Re: FAI CD for etch using FAI 3.1.8 available
Hi, On Friday 20 April 2007 20:41, Holger Levsen wrote: If you download the iso from faicd.d.n today, its still broken, you need to boot with vmlinuz initrd=initrd.gz boot=casper - hopefully tomorrow the syslinux-config is fixed (its autobuild daily) and you can just boot the cd. (Which will erase your harddrive :) The iso from faicd.debian.net as of today works fine. (It installs a very basic system.) Is there also some way to build such a faicd from my own config space? Much like the old FAI-CD... Will this kind of faicd be the future of FAI-CD in general? I don't entirely rely on FAI-CD for our installs in general, but it would be a very nice way for bootstrapping our entire network, some I'm somewhat missing this feature :-) I haven't been engaged in the discussions about FAI-CD in general lately, so I might be missing something, sorry. Best, Michael pgpeqapSrddfX.pgp Description: PGP signature
Re: Chnages in grub menu.lst will be overwritten
[...] Is there any other script/class that works after all classes ? In /usr/local/share/fai I haven't found any entry. You can always try the class LAST, maybe by adding the fcopy command to scripts/LAST. HTH, Michael pgpOXGEEb3cYV.pgp Description: PGP signature
Re: sfdisk goes mad -- or -- my borked disk_config
Hi, I'm using 3.1.8ubuntu1 with feisty (No ubuntu change touches the config parts of fai) With the diskconfig file below (part1 is DellUtil vfat, part2 is Windows XP ntfs), fai computes way to high cylinder number (log below). [...] gives this in fai.log: Mapping disk name disk1 to hda Creating partition table: LC_ALL=C sfdisk -q /dev/hda /var/log/fai/current/partition.hda Warning: The partition table looks like it was made for C/H/S=*/255/63 (instead of 116280/16/63). For this listing I'll assume that geometry. sfdisk: I don't like these partitions - nothing changed. (If you really want this, use the --force option.) [...] As the new disk configuration computed by sfdisk seems to be fine, you might want to try the --force option, it did work for me in such cases - you've got backups, don't you? :-) No warranties, Michael pgpuXdaSBwCCA.pgp Description: PGP signature
Re: tftp-hpa
Hi all, There seems to be a problem installing tftp-hpa from the debian repositories at the moment. I can't even download it via aptitude download tftp-hpa? That should be sort of a local problem, http://packages.qa.debian.org/t/tftp-hpa.html says that it should be in all distributions. Best, Michael pgp78KDi7S3XT.pgp Description: PGP signature
Re: FAI won't install snort?
OK, I found the fai.log. Here's what I found in it about why snort didn't install: ---Snip--- Unpacking snort (from .../snort_2.3.3-11_i386.deb) ... mkdir: cannot create directory `/var/log/fai/current': No such file or directory dpkg: error processing /var/cache/apt/archives/snort_2.3.3-11_i386.deb (--unpack): subprocess pre-installation script returned error exit status 1 -End Snip-- I'm not sure what to do to fix this? During the install there is a /var/log/fai/current directory? Right? Do you have any ideas? This error seems like really weird, what is snort doing about FAI's log directory!? But anyway, you should probably check /var/lib/dpkg/info/snort.preinst to see what snort is trying to do there. HTH, Michael pgp5hFsCllo4z.pgp Description: PGP signature
Re: Strange error w/ fai-mirror
[...] To really fix it please file a bug with the Debian BTS to get this issue solved as soon as possible, it probably doesn't even take too much effort. For this bug I would report that fai-mirror fails when detecting two archs in a mirror? That seems like one bug (i.e. better error message). I also think I'm going to read into this a little more (when i have time) to see if I can figure out why fai-mirror is populating /srv/fai/mirror with more than one arch or would that type of investigation be generally discouraged? (I'm a bit new to open source and contributing to projects, but would like to help more if i could) Such an investigation is never discouraged, at least not within Debian, rather the converse is true: Whenever you find some bug or way to improve the software, this will be highly appreciated. It would be cool if you could file a bug report, such that this issue gets documented (publically), feel free to add the information that you are already investigating this issue. Thanks for your efforts, Michael pgpoJ0xxW4Ml7.pgp Description: PGP signature
Re: Strange error w/ fai-mirror
[...] /srv/fai/mirror/dists/etch/main/binary-i386/Packages /srv/fai/mirror/dists/etch/main/binary-amd64/Packages I use this machine to do fai installations for both 32bit and 64bit machines. Originally I set things up on original machines and copied them over. I was hoping to make things work on a single machine without having to set things up on two different machines. For a workaround i'll continue to do things on separate machines (a 32bit and a 64bit) and copy the results over to my server unless anyone has a better suggestion? Hmm, for the fix: It probably suffices to choose different mirror directories rather than different machines, but that's up to you. To really fix it please file a bug with the Debian BTS to get this issue solved as soon as possible, it probably doesn't even take too much effort. Best, Michael pgpR2NpVvjgcS.pgp Description: PGP signature
Re: issue.linuxlogo files...
