a former release. The FAI guide does not reflect the changes in FAI 3 but I hope this will be fixed before etch is released. One of the major and possibly incompatible changes are renamed variables. Now all important variables are described on this wiki page: http://faiwiki.informatik.uni-koeln.de/index.php/Variables FAI 3.0 is tested with etch. You can get it from any Debian mirror or by adding this entry to sources.list: deb http://www.informatik.uni-koeln.de/fai/download etch koeln These are the user visible changes: fai (3.0) unstable; urgency=low - WARNING: the fai-guide does not yet reflect the changes from FAI 2 to FAI 3 - rename NEWS to NEWS.Debian - /etc/fai/sources.list moved to /etc/fai/apt/sources.list - lots of new pathes: use /srv/tftp/fai instead of /boot/fai use /srv/tftp/fai/pxelinux.cfg instead of /boot/fai/pxelinux.cfg use /srv/fai/config instead of /usr/local/share/fai use /srv/fai/nfsroot instead of /usr/lib/fai/nfsroot use /var/lib/fai/config instead of /fai use /var/lib/fai/mirror instead of /mnt - changed file names variables.sh is now variables.log - new config file /etc/fai/NFSROOT which contains the package list for the nfsroot - changes in variables remove FAI_REMOTESH and FAI_REMOTECP from fai.conf, use FAI_LOGPROTO instead remove $installserver and $mirrorhost in fai.conf replace FAI_LOCATION by FAI_CONFIG_SRC, use default if undefined: "nfs://`hostname`$FAI_CONFIGDIR" LOGSERVER, LOGREMOTEDIR, LOGPASSWD removed from fai.conf, set them in the config space MNTPOINT now set to /media/mirror SERVERINTERFACE not needed by default in make-fai-nfsroot.conf NFSROOT moved to make-fai-nfsroot.conf FAICLIENTS, NFSROOT_PACKAGES removed from make-fai-nfsroot.conf $terror renamed to $task_error all flags from FAI_FLAGS are now defined as $flag_xxx $time_zone renamed to $TIMEZONE $videodrv renamed to $VIDEODRIVER $FAI_CONSOLEFONT renamed to $CONSOLEFONT $FAI_KEYMAP renamed to $KEYMAP $rootpw renamed to $ROOTPW $moduleslist renamed to $MODULESLIST new variable FAI_ALLOW_UNSIGNED - task_updatebase does not use updatebase.log any more - task_updatebase now logs to fai.log - all IP addresses of the install server are added to $NFSROOT/etc/hosts - automatic reboot if no errors are found (if error.log is empty) - new task dirinst for creating a chroot environment - try ftar $FAI/basefiles before extracting base.tgz - install_packages: new option -p (see man install_packages) - make-fai-nfsroot: add option -V (see man make-fai-nfsroot) - fai-chboot: change in options, -d is now -D, -r is now -d, -r is not used any more - add policy-rc.d support - move some -c to -C, so now all scripts use -C for cfdir, setting $FAI_ETC_DIR does the same - fai-cd can now eject the CD after installation - fai-mirror remove option -a (now default), add -B - default is kernel 2.6.17 - convert cfengine (version 1) scripts to cfengine2 - rename class XFREE to XORG - new script ainsl, implements AppendIfNoSuchLine from cfeninge in Perl - config space can be checked out from a subversion repository -- Thomas Lange <[EMAIL PROTECTED]> Wed, 27 Sep 2006 14:41:11 +0200 -- regards Thomas
Re: Testing changes in FAI
> On Mon, 25 Sep 2006 11:14:58 -0400, Paul Lussier <[EMAIL PROTECTED]> said: > What is the timeline for FAI 3, and is there some type of roadmap or > feature list for it available? I looked around on the wiki, but > didn't see anything there. FAI 3.0 will be release this week, if nothing serious happens. FAI 3.0 will not include an updated version of the FAI guide. This will be added later. Currently there's no roadmap or feature list available, but this is my own TODO list: http://faiwiki.informatik.uni-koeln.de/rails/todos/list You can read the NEWS directly from the svn tree here: http://svn.debian.org/wsvn/fai/trunk/debian/NEWS?op=file&rev=0&sc=0 If you like to see other new features for FAI 3 please discuss that on the linux-fai-devel mailing list. -- regards Thomas
Re: saving logs via ssh(scp)
> On Thu, 21 Sep 2006 17:20:02 +0200, Juraj Holtak <[EMAIL PROTECTED]> said: > Saving logs via ssh(scp) doesn`t work in beta. Please send a detailed bug report to the BTS (bugs.debian.org). I will not work on such a short bug report. If you need quick help for minor problems, come to #fai on irc.debian.org and do not use the mailing list for this. -- regards Thomas
Re: sendmail and FAI