Hi all, During the FAI install these two symlinks are created: /etc/issue.linuxlogo - /var/run/linuxlogo/issue.linuxlogo /etc/issue.linuxlogo.ascii - /var/run/linuxlogo/issue.linuxlogo.ascii and the files issue.linuxlogo and issue.linuxlogo.ascii are copied to /var/run/linuxlogo. What config files or scripts is this done with? I would like to remove/replace these with our own. I guess these changes are performed by the linuxlogo package, which is currently installed by the FAIBASE class of the simple example. This is just a wild guess, though. Best, Michael pgpzplEInWaAw.pgp Description: PGP signature
Re: Strange error w/ fai-mirror
I'm running 3.1.6. After running the command and some downloading the script produces an error... [apt doing stuff.] Calling apt-move /usr/bin/fai-mirror: line 319: $pfile: ambiguous redirect overlord:/srv/fai/config/package_config# Is this something I've done wrong or a problem with the script? Definitely a problem of the script as it should deal with such situations properly, but in essence it means that pfile=$(find $mirrordir/dists -name Packages) returns more than one match (or none?) Could you try this find command manually after the script has failed? $mirrordir is the directory you are using for creating a mirror. Thanks, Michael pgpJZMrDcVK8T.pgp Description: PGP signature
Re: booting FAI kernel with initrd
[...] 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 I doubt that you really need these, actually I did not change my initramfs.conf and it still works. 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: If you don't change initramfs.conf, you don't even need to run mkinitramfs ... - just copied over the kernel and the initrd that was already there. 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 For the sake of completeness, I'm using the following pxelinux.cfg/default: label fai-generated kernel vmlinuz-2.6.17-2-486 append ip=dhcp devfs=nomount FAI_ACTION=install root=/dev/nfs FAI_FLAGS=verbose,sshd,createvt,syslogd initrd=initrd.img-2.6.17-2-486 nfsroot=192.168.248.29:/srv/fai/nfsroot Actually one shouldn't need the additional nfsroot=..., if it is already provided by DHCP; however some versions of initramfs-tools seem to be buggy (and I don't know whether it got fixed already) and fail to parse the DHCP results. Best, Michael pgpCOu0suY1i6.pgp Description: PGP signature
Re: booting FAI kernel with initrd
[...] Actually one shouldn't need the additional nfsroot=..., if it is already provided by DHCP; however some versions of initramfs-tools seem to be buggy (and I don't know whether it got fixed already) and fail to parse the DHCP results. Right, I fail to get an installation started when not having nfsroot set on kernelk command line. So my orginal question, if/how this can be done with getting the nfsroot location via dhcp seems not yet solved In most situations it's enough to set it on the kernel command line, but it's not 100% compatible to the strategy of putting these features in the Kernel. I think the initramfs-tools bug related to these problems is http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=387808 - and it isn't fixed yet... I guess it's just some small bug in the scripts running in the busybox, but I haven't found the time to debug it. Best, Michael pgpoXoDZuZ67c.pgp Description: PGP signature
Re: [FAI] Install error
Michael Tautschnig [EMAIL PROTECTED] escribió: That's so unusual! The two anomalies I found are the main theme of theese mails, located in Calling task_instsoft step, and this, at the end of the installation: Could you please send all your log files? Please send them to me via private mail, not via the list. Best, Michael pgpyui2lZP3nQ.pgp Description: PGP signature
Re: [FAI] Install error
[...] (weird errors...) Did you check all the logs? I guess there must be some other anomalies than just these problems with apt - and your system being read-only is very very strange (and probably one of the causes of this problem). Best, Michael pgpJYFJQhf3Ak.pgp Description: PGP signature
Re: [FAI] Install error
Hi! In the installation of the client, if fails with errors of this kind: Could't stat source package list http://amd64.debian.net sarge/main Packages (/var/lib/apt/lists/amd64.debian.net_debian-amd64_dists_sarge_main_binary-amd64_Packages) - stat (2 No such file or directory) Could it be as simple as amd64.debian.net being irresponsive for a few seconds or some proxy at your site being overloaded? Or rather - is it reproducible? Thanks, Michael pgpYWlW9oM1B7.pgp Description: PGP signature
Re: DHCP and e1000
[...] So maybe it is a TFTP problem after all? I also tried the suggested tftp-hpa with the same outcome. Any help would be appreciated. When using tftpd-hpa, have you added the following options: OPTIONS=-l -vvv -r blksize in /etc/default/tftpd-hpa? Best, Michael pgpFyQ8NHDleY.pgp Description: PGP signature
Re: fcopy feature request: report if a file was copied
[...] yes but imagine fcopy -ir /etc/postfix/ and i need a postinst script with /etc/init.d/postfix reload i have to place postinst inside /etc/postfix/main.cf/BASE/postinst /etc/postfix/master.cf/BASE/postinst right? Nearly correct - there is no such directory as BASE, but rather you'll have the files /etc/postfix/main.cf/BASE and /etc/postfix/main.cf/postinst and if i did change master.conf and main.conf postinst would be called two times. and what if i have 30 files or more in /etc/postfix like virtual tables ... Would you really change all of them at once? And furthermore, in case of postfix there is usually no need for reloading it, it will at most be a call to postmap (which should definitely go into the postinst script). [...] Even though I do see (very few) applications for such a code, the same objections as for your -k,--keep-permissions apply here: It might add quit a bit of code, which may cause quite a lot of trouble. Don't get me wrong, I'd hope for a lot more of such ideas, but they must be discussed thoroughly, even more so when touching mission-critical tools such as fcopy. Best, Michael pgphlfbnvz0lR.pgp Description: PGP signature
Extending fcopy