> On Thu, 21 Sep 2006 17:21:24 +0200, Juraj Holtak <[EMAIL PROTECTED]> said: > is anybody using sendmail on FAI clients? Especially how do you deal with > sendmailconfig confirmations. Yep, I'm using FAI and sendmail on sarge systems withou any hassle. ./files/etc/mail ./files/etc/mail/Makefile ./files/etc/mail/Makefile/NULLCLIENT ./files/etc/mail/Makefile/MAILSERVER ./files/etc/mail/sendmail.mc ./files/etc/mail/sendmail.mc/NULLCLIENT ./files/etc/mail/sendmail.mc/MAILSERVER ./files/etc/mail/sendmail.mc/dreiser ./files/etc/mail/sendmail.mc/einser ./files/etc/mail/access ./files/etc/mail/access/kueppers ./files/etc/mail/local-host-names ./files/etc/mail/local-host-names/MAILSERVER ./files/etc/mail/local-host-names/kueppers ./files/etc/mail/databases ./files/etc/mail/databases/MAILSERVER ./scripts/CS_KOELN/31-sendmail faiserver[my_config_space]> cat ./scripts/CS_KOELN/31-sendmail #! /bin/sh fcopy -r /etc/mail $ROOTCMD sh -c "cd /etc/mail ; make" Hope this helps. -- regards Thomas
Re: unsigned repositories (PROBLEM FOUND!)
>>>>> On Thu, 21 Sep 2006 16:45:05 +0200, "Henning Sprang" <[EMAIL PROTECTED]> >>>>> said: > On 9/21/06, Thomas Lange <[EMAIL PROTECTED]> wrote: >> --- subroutines-linux (revision 3953) >> +++ subroutines-linux (working copy) > Should we do it like that/at this place? Shouldn't we just provide a > hook in the simple examples? I think we need a more general solution for enabling or disabling unsigned repositories. This was just a quick shot. -- regards Thomas
Re: unsigned repositories (PROBLEM FOUND!)
> On Thu, 21 Sep 2006 15:03:36 +0200, Juraj Holtak <[EMAIL PROTECTED]> said: > ls /etc/apt/apt.conf.d/ > 10fai 70debconf 90fai > so the 10fai and 90fai is MISSING. I hope you will test this patch and send me feedback. Index: subroutines-linux === --- subroutines-linux (revision 3953) +++ subroutines-linux (working copy) @@ -187,6 +187,7 @@ fi fi [ -d $FAI_ETC_DIR/apt ] && cp -r $FAI_ETC_DIR/apt/* $FAI_ROOT/etc/apt/ +[ -d $FAI_ETC_DIR/apt/apt.conf.d ] && cp -r $FAI_ETC_DIR/apt/apt.conf.d/*fai $FAI_ROOT/etc/apt/apt.conf.d -- regards Thomas
Re: $FAI_REMOTESH and $LOGSERVER in FAI~beta
> On Thu, 21 Sep 2006 15:04:40 +0200, "Henning Sprang" <[EMAIL PROTECTED]> > said: >> LOGSERVER has been moved to CS. > What does CS mean? Config space, so it's your /srv/fai/config tree. > Not sure, try if it works to define that variable in the config file > /etc/fai/fai.conf. If not it's probably worth a bugreport. I should be possible to define all variable in in fai.conf. But some people like to define it depending on a class. $SERVER is normally defined by a DHCP call. Have a look at boot.log. > That's why Thomas is calling for beta tests since weeks! I think it's > not yet too late to fix some things, but it might be a pointer for > others to test FAI 3 and comment on the impact of all the changes - > there should be some time, but not much, left to fix problems. Thanks. That's excatly my oppinion. -- regards Thomas
Re: $FAI_REMOTESH and $LOGSERVER in FAI~beta
> On Thu, 21 Sep 2006 14:02:01 +0200, Juraj Holtak <[EMAIL PROTECTED]> said: > have these been completly removed or just renamed? In FAI 3 variables have been renamed. Have a look at http://faiwiki.informatik.uni-koeln.de/index.php/Variables and http://svn.debian.org/wsvn/fai/trunk/debian/NEWS?op=file&rev=0&sc=0 -- regards Thomas
Re: unsigned repositories
> On Thu, 21 Sep 2006 12:26:08 +0200, Juraj Holtak <[EMAIL PROTECTED]> said: > as you can see from my previous email (fai-client), I have the same problem > with unsigned repositories even with your repositories for the developer pool > any idea what it could be? No. ATM I do not have the time to look into the unsigned/signed repository topic. I hope that someone else will care about this. -- regards Thomas
Re: problem with custom repository and aptitude
> On Wed, 20 Sep 2006 18:20:08 +0200, Juraj Holtak <[EMAIL PROTECTED]> said: > i have a custom rep with my own packages. it`s accesible via ftp. it`s not an > apt-secure repository, no Release.gpg file. I'm have no problems using an unsigned repository with FAI 3betaXXX. Please try the newest beta version and send the error message. -- regards Thomas
FAI at the city of munich
There's a short announcemnt about the use of Debian and FAI at the city of munich. http://times.debian.net/1020-Munich-Desktop -- regards Thomas
Re: Kernel Panic with custom kernel
> On Wed, 20 Sep 2006 12:02:27 +0300, Vicky <[EMAIL PROTECTED]> said: > Thanks for the help and I was able to start with the installation but > while the fai was trying to locate the HDD, it failed. I realised that > I was using it under the VMWare where the HDD is configured as scsi. I > compiled scsi drivers in the kernel. but again I am gettting the same. > Kernel panic. (this error is just the same as I was getting without > the CONFIG_ISO9660_FS). Please start with the default kernel configuration from fai-kernels. Then adjust this configuration step by step to your needs and check often if the kernel still can boot your vmware system. I bet your are missing some important driver or configuration in your kernel. -- regards Thomas