Hi all, Janning has proposed quite a few enhancements to fcopy lately. I think the resulting discussion should be bundled in a single thread and it should actually move to linux-fai-devel (@Janning: don't know, whether you are subscribed to this one already). Just to summarize, the suggestions were: - some -k/--keep option (which I'd rather call --keep-permissions :-) ) to retain the permissions of the file which should be overwritten (BTW: what should happen if there is no such file?) - some way to return the number of files that were actually copied. My objections were - Adding code to a mission-critical tool like fcopy is dangerous. - More options may confuse users. - The intended behaviour may be simulated using pre-/postinst scripts. In my opinion the first and the second are still valid, however, the third has been proven to be _really_ clumsy. So I guess these options should be added, if a proper patch can be provided. Hoping for further discussion, Michael pgpldixqDwsaE.pgp Description: PGP signature
Re: fcopy feature request: report if a file was copied
hi, as i use softupdate a lot , it would be nice to have a fcopy feature which reports if any files or file permissions were changed. Look at my script in BASE/50-ntp [...] I guess postinst is what you really want (and it is already there, after all). See the man page. However, the permission thing is obviously not captured (yet). Best, Michael pgp5yRmGlAlET.pgp Description: PGP signature
Re: fcopy feature request: report if a file was copied
Am Mittwoch, 6. Dezember 2006 16:13 schrieb Michael Tautschnig: hi, as i use softupdate a lot , it would be nice to have a fcopy feature which reports if any files or file permissions were changed. Look at my script in BASE/50-ntp [...] I guess postinst is what you really want (and it is already there, after all). See the man page. you are right, but - postinst is called for every file which is copied - Which is what you intended, as I understood it - this does _not_ mean that postinst is called for every file that fcopy is run on, but only in those cases where the file is _really_ copied, i.e., overwritten. Best, Michael pgpMQXkhLyFLJ.pgp Description: PGP signature
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. Did you solve this issue already? Otherwise this is probably just what #360184 is about: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=360184 Best, Michael pgpahTlDDPFfu.pgp Description: PGP signature
Re: config directory doesn't mount and ssh problem
[...] Then, is there any error message you could see when it should have mounted it automatically? [...] Still no answer to this one - it's hard to debug a remote problem without an error message :-) Cheers, Michael pgpLWzCAFBMjT.pgp Description: PGP signature
Re: config directory doesn't mount and ssh problem
[...] I sent in message with attached log files trying again preplexed At least it hadn't reached me before; did you read through fai.log before sending it? It contains the answer :-) Copied from your first mail in this thread: FAI_CONFIG_SRC=nfs://192.168.1.1/srv/fai/config However, fai.log reads as follows (quoting it completely as it is fairly short) === - Fully Automatic Installation for Debian GNU/Linux FAI 2.10.1ubuntu1, 17 May 2006Copyright (c) 1999-2006 Thomas Lange [EMAIL PROTECTED] - Calling task_confdir Kernel parameters: ip=dhcp FAI_CONFIG_SRC=nfs://192.168.1.1/srv/fai/config FAI_ACTION=sysinfo root=/dev/nfs nfsroot=192.168.1.1:/srv/fai/nfsroot,v3,tcp,rsize=32768,wsize=32768 FAI_FLAGS=verbose,sshd,createvt BOOT_IMAGE=fai/vmlinuz-install Reading /var/log/fai/current/boot.log Warning /etc/init.d/rcS: Neither $FAI_LOCATION nor $FAI_CVSROOT are defined. WARNING: directory /var/lib/fai/config/class not found. Can't connect to monserver on port 4711. Monitoring disabled. Calling task_setup FAI_FLAGS: verbose sshd createvt Press ctrl-c to interrupt FAI and to get a shell Calling task_defclass Directory /var/lib/fai/config/class not found. Following subdirectories are found: find: warning: you have specified the -maxdepth option after a non-option argument -type, but options are not positional (-maxdepth affects tests specified before it as well as those specified after it). Please specify options before other arguments. /var/lib/fai/config cp: cannot stat `/var/log/fai/current/disk_var.sh': No such file or directory Aborting. === So, there are these warnings :-)) And: There is the version number: 2.10.1ubuntu1, which is strictly less than 3 :-) So - there is no such thing as FAI_CONFIG_SRC in this version; you need to upgrade your nfsroot first... Best, Michael pgpuIXi4C5K6C.pgp Description: PGP signature
Re: config directory doesn't mount and ssh problem
Hi, My pxe booted client does NOT mount the config directory /srv/fai/config. I can login to the client and manually do it successfully. The setup I'm using is syslinux. The dhcp broadcast takes about a minute to complete successfully...something about the broadcast going through the switch, I.e., I not directly on the local subnet but connected via some VLAN magic setup of port configuration. Perhaps this accounts for why the config directory isn't being mounted? I.e., timeout problems [...] Could you please provide some error messages or the like? What about mounting it manually, does that work? Best, Michael pgpv4qloYRPCA.pgp Description: PGP signature
Re: config directory doesn't mount and ssh problem
Michael Tautschnig wrote: [...] Could you please provide some error messages or the like? What about mounting it manually, does that work? Could of sworn that I wrote that mounting manually worked. Then, is there any error message you could see when it should have mounted it automatically? Cheers, Michael pgpHV33qNHxsS.pgp Description: PGP signature
Re: Maintaining Xen with FAI - Questions
Hello, I've got several Questions about using FAI to install a Xen virtual Domain. - First of all, what did you use to make custom kernels? I tried to use make-kpkg for generating xen dom0 and domU kernels, for the dom0 kernel it works fine, I am using the xen hypervisor provided by the debian repository, that works ok together with homemade Dom0 linux- kernel. Please correct me if I already have done something stupid at this time. I'm using a custom built xen kernel as well, because Debian kernels for xen lack pae support :-( - If I install the whole xen-system by apt-getting it, i can start a FAI Installation with the initrd provided by the deb package and the xenu kernel i generated by myself, of course i have to install my domu kernel over the xen system installation, also not very clean, but i dont know how to build a .deb-package out of the xen-utils or the hypervisor when i compile it from source. But if i try to use ONLY my domu kernel it doesnt start. XEND.log says domain has crashed. Building deb-pkgs from xen-source seems pretty advanced to me, because there are several make targets, dom0, domu and xen-utils with hypervisor,so how did you guys manage this? Why don't you use the packages provided by Debian? Here we've got: [EMAIL PROTECTED]:~# dpkg -l | grep xen ii libc6-xen 2.3.6.ds1-7 GNU C Library: Shared libraries [Xen version ii linux-xen0-2.6.18-686xenbigmem F.1 Linux xen kernel binary image for version 2. ii xen-hypervisor-3.0-unstable-1-i386-pae 3.0-unstable+hg11561-1 The Xen Hypervisor on i386-pae-class ii xen-utils-3.0-unstable-1 3.0-unstable+hg11561-1 XEN administrative tools ii xen-utils-common 3.0.3-0-1 XEN administrative tools - common files (System is entirely from Debian/etch) - Is there any essential support I have to setup in my domu kernel to get FAI to start running without an initrd?? I've checked there is already NFS support built into my domU-kernel. But what else do i need to be ready to use fai with xen? Network, disk, NFS should be about it... - Do you use seperate kernels for dom0 and domU? the debian pkg from official server installs one kernel for both. What are the advantages of using just one for both purposes? The same kernel everywhere... The advantage of using one for both should be obvious: It takes a lot more time to build and maintain two variants of your kernel. And I don't know about any disadvantages... Best, Michael pgp1bjl4E3xFt.pgp Description: PGP signature
Re: Problem with setup_harddisks
[...] Which chipset are you using, is it plain old P-ATA or S-ATA? I'm using S-ATA Ok, which chipset respectively which driver do you expect to be loaded? Best, Michael pgp60eU34B2m4.pgp Description: PGP signature
Re: Static IP Configurations
I'm using fai 2.8.4 with debian sarge. I've been working with fai for the past couple of days and have made it partially through the documentation. Is it possible to bootstrap a host to have a static interface configuration rather than a dhcp configuration? Any high level suggestions for how to go about that? Thus you probably can't use PXE I guess; so you'll need to create a fai-CD (which should work fine for sarge), there you'll find examples in the grub menu of how to do that. Best, Michael pgpCP6it0uCdq.pgp Description: PGP signature
Re: Extra partition being created
It would appear to me as though somehow an extra partition is being created during the disk setup. Here is my partition setup file, the partition.sda from the logs, is there a script somewhere adding an extra partition? #disk_config/FAIBASE # generic disk configuration for one small disk # disk size from 500Mb up to what you can buy today # # type mountpoint size in mb [mount options] [;extra options] disk_config disk1 primary / 3000- rw,errors=remount-ro ; -c -j ext3 logical swap 1000 rw # logical /homepreserve9 rw,nosuid; -m 1 -j ext3 #logs/partition.sda glue:/var/log/fai/node-27/last# cat partition.sda # partition table of device: /dev/sda unit: sectors /dev/sda1 : start=63, size= 33784632, Id= 83, bootable /dev/sda2 : start= 33784695, size= 2056320, Id= 5 /dev/sda3 : start= 0, size= 0, Id= 0 /dev/sda4 : start= 0, size= 0, Id= 0 /dev/sda5 : start= 33784758, size= 2056257, Id= 82 The partition on /dev/sda2 is extra and not something I defined. I just want one big partition and a 1 gig swap, is my configuration wrong? Furthermore, I looked at the partition table via fdisk, these partitions are assigned to the same starting point? /dev/sda2 is what you implictly requested - an extended partition (you said the swap partition should be a logical one, so setup_harddisks inferred an extend partition must be created). Best, Michael pgppeVYXbwita.pgp Description: PGP signature
Re: Extra partition being created
[...] Oh, I see, because I said it was a logical disk so an extended partition had to be created to house the logical disk. Its been a while since I've done some of this stuff, but is that more or less correct? It's completely correct. Best, Michael pgpMKiBVeSA6R.pgp Description: PGP signature
Re: Problem with setup_harddisks
Hello! I've got a problem booting a client, my installation dies with: /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.. I have udev in my FAI-Server, and the package udev in make-fai-nfsroot.conf. Is your disk detected? (check /tmp/fai/dmesg.log) Which harddisk controller driver is supposed to detect your disk? Is the appropriate module loaded? BTW - which kernel are you using? Thanks, Michael pgpvQ8YIjEXWx.pgp Description: PGP signature
Re: And a question about writing scripts
[...] Now to my problem. I use a couple classes for scripts, DEFAULT, FAIBASE, I386. In two of my directories not all scripts are being executed. In the default directory it executes 10, 11 and 20, but not 12. In my I386 directory it executes 22 but not 23. Below are the contents of these directories, can anyone tell me what I'm doing wrong? Is there more to this step than just a numbering scheme? [...] Have you checked that all of the scripts have proper permissions, i.e., are they readable and executable? Cheers, Michael pgpXg0KrKeCp6.pgp Description: PGP signature
Re: And a question about writing scripts
Michael Tautschnig wrote: [...] Now to my problem. I use a couple classes for scripts, DEFAULT, FAIBASE, I386. In two of my directories not all scripts are being executed. In the default directory it executes 10, 11 and 20, but not 12. In my I386 directory it executes 22 but not 23. Below are the contents of these directories, can anyone tell me what I'm doing wrong? Is there more to this step than just a numbering scheme? [...] Have you checked that all of the scripts have proper permissions, i.e., are they readable and executable? I sure have. Everything looks pretty good from the outside. Here's my listing: [...] (looks good) Could you provide an ls -lR scripts/ ? Just to have another quick look at all your scripts. Are you using NFS to access the config space? Is there a way to get a more verbose output from fai-do-scripts to help trouble shoot the error? I guess doing an export debug=1 somewhere, e.g., in class/DEFAULT.var should give you quite a lot of debug information. HTH, Michael PS.: If you prefer more of an instant messaging way of help just join #fai on irc.oftc.net pgpWp7ktx1JOz.pgp Description: PGP signature
Re: Need help understanding the partition/mountdisks tasks