Re: Conflicting MTA Packages and fai-mirror
> On Wed, 20 Sep 2006 11:04:05 +0200, Rene Caspari <[EMAIL PROTECTED]> said: > I want to use two different MTA (Postfix and Exim) for severeal FAI > classes. If I use fai-mirror to build the mirror, it fails with a > conflict, that Postfix is not installable. > Maybe apt fails because of two concurrent MTA packages...? Yes, this is a "feature" of apt. Even if you use -d (for download-only) it complains if there are conflicts. Maybe apt should get a new option --ignore-conflicts. But here's the workaround. install_packages -m MAX or setting the variable MAXPACKAGES to 1 will install^wdownload only one package at a time, so there should be not conflicts. -- regards Thomas
Re: Testing changes in FAI
> On Tue, 19 Sep 2006 22:34:26 -0400, Paul Lussier <[EMAIL PROTECTED]> said: > Is there currently a way to answering the question "What would FAI > do?" for a given host *other* than actually FAI'ing that host? Maybe the new action in FAI 3 "dirinst" will help you most of all. There you can execute all parts of FAI (except for partitioning) and see what FAI is doing. -- regards Thomas
Re: Testing changes in FAI
> On Tue, 19 Sep 2006 22:34:26 -0400, Paul Lussier <[EMAIL PROTECTED]> said: > Is there currently a way to answering the question "What would FAI > do?" for a given host *other* than actually FAI'ing that host? Short answer: No and Yes. Long version: Not completely. You scripts in class/ for defining classes may also depend on the hardware found, so the list of classes does not only depend on the hostname. But if you are sure that the hostname is the most important information to define classes this can be done. > - which classes a given host is in we can set HOSTNAME and then call fai-class, which gives us all the classes. > - which disk config would be used you can call setup_harddisks with the option -X which does only a test, no real formating is done and you see what setup_harddisks would have done. > - what packages would be installed (and what dependencies would get >dragged in) Use install_packages -l for that. -l Print only the list of all packages that would be installed, then exit. If you liketo know all packags that will be installed because of dependencies, I have to enhance the script a little bit, but it's possible using the -s or --dry-run option of apt-get. > - which scripts would run and what the output/results of those >scripts would be fai-do-script has the option -n. But it will be impossible to get the output of those script without really executing them. cfagent also has a --dry-run flag, but IMO it's impossible for shell or perl scripts. > - what files would be installed (via fcopy) use -n. > Sort of a lint for FAI type of thing. IMO this is less than a lint thing, because you will not detect syntax errors in your scripts. I had the same idea as Michael who suggest a new action like test. IMO this would be the best way to implement such a functionality. -- regards Thomas
Re: Kernel Panic with custom kernel
> On Tue, 19 Sep 2006 13:52:35 +0300, Vicky <[EMAIL PROTECTED]> said: > I need the boot splash support into the kernel...and there are many > other drivers to be added/removed from the kernel alsobut the main > thing is the boot splash... Your config file at http://paste.debian.net/13255 maybe is not complete. The last line says: CONFIG_EXT3_FS_POSIX_ which can be the end of the config file. Your kernel config lacks the ISO 9660 driver which is needed for CD-ROM access. Maybe other drivers are also missing. I recommend to start with the kernel config from the original fai install kernel. You need this line CONFIG_ISO9660_FS=y -- regards Thomas
Re: Kernel Panic with custom kernel
> On Mon, 18 Sep 2006 19:42:47 +0300, Vicky <[EMAIL PROTECTED]> said: > I am using fai ver. 2.10.5 on debian testing. > with the basic installed fai I can make the cd and install the demo fai server. > I made a custom kernel which boots into vga mode (bootsplash) and is > working in my debian. but when I put the same kernel into the fai > nfsroot (dpkg -x package.deb), create the cd with fai-cd (along with a > proper local mirror), boot with the cd, it boots into vga mode but > doesnt go beyond Gibraltar. Why do you need a custom kernel for the nfsroot? Is the default install kernel not working? Which kernel version are you using? -- regards Thomas
Re: Screenshot of kernel panic.