I need to write some scripts that will modify the file system on the disk. I'm lookin in the manual for partition mountdisks and it says that my local disk should be mounted according to /tmp/fai/fstab relative to $FAI_ROOT, but I don't quite understand how things are being mounted. How is /tmp/fai/fstab generated? Is this the same as the fstab in my logs directory? If I had a single / partition, where would this be mounted to during the install? /tmp/fai/fstab is generated by setup_harddisks, which is called by the partition task; and yes, it is the same as backuped in the logs directory. Upon installation, everything is mounted below /tmp/target, so / would be /tmp/target, other partitions, e.g. /usr, would then be mounted as /tmp/target/usr HTH, Michael pgpCCWYWlk1Iz.pgp Description: PGP signature
Re: [Fai-commit] r4140 - trunk/bin
Author: lange Date: 2006-11-04 19:38:24 +0100 (Sat, 04 Nov 2006) New Revision: 4140 Modified: trunk/bin/fai Log: remove set +a for dirinstall, so $fai_rundate will be exported (closes: #390373) Do you really think this is the only way to fix the bug? I have not looked into these things that much, but why can't you simply export $fai_rundate explicitly? This modification might change the way things work in a very very subtile way... Best, Michael pgpSswI9i6Gjq.pgp Description: PGP signature
Re: [Fai-commit] r4140 - trunk/bin
On Sat, 4 Nov 2006 20:31:03 +0100, Michael Tautschnig [EMAIL PROTECTED] said: remove set +a for dirinstall, so $fai_rundate will be exported (closes: #390373) Do you really think this is the only way to fix the bug? I have not looked into these things that much, but why can't you simply export $fai_rundate explicitly? This modification might change the way things work in a very very subtile way... The but for others actions we also export all variables and do not use set +a. So, now dirinstall behaves the same as other actions. I'm not sure why I have added set +a for dirinstall in the past. I remember some discussions about not automatically exporting variables anymore on the dev weekend; maybe you intended to test that using dirinstall? Whatever - I think (not) exporting variables should either happen for all or for none of the actions. Best, Michael pgpnnZN0ItaU1.pgp Description: PGP signature
Re: kernel 2.4 and etch (was Re: Bug#367476: marked as done (make-fai-nfsroot doesn't create a proper dev/ directory))
Hi, Thomas Lange [EMAIL PROTECTED] wrote to [EMAIL PROTECTED] A 2.4 kernel is not supported by etch any more. That is wrong. Kernel 2.4 is (supposed to be) supported by etch, but not provided by etch. Then, how is 2.4 supposed to work debootstrap, which doesn't even create the required /dev-nodes? Thanks, Michael pgpzZlhwTHNKC.pgp Description: PGP signature
Re: Installing kernel's for local boot
[...] 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. Alright, so I'd make a local repository on my fai server, place that debian package inside of it, and add it to package_config. Does this handle updating menu.lst and the modules for me? (sorry if that's a noob question, previously I just wrote scripts to manipulate these files for me) I'm not absolutely sure on this one, because I don't manage my menu.lst this way, but usually either there is no need to modify menu.lst (as all kernels make /boot/vmlinuz a proper symlink) or menu.lst is updated automatically, if it contains appropriate placeholders as in the version shipped by Debian. About the modules: Did you mean /etc/modules? If it does need updating (which is quite rare in times of udev), you can do it setting MODULESLIST in one of the .var files in the class/ directory. This variable is then handled by scripts/FAIBASE/10-misc, if you haven't changed the example configs too much. HTH, Michael pgpFndvBbvHeE.pgp Description: PGP signature
Re: Installing kernel's for local boot
[...] Anyhow, following that stuff.. what's the best way to configure fai to install a specific kernel for class? Since the howto talks about seriously deprecated things, I'm assuming a better was has been put in place. If there's no fancy way, does anyone do a brute force approach where the kernel is copied by a script to /boot along with all the other necessary things (menu.lst, modules, etc).? Do I get that right that you are asking on how to install a kernel on client? I'm just a bit confused because all of the text removed above was about the kernel that the client will boot during installation... Anyway, if my guess is correct: Are you planning to install a distro-kernel like linux-image-2.6 or do you want to install your custom built kernel? If the latter is true, have you built a Debian-package using make-kpkg or do you intend to install the plain binaries (including modules)? As all variants mentioned above require a different approach, please answer these questions before we can help you further on. Regards, Michael PS.: Sorry for the very outdated manual, we are trying to fix that as soon as possible... pgp4yPhyxdVaL.pgp Description: PGP signature
Re: FAI 3.0 released
[...] Looking at the massive changes: Is there any hope of updating clients from fai 2.10.5 to fai 3.0 with a fai softupdate? I imagine the Debian side should mostly work (as described in another thread, I'm using X, libc, KDE, udev/dbus/hal, oo.org etc. from etch already, so there is not that much left), but I'm not sure about how fai will react. [...] As posted on linux-fai-devel 2 days ago: just do it :-) See, e.g., http://www.uni-koeln.de/bin2/maillist/linux-fai-devel/20061004.002451.2/173053 HTH, Michael pgputlClOGJiP.pgp Description: PGP signature
Re: FAI3 / mdadm
Hi *, I'm trying to set up an FAI3 etch installation on a sarge machine (both amd64). I have a strangeness concerning the installation of the mdadm package (I'm installing on a raid1 which is working fine under fai2.10.5/sarge): Building the nfsroot (using debootstrap 0.3.3) everything works out fine, but during the installation of a new machine the mdadm package won't finish installing with the error: --- Setting up mdadm (2.5.3.git200608202239-6) ... dpkg: error processing mdadm (--configure): subprocess post-installation script returned error exit status 1 --- I'm wondering if this error is etch or FAI3 related? To me it seems strange that the nfsroot is building fine, but the installation goes wrong. I've seen these errors here as well, but didn't find the time to debug them (and didn't care about mdadm at all). Could you please try to do this? Would be very nice to know the reason of this problem. Some hints: Try chroot $NFSROOT apt-get -f install Does it then try to reinstall mdadm? Does it ask any questions? If that doesn't work, please insert a set -vx in the second line of $NFSROOT/var/lib/dpkg/info/mdadm.postinst and retry the apt-get -f install Thanks in advance, Michael pgpJvWmzKW4gS.pgp Description: PGP signature
Re: disk_var.sh copied during softupdates?
[...] (will reply to that part later on) Even though this shouldn't cause too much trouble, it does result in an error message on systems that were installed using fai 3.0 because AFAIK the path was changed. Any ideas how to fix that? Are there other problems when using FAI 3.0 for softupdate when the install client was installed with 2.X? IMO there must be problems when using variables in customization scripts, which had been renamed. Actually things were less troublesome than expected, but I had prepared my configspace already; however, the only things I did were: - update fai.conf - s/DO_INIT_TASKS/do_init_tasks/ - s/target/FAI_ROOT/ - explicitly export all the variables - added class/00transition.source: if [ -z $do_init_tasks ] ; then export do_init_tasks=$DO_INIT_TASKS fi I don't know, whether all these steps were really required, but this way it seems to work without any problems. Even fai-client itself was upgraded via a softupdate :-) Cheers, Michael pgp3ZVn3RFREg.pgp Description: PGP signature
Re: sendmail and FAI
Hi, is anybody using sendmail on FAI clients? Especially how do you deal with sendmailconfig confirmations. Is sendmail using debconf or what kind of confirmations are these? I think I've never used sendmail on Debian ... Thanks, Michael pgpJWC5FoZOyU.pgp Description: PGP signature
Re: Testing changes in FAI
[...] Is there currently a way to answering the question What would FAI do? for a given host *other* than actually FAI'ing that host? Not that I would know of. However, many of us here would like to see that in FAI :-))) I'm thinking of some kind of script which would recurse through the config hierarchy and evaluate files and scripts to determine things like: - which classes a given host is in I think you will need to run this test on the acutal machine because the classes usually depend on the software/hardware configuration of the machine being installed. However, these scripts usually should not touch anything, just read-only access. - which scripts would run and what the output/results of those scripts would be Which scripts are you talking about? Those in class/ or those in scripts/? - which disk config would be used Ok, that's quite easy once you know which classes are there. - what files would be installed (via fcopy) fcopy now has an option --dry-run (probably available in 2.10.x), it might be helpful (alternatively the environment variable FCOPY_DRYRUN may be set). - what packages would be installed (and what dependencies would get dragged in) This will probably demand a patch to install_packages, but it seems feasible. In theory, this script could be given a hostname of some machine and run this test as if it were that host doing the actual FAI install. Sort of a lint for FAI type of thing. Obviously, since that host would not actually be installing, this script would not be able to actually do certain things like partition disks, fcopy files, and install packages. But it could certain provide a significant amount testing to any given configuration change and inform you whether or not the change would result in the desired outcome when you do actually go and install the host. The FAI softupdate feature seems like it might be a good place to start, but appears to have to actually run on the host. I don't think fai softupdate is the way to go, because that will just mean that you skip some tasks, such as partitioning. I'd rather go for a new method (like dirinst, which has recently been added), something like fai test That will definitely require some work on FAI, but it would still seem very very helpful, especially for the developers. To this end I suggest adding something similar to --dry-run/FCOPY_DRYRUN to all relevant scripts; then simply exporting FCOPY_DRYRUN (and maybe INSTALLPACKAGES_DRYRUN and the like) from /usr/sbin/fai would already do quite a bit of the job. Regards, Michael pgpNzS7c6BwWN.pgp Description: PGP signature
Re: Testing changes in FAI
On Wed, Sep 20, 2006 at 09:46:35AM +0200, Thomas Lange wrote: 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. For us it would be much more important to simulate softupdates than installs. So one new action won't do it. Rather a -n option to the fai script which makes the following action into a dry-run. This is also what one would expect :) Well, the difference as far as simulation is concerned is really a minor issue; actually, running the fai test action would cater for both; actually, it would be more like a softupdate anyway. This is due to the following fact: On whatever system you run it, it would, e.g., have apt grab the currently installed packages and apt would only print the packages newly installed. The same holds for fcopy. It's IMHO just an additional output of the setup_harddisks utility, which you can ignore anyway. Testing an installation that way is probably even harder, as we might need the base.tgz to be extracted to perform any kind of apt simulation at all!? Regards, Michael pgpqeHMuvBTWp.pgp Description: PGP signature
Re: Kernel Panic with custom kernel
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... Why do you need the boot splash during an installation for? And which drivers do you need to have added? You might want to file a wishlist-bug for these... Thanks, Michael pgpqqELCMTb9X.pgp Description: PGP signature
Re: Build nfs root for Woodcrest
On Wed, Sep 13, 2006 at 04:30:22PM +0200, Thomas Lange wrote: 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? Since I need the arcmsr module, almost every kernel will do if if can be patched. If the 2.6.17 source is available, I'm fine... - 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. [...] Is so incomplete that you can't use it? In case you are using it for some server applications only, it should probably be alright. What one could do: - Boot your x86_64 with some sort of live-CD or anything else that gets you into a native x86_64 environment - install the necessary FAI packages there and run make-fai-nfsroot - find some way of installing your tailored kernel * make-fai-nfsroot -k * have the initial make-fai-nfsroot install it * cp $kernel $NFSROOT ; chroot $NFSROOT ; dpkg -i $kernel * most probably even dpkg -x $kernel $NFSROOT will do - copy your $NFSROOT to your FAI server If the first 2 steps aren't possible at all (probably due to your RAID controller), grab someone elses NFSROOT for x86_64 (e.g., I do have one), it might be sufficient to get your own x86_64 running to build an NFSROOT yourself. HTH, Michael signature.asc Description: Digital signature
Re: Re: package postgresql fail
All i see in log files is that the start of postgres get fail (ok it's dosn't matter) Somewhere in software.log it should say something like Setting up postgresql-7.4 ... Does that show up in your logs? If it does, is it followed by any errors? Please post the relevant parts of your logfile here. As far as I could see from the scripts used to setup the postgresql-7.4 Debian package, the initdb command should be run from the postinst, which is what happens after the Setting up ... line. HTH, Michael signature.asc Description: Digital signature
Re: your mail
Hi! Just a few notes: This list prefers English, even though many of the subscribers do indeed speak german. AND PLEASE find a proper subject - mails without a subject tend to be eaten by SPAM filters. Hallo ich hab ein Problem mit dem Erstellen von 'postgresql' als Paket #cat /package_config/POSTGRESQL PACKAGE aptitude-r postgresql postgresql-client #cat /debconf/POSTGRESQL postgresql postgresql/initdb/location string /var/lib/postgres/data postgresql postgresql/upgrade/preserve_locationstring $PGDATA/.. postgresql postgresql/settings/day_month_order select European postgresql postgresql/upgrade/policy boolean true postgresql postgresql/settings/locale select C postgresql postgresql/enable_lang boolean true postgresql postgresql/purge_data_too boolean false postgresql postgresql/very_old_version_warning boolean true postgresql postgresql/upgrade/dump_locationstring $PGDATA/.. # Should your missing postgresql.env configuration file be restored? postgresql-client postgresql-client/missing-postgresql.env boolean true postgresql postgresql/convert-pg_hba.conf boolean true # Postgres ist installiert aber nicht fertig konfiguriert (initdb fehlt ) der start geht auch fail da 'invoke-rc' kein fake Skript ist About which kind of start are you talking here - during the installation or the start after the reboot? And in which sense is it failing to start? wenn ich nach der Fai Installation reboot #debconf-set-selection /debconf/POSTGRESQL aufrufe und dann #aptitude postgresql ist die Installation ohne Interaktion mit 'initdb' ausgeführt worden. hat jemand eine Ahnung wie es sonst geht könnte ? I don't know anything about postgresql, so my question might be somewhat mistaken, but what is initdb about - is it the initialisation of the storage? Furthermore, you might want to find out where initdb is initialised (normally) and have a look at the script, it might tell you a lot more. Regards, Michael signature.asc Description: Digital signature
Re: setup_harddisks doesn't find any disk
Well, Well, to start, some good news : I have configured a new fai server. Now everything is ok for ata disk : I can install a demohost with ata, but (bad news !) not with SATA. The modules are now ok and they can be loaded. When I try an installation with SATA, I've got those error messages : [...] Your kernel log makes me think this is for sure not a bug in FAI - your kernel is having some trouble (with the controller); you might want to try some newer kernel version. Regards, Michael signature.asc Description: Digital signature
Re: setup_harddisks doesn't find any disk
Hi, thanks for your answer. I started my client with the boot cd-rom and then checked that they are files in /dev, and yes, I can see all the devices (regarding hard drive I've got hdx and sdx) Kernel version : 2.6.16-fai-kernels #1 smp My hardware : I'm using an amd 64 3000+ with a gigabyte motherboard : [...] IDEamd74xx 10de0054YesnVidia CorporationCK804 Serial ATA Controllersata_nv 10de0055YesnVidia CorporationCK804 Serial ATA Controllersata_nv 10de005c-nVidia CorporationCK804 PCI Bridge [...] What I don't understand is why it is the same thing for an ide and sata file ? It is like the modules are not loading ? I don't know whether the modules aren't loading or the controller isn't supported/the modules aren't provided by that kernel - Holger, could you help? IDE vs. SATA probably won't make any difference as it is the controller (which is probably the same for both), which isn't supported. HTH, Michael signature.asc Description: Digital signature
Re: how to pin a locally compiled package?
Sorry, I've got myself all confused. I've got some samba packages which I've recompiled. samba-common-3.0.14a-3sarge1SBC Now that samba-common-3.0.14a-3sarge2 is out, this is overriding my own package. Which is fine. I understand. I just need to pin the packages with a priority of 1001 and apt won't upgrade them. But it doesn't work. I'm fcopy-ing /etc/apt/preferences, and post-install an aptitude upgrade downgrades to my local package. Which is nice. But I want fai to prefer my local package. I suspect I need to modify /etc/apt/preferences on the fai server - is this right, and do I need to do a make-fai-nfsroot after? I'm sure it's in the docs or on the mailing list, but sometimes, you know, you just can't SEE it? If I got your situation right, all you need to do is fcopy-ing /etc/apt/preferences _before_ the actual package installation begins. You might want to use a hook like updatebase.DEFAULT to do so. HTH, Michael signature.asc Description: Digital signature
Re: fai bugs against policy - lets get this fixed NOW...