> On Sat, 16 Sep 2006 22:57:59 -0300, Edmundo Valle Neto <[EMAIL > PROTECTED]> said: > FAI-CD puts all data needed for a fully automatic installation on a CD. > So not network is needed and you can install the whole system from CD. > The script creates an ISO image that includes the nfsroot, the > configuration space and a partitial Debian mirror. > The network configuration of my server was totally different from my > virtual network. (That was a 10.0.0.0/8 network). It always configures > the server with the IP 192.168.1.250 and a gateway 192.168.1.254 (what > is this?). Yes. This is an example of the network configuration. IMO it's not possible to guess your network configuration and to setup the install server from fai-cd using this configuration. Therefore I use a default network configuration. I you have a better solutiion for this, please email it to me. If you want to use the install server (which was installed from fai-cd) you have to manually adjust the network parameters. Then it should work. > I corrected the network configuration and installed the server and the > client without error messages from the repositories and from apt-proxy, > I got rid the custom names too, and tried to install atom25, but at the > end the result is the same as in the first e-mail: Kernel Panic When does the kernel panic occur? When booting the installation system (network boot, kernel from PXE with nfsroot) or after the client was installed, when it reboots the newly install system from local disk? > Edmundo Valle Neto escreveu: >> I tried some more times and during installation appears some errors Please do not cite all text again and agin. -- regards Thomas
Re: Screenshot of kernel panic - SOLVED.
> On Sun, 17 Sep 2006 12:24:15 -0300, Edmundo Valle Neto <[EMAIL > PROTECTED]> said: > So, the current fai-cd image doesnt work with a default configuration of > a vmware virtual machine, and probably doesnt work with LSI Logic SCSI I have no problems with FAI 2.10.5 or FAI 3beta and fai-kernels 1.10.1 and above and vmware 4.5. Which vmware version are you using? -- regards Thomas
Re: Trouble building NFSROOT (fai 3~beta21)
> On Fri, 15 Sep 2006 19:10:03 +0200, Toni Mueller <[EMAIL PROTECTED]> said: > Upgrading /usr/lib/fai/nfsroot > E: Couldn't find package fai-nfsroot > That needs /etc/fai/apt/sources.list with an entry to get the fai-nfsroot package. I think there's something missing. > What's that? I can easily download such a package using apt, but I That uses the default /etc/apt/sources.list. -- regards Thomas
Re: Build nfs root for Woodcrest
> On Thu, 14 Sep 2006 16:57:41 +0200, Steffen Grunewald <[EMAIL PROTECTED]> > said: > so a simple "ln -s ../sbin/cfagent /usr/bin/cfengine" will do, without > changing the scripts? Do not do ln -s. Install the cfengine2 package. We had pnly to do minor changes to the cfengine scripts. -- regards Thomas
Re: Build nfs root for Woodcrest
> On Thu, 14 Sep 2006 16:10:02 +0200, Steffen Grunewald <[EMAIL PROTECTED]> > said: > Setting up mdadm (2.5.2-7) ... > W: failed to load RAID subsystem. > Generating device nodes... done. > Generating mdadm.conf... done. > invoke-rc.d: initscript mdadm-raid, action "start" failed. > dpkg: error processing mdadm (--configure): > subprocess post-installation script returned error exit status 1 > Is there any quick fix? No. I just started to adjust FAI beta50 for etch. All versions before need manual change to work woth etch. > E: package cfengine has no installation candidate Just replace cfengine with cfengine2 > the two are really interchangeable - at a first glance there is no /usr/bin/cfengine With cfengine2 it's /usr/sbin/cfagent -- regards Thomas
Re: Build nfs root for Woodcrest
> On Wed, 13 Sep 2006 15:55:52 +0200, Steffen Grunewald <[EMAIL PROTECTED]> > said: > - a new kernel has to be used (for support of newer e1000 chips, and > Areca RAID controllers) fai-kernels 1.12 includes the 2.6.17 kernel. It that ok for you? > - it has to be a complete x86_64 installation > - etch for x86_64 is still incomplete I know. Since today I only tested the FAI packages (even FAI 3 beta20) on sarge machines. I will start with etch support tomorrow but it will take some days to find all quirks in etch. > What I need now are suggestions how to build a nfsroot that uses etch > (and possibly fill in missing parts from sarge?). Using make-fai-nfsroot -c you can specify a different dir for /etc/fai (e.g /etc/fai-etch) -- regards Thomas
Re: FAI on compaq alpha?
> On Thu, 7 Sep 2006 10:54:46 +0300, "T.Tugay BILGIN" <[EMAIL PROTECTED]> > said: > I would like to ask you if FAI works on Compaq Alpha AXP platform based clusters? Is this normal Alpha architecture? IMO it should be possible. You only have to write a hook for the partitioning, but that's easy. FAI is very generic except for the boot loader and partitioning part. Also have a look at this bug: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=330915 IMO This bug is wrong. It's not completly broken, you just have to write the hooks and compile an install kernel. > Please no HTML mails! -- regards Thomas
Re: package postgresql fail
> On Wed, 6 Sep 2006 15:58:00 +0200, [EMAIL PROTECTED] said: > All i see in log files is that the start of postgres get fail (ok it's > dosn't matter) This is not a bug, it's a feature of FAI (an other installers). FAI does not allow to start daemons inside the chroot during installation. Therefore we faked the start-stop-daemon program. IMO that's the reason why postgresql can't be set up properly. Now you have two option: Force postresql to start its daemon maybe by executing the original start-stop-daemon command inside the nfsroot (its: /sbin/start-stop-daemon.distrib). Write a rc2.d startup script that is run once during the first boot and do all the postgres initialsation there. Normall those script do rm -f $0 at the end, so it will only be executed once. /files/etc/rc2.d/S99fai-setup/FAISERVER is an example for this. -- regards Thomas
Re: setup_harddisks doesn't find any disk
> On Fri, 25 Aug 2006 16:48:38 +0200, "Thomas Wuppermann" <[EMAIL > PROTECTED]> said: > which was the solution of this problem? > FAI-Kernel? Which fai version? Which fai-kernel version? Which debootstrap version? -- regards Thomas
Re: Trouble With Brand New NFSROOT
> On Tue, 22 Aug 2006 16:17:58 +0200, Toni Mueller <[EMAIL PROTECTED]> said: > ipv6: version magic '2.6.16-fai-kernels SMP 386 gcc-4.1' should be '2.6.16-fai-kernels SMP 386 gcc-4.0' I have seen this on a Ubuntu system but have no solution. IMO modprobe or insmode was build with a different compiler version than the modules, and so they are complaining about this. -- regards Thomas
Re: sources.list in softupdate
> On Thu, 10 Aug 2006 11:45:34 +0200, Adrian von Bidder <[EMAIL PROTECTED]> > said: > Yes, I saw that now, but I don't like requiring two softupdates until it Our code is good when you are using different sources.list files during installation and after installation. Maybe you do not need this feature. > takes effect. My solution is to do it in prepareapt.FAIBASE. I guess I > should prepend it with "[ "$FAI_ACTION" == "softupdate"] &&" Yes, that's a good idea. But it's better to name the hook preparepat.DEFAULT and use the fcopy inside this hook. But the you can no have different sources.lsit file during and after the installation. > The other thing I don't like about this solution is that I now need to have > the sources.list in both /etc/fai and /srv/fai/config/files. Can I have > the copy in /etc/fai be a symlink, or will the scripts barf? IMO a symlink will not work, -- regards Thomas
Re: sources.list in softupdate
> On Thu, 10 Aug 2006 09:24:38 +0200, Adrian von Bidder <[EMAIL PROTECTED]> > said: > How do I (using the officially blessed method ;-) distribute a new > sources.list in softupdates? (right now, I'm adding a new repository, > later, I'll probably want to switch the systems from sarge to etch.) In the simple examples we do it with this line in a shell script using the fcopy command: scripts/LAST/50-misc: fcopy -i /etc/apt/sources.list -- regards Thomas
Re: initrd?