Hi, On Monday 10 July 2006 13:12, Henning Sprang wrote: Things for which the rules in Debian are set for etch should really be fixed as soon as possible - if there's a final decision in Debian for which FHS version to go, then the bug should be handled. As written in my mail which started this thread, Debian has changed the policy to mandate FHS 2.3. Thats the latest FHS version, so it's final. Nice to know, however, ... Also, I dont see how changing the default setting of a configuration variable effects any existing installation, you need to modify fai.conf to use it, and modified config files are not touched on upgrades. even though I agree that this should not cause any harm, there might be some unexpected side effects. Thus I urge for the modified package being tested before it gets uploaded to unstable, and I'd suggest we do the testing at Linuxhotel. Best, Michael signature.asc Description: Digital signature
Re: forcedeth
Trying to install some new AMD64 computers with NForce chipset, I noticed that the forcedeth driver is missing in the 2.6.16 installation kernel. Of course, I solved the problem by compiling my own kernel with CONFIG_FORCEDETH=y. Since most newer (939/AM2) AMD motherboards use nforce, I suggest that forcedeth is added to the standard kernel. Please file a bug with severity wishlist. Thanks in advance, Michael signature.asc Description: Digital signature
Re: subnet depending fai installation
Hi there, Has anyone changed the class '50-host-classes' that FAI is able to recognized from which subnet the client comes from?!? We want to have different disk_config files for different subnets. Can anyone help us? Which subnets do you intend to distinguish? Could you give some examples? I'd start out by parsing `hostname -i`. Thanks, Michael signature.asc Description: Digital signature
Re: GRUB, FAI, and RAID1
Michael Tautschnig [EMAIL PROTECTED] writes: Yes. So - what did the grub installation tell you there? Nothing, that's the problem. Is it at least telling you that the command is run? That's what I was thinking. However, I can't seem to figure out how grub *should* be installed properly from within FAI wrt a RAID1 set. To what extend is installation from within FAI different from a normal installation of grub? Well, this is also in combination with a RAID1 set across 4 disks. So, I guess it differs in that the root (/) partition is not on any one disk, but an md device. So, does grub get installed to the MBR, or to the root partition on md0? Also, should I be using grub-install (which I've never had much luck with even outside of FAI), update-grub, or echo EOF ... | grub ? Currently, my install seems to be hanging on the grub install: :/tmp/fai# ps auxww | grep grub root 14180 0.0 0.0 2580 1280 tty1 S15:44 0:00 /bin/sh /sbin/grub-install --no-floppy --root-directory=/tmp/target /dev/sda root 14282 0.0 0.0 6004 548 tty1 D15:46 0:00 //sbin/grub --batch --no-floppy --device-map=/tmp/target/boot/grub/device.map root 14297 0.0 0.0 1544 500 pts/0S15:48 0:00 grep grub I can almost guarantee that the reason that second line is in a D state is that it's waiting on a floppy drive which doesn't exist, this despite the presence of --no-floppy option. Is this command run by FAI or are you doing it manually? However, here (http://www.linuxsa.org.au/mailing-list/2003-07/1270.html) it states that you need to install grub on all of your disks. In this case you will need to adapt the scripts in the GRUB/ directory to fit your needs. HTH, Michael signature.asc Description: Digital signature
Re: GRUB, FAI, and RAID1
Thomas Lange [EMAIL PROTECTED] writes: On Thu, 22 Jun 2006 12:25:42 -0400, Paul Lussier [EMAIL PROTECTED] said: I'm trying to install to a RAID1 set using FAI. Everything goes extremely well Are you sure? Have a look at all log files, especially shell.log. Yes. So - what did the grub installation tell you there? GRUB GRUB _ Maybe the MBR was not written correctly. Did you define the class NOMBR? It should not be defined. No, we have no class called NOMBR, nor does anything define it for any of our systems. There were no stage files at all. So the installatiof of grub must have an error. That's what I was thinking. However, I can't seem to figure out how grub *should* be installed properly from within FAI wrt a RAID1 set. To what extend is installation from within FAI different from a normal installation of grub? Regards, Michael signature.asc Description: Digital signature
Re: sources.list per host
Hi, I there a way to have a sources.list during the installation (and when the system is installed) for some specific hosts. E.g. we have rolled our own multipath packages and I don't want to include that repository on each machine. I could use a script that appends that repository and then do an update; and install using that repository. But then I can't use the package_config/CLASS for adding software. An ideal way would be to be to say if class X is defined use sources.list.X instead of sources.list. Would it suffice if you fcopy'd sources.list at the right time? We're doing that to some extend, using the hook updatebase.DEFAULT: #!/bin/bash if ! echo $classes | grep -q FAICD ; then fcopy -Bi /etc/apt/sources.list fi fcopy -Bi /etc/kernel-img.conf fcopy -Bi /etc/apt/preferences HTH, Michael signature.asc Description: Digital signature
Re: sources.list per host
Michael Tautschnig wrote: Would it suffice if you fcopy'd sources.list at the right time? We're doing that to some extend, using the hook updatebase.DEFAULT: Ah yes the hooks! I tried putting a cfengine script as a hook but fai.log says: Calling hook: updatebase.DEFAULT oeral: (CFINPUTS is set to nothing) oeral: Can't open file cfengine.conf updatebase.DEFAULT FAILED with exit code 1. I think I should set some variables, or is it not possible to use cfengine here? I've got no idea about cfengine, I'd use sed/awk/perl instead. Sorry... Michael signature.asc Description: Digital signature
Re: xen and fai
Michael Tautschnig wrote: Michael Tautschnig wrote: So - what? What kind of thing is mapper/itext - a partition, the whole SAN? mapper/itext is a LUN that is exported by our SAN. Thanks; do you have any pointers since when Xen should support partitioning from within the dom-U? AFAIK that simply isn't supported... We are using a recent xen Package: xen-hypervisor-3.0-amd64 Source: xen-3.0 I can manually use fdisk in my virtual domain to partition the disk. Cool - nice to know! This the error I get in my fai.log: sfdisk: ERROR: sector 0 does not have an msdos signature /dev/sda: unrecognized partition table type No partitions found Hmm - I wonder why this causes the partitioning to fail!? A quote from one of my installations of a normal system, actually a dom-0: Partitioning local harddisks setup_harddisks version 0.35fai Probing disks: /dev/sda Disks found: sda sfdisk: ERROR: sector 0 does not have an msdos signature /dev/sda: unrecognized partition table type No partitions found Using config file: /fai/disk_config/XEN0 Creating partition table: LC_ALL=C sfdisk -q /dev/sda /tmp/fai/partition.sda sfdisk: ERROR: sector 0 does not have an msdos signature /dev/sda: unrecognized partition table type No partitions found Creating file systems: mke2fs -q -j /dev/sda1 mkswap /dev/sda2 mke2fs -q -j /dev/sda3 mke2fs -q -j /dev/sda5 mke2fs -q -j /dev/sda6 mke2fs -q -j /dev/sda7 mke2fs -q -j /dev/sda8 mke2fs -q -j /dev/sda9 mke2fs -q -j /dev/sda10 mke2fs -q -j /dev/sda11 Write fstab to /tmp/fai/fstab Write FAI variables to file /tmp/fai/disk_var.sh Calling task_mountdisks So it is telling me the same, but nevertheless continues to work; actually it is to be expected that there is not partition table on a new set of harddisks... HTH, Michael signature.asc Description: Digital signature
Re: apt.conf.d/*
On Wed, 14 Jun 2006 11:32:50 +0200, Rudy Gevaert [EMAIL PROTECTED] said: APT { // Options for apt-get Get { Assume-Yes true; Fix-Missing true; Show-Upgraded true; Purge true; // really purge! Also removes config files The above config should only be used when doing an installation (and then the config file is in the nfsroot) Yep, you're right. and when doing a softupdate. No, must not use this config during softupdate. For doing the softupdate, apt can be called with a specifiek config file. I'm not sure if this apt.conf will be installed to the new system. This may be a bug if it's done. What is so wrong about using this file for softupdates as well? After all, you ought to have the relevant config files in FAI anyway!? Regards, Michael signature.asc Description: Digital signature