> On Tue, 01 Aug 2006 15:02:29 +0200, Patrick Cornelißen <[EMAIL > PROTECTED]> said: > Maybe we're going to equip our new workstations with the asus m2v board. > This has the disadvantage that some revisions need a driver that is only > available as a module. > I remember a statement from Froscon that initrd support for the > installkernel is not ready yet. What do we have to do to use an initrd? Contact holger levsen, and help him. He's working on install kernels with initrd. It's not that difficult for net booting with nfsroot, but before we can include it into FAI we need some solution for fai-cd to support it. -- regards Thomas
Re: request for comments on Getting_fai-logs_a_better_way
> On Mon, 31 Jul 2006 11:39:37 +0200, Ingo Wichmann <[EMAIL PROTECTED]> > said: > I've put up an acticle about how to get logfiles to the faiserver in the fai-wiki: > http://faiwiki.informatik.uni-koeln.de/index.php/Getting_fai-logs_a_better_way I've added some comments on it there. BTW. I should be very easy to use syslog instead of faimond. Using syslog for the short messages which are currently received by faimond makes sense, but IMO it's not useful to use syslog for all the long log files inside fai. If you want to use syslog instead of faimond, just redefine the subroutine sndmon() and luse the command logger for sending messages to a syslog server. -- regards Thomas
Re: bootloader task
> On Thu, 27 Jul 2006 11:34:53 +0200, Holger Levsen <[EMAIL PROTECTED]> > said: > But there should also be an option NOT to install/configure the bootloader... > some systems dont have a bootloader (think vserver etc), others have it vserver has no kernel on its own, so it's easy to detect that we do not need a boot loader. Maybe xen is more difficult. > manually installaed. Manually? No! Never! I think for FAI the kernel and the bootloader is just another software package that needs to be installed, configured or upgraded. I do not like to treat it different from other software package (if this is possible). So let's think of solutions that handle these package like all other packages. -- regards Thomas
Re: bootloader task
> On Wed, 26 Jul 2006 16:06:45 +0200, Ingo Wichmann <[EMAIL PROTECTED]> > said: > The bootloader is actually configured in task_configure, isn't it? Thats Yep, bootloaders are configured there. > not too nice for softupdates and non-debian distros. Shouldn't there be > a task_bootloader? Why? I think even for softupdates, you sometimes like to reconfigure your bootloader. It's only important that the bootloader configure code is idempotent :-) -- regards Thomas
Re: Kernel panic after Unpacking initramfs
> On Wed, 19 Jul 2006 11:08:11 -0700, "Russel Hill" <[EMAIL PROTECTED]> > said: > We already have a custom kernel for our installed systems. I was > hoping to turn on some config options and be able to use the same > kernel for our fai CDs. I turned in CRAMFS and BLK_DEV_INITRD. When I > boot in this configuration, I get the following: > Unpacking initramfs...<0>Kernel panic - not syncing: no cpio magic Building the FAI CD uses the command mkinitrd-cd. AFAIK this can't use kernels with initramfs. -- regards Thomas
FAI 2.10.5 released
FAI 2.10.5, the "do-not-trust-NMU-uploads" release has been uploaded. This fixes buggy patches from the two NMUs (2.10.3 and 2.10.4). It was really hard work to this at 30 degrees in my office. Not much fun work. 2.10.5 is the last release that is tested with sarge. From now on we're going towards FAI 3 and will break sarge compability. The next upload will use the new version number but maybe with a beta in the version name. Use this line in sources.list to install the new version: deb http://www.informatik.uni-koeln.de/fai/download sarge koeln -- regards Thomas
Re: Conflicting inetd packages
> On Fri, 30 Jun 2006 12:29:24 -0700, "Russel Hill" <[EMAIL PROTECTED]> > said: > make-fai-nfsroot appears to be installing both openbsd-inetd and > netkit-inetd. I'm not certain how that's possible since openbsd-inetd In my nfsroot I only have netkit-inetd. > We haven't been able to find any way to resolve this conflict. If you have only a problem with fai-mirror, try export MAXPACKAGES=1 fai-mirror ... This will download every package with a single apt-get/aptitude call -- regards Thomas