Re: xen and fai
Hi, I can install a xen virtual machine with FAI when I export my partitions to the virtual machine. I would like to export a disk to my virtual machine and install (and partition) it with FAI. When I boot with FAI it sees the disk as /dev/sda. But FAI reports /dev/sda: cannot get geometry If then do cfdisk /dev/sda I can partition that disk. Does anybody have any clues? Is it possible at all to export a disk -- and partition it -- in a domain other than domain 0? I don't think so... Regards, Michael signature.asc Description: Digital signature
Re: xen and fai
Michael Tautschnig wrote: Hi, I can install a xen virtual machine with FAI when I export my partitions to the virtual machine. I would like to export a disk to my virtual machine and install (and partition) it with FAI. When I boot with FAI it sees the disk as /dev/sda. But FAI reports /dev/sda: cannot get geometry If then do cfdisk /dev/sda I can partition that disk. Does anybody have any clues? Is it possible at all to export a disk -- and partition it -- in a domain other than domain 0? I don't think so... In domain 0 I have my images (that are on a san). I export in domain 0 the following: disk=['phy:mapper/itext,sda,w'] So - what? What kind of thing is mapper/itext - a partition, the whole SAN? Thanks, Michael signature.asc Description: Digital signature
Re: xen and fai
Michael Tautschnig wrote: So - what? What kind of thing is mapper/itext - a partition, the whole SAN? mapper/itext is a LUN that is exported by our SAN. Thanks; do you have any pointers since when Xen should support partitioning from within the dom-U? AFAIK that simply isn't supported... Regards, Michael signature.asc Description: Digital signature
Re: [Fai-commit] r3511, people/michael/features/setup_harddisks_2/implementation
Some comments on the current disk_config format at http://faiwiki/index.php/Setup_harddisks_2 config ::= disk_config lvm | disk_config raid | disk_config end | disk_config disk[[:digit:]]+( option)* | disk_config [^[:space:]]+( option)* /* fully qualified device-path or short form, like hda, whereby full * path is assumed to be /dev/hda */ Where/why do we need disk_config end? It does exist in the current version of setup_harddisks, so I thought we should keep it. But I don't know the use of it either... | disklabel:(msdos|sun) What about GPT disklabels and other type we need for other hardware architectures? No problem, these are just all the disklabels I know... Feel free to add the somewhere or -- better -- send me the list of labels :-) size ::= [[:digit:]]+%?(-[[:digit:]]+%?)?(:resize)? maybe it's nice to specify the resize option in the same manner as preserve. I'll add it to people/michael/features/setup_harddisks_2/implementation/shdd2-parser - done. Best, Michael signature.asc Description: Digital signature
Re: Windows Partition has to be preserved
Hi there, i have the following problem. We have a running Windows-system here, with three partitions. One primary, two logical. A forth logical one is reserved for the Debian installation with FAI. My config-file looks like that: disk_config disk1 primary - preserve1 ro primary / 50-1 rw;error=remount-ro ;-j ext3 logical - preserve5 ro logical - preserve6 ro ; logical swap 1000 logical /home 50- rw ;-j ext3 [...] So we saw those negative numbers in the sfdisk script; is it possible at all for FAI to fit the above scheme onto the disk? As you say, you've already got 1 primary and 2 logical partitions on your disk. Does the underlying extended partition already occupy the whole disk? Thanks, Michael PS.: This may also be a situation setup_harddisks currently can't handle, but I'm not sure on this one. signature.asc Description: Digital signature
Re: Windows Partition has to be preserved
Hi there, i have the following problem. We have a running Windows-system here, with three partitions. One primary, two logical. A forth logical one is reserved for the Debian installation with FAI. My config-file looks like that: disk_config disk1 primary - preserve1 ro primary / 50-1 rw;error=remount-ro ;-j ext3 logical - preserve5 ro logical - preserve6 ro ; logical swap 1000 logical /home 50- rw ;-j ext3 But I 've got a error message saying this: cannot copy disk_var.sh to directory... At what point of the installation is it saying so? After running all scripts? What version of FAI are you using? If I comment out the first line of the above disk_config it works!!! Do you mean the line primary - preserve1 ro? In this case you will lose your Windows partition... So how has the disk_config file look like, that it works? Your disk_config looks fine, this may be a bug in FAI; however, it is a minor issue that won't affect your system at all. Actually, I don't know the reason why this file is copied at all. Regards, Michael signature.asc Description: Digital signature
Re: Windows Partition has to be preserved
I'm using FAI 2.10.1 with FAI-Kernel 1.10.3. The error-message comes up after: Calling task_partition: ... 'Here is the partition table' ... Setup_harddisks version 0.38fai Probing disks = /dev/hda Disks found: hda. Using config file: /fai/disk_config/FAIBASE Mapping disk name disk1 to hda Creating partition table: LC_ALL=C sfdisk -q /dev/hda /var/log/fai/current/partition.hda Cp: cannot stat '/var/log/fai/current/disk_var.sh': No such file or directory Setup_harddisks did not create /var/log/fai/current/disk_var.sh file. ... and the system stops there. Another fact is, if I use the same disk_config/FAIBASE and there is no Windows XP installed on the machine, it works Maybe there is some trouble with the partitioning Windows vs. Linux (sfdisk)?!? Yes, I know that I loose the data, if i comment out the first line, but it doesn't matter. It is only a test-system with a blank Win XP installation. But I should do the same procedure with 30 other boxes, and there, the data on the Windows partitions is critical. Have you an idea what is going wrong there? Do you find any further information in /tmp/fai/format.log? I suspect this to be a disk geometry issue, see also, e.g., http://www.mail-archive.com/linux-fai@rrz.uni-koeln.de/msg03236.html HTH, Michael signature.asc Description: Digital signature
Re: cfengine error during fai-based install
Hallo Michael! Yes, im sure squid is successful installed include running the postinst script. The Part which I did not posted works successful with cfengine. I replace strings in the squid.conf with ReplaceAll. Like: ReplaceAll ^\# http\_port 3128 With http_port 8080 What about some script that runs before your first cfengine attempt and backups /etc/squid/squid.conf somewhere such that you can later try to run cfengine manually on the backuped version? In the last days i try to setup cgengine2 to resolve the problem. but cfengine2 wont not successful install with fai or after the first reboot. i use this to install cfengine correctly, but not beautiful but works: apt-get install cfengine2 agt-get remove cfengine2 apt-get install cfengine2 - now dpkg starts to configure cfengine ! Which version of cfengine2 are you trying to install (stable/testing/unstable)? What is the exact error message? Regards, Michael signature.asc Description: Digital signature
Re: fai and 64 bit
Hi, These are my first steps in 64bit land and have a problem with setting up FAI I've installed a 64 bit kernel on a 32 bit system (Dell PE1850). I then created a 64 bit chroot on the system. In the chroot I then compiled a 64 bit kernel. In the same chroot I have then created a fai nfsroot with with that kernel. Now when booting from that nfsroot it stops on the following error: request_module: runaway loop modprobe binfmt-464c I'm using a 2.6.16.20 kernel in that nfsroot. What can I do to solve this problem? What about this thread: http://lists.debian.org/debian-amd64/2004/01/msg00129.html HTH, Michael signature.asc Description: Digital signature
Re: fai and 64 bit
[...] What about this thread: http://lists.debian.org/debian-amd64/2004/01/msg00129.html The thread suggests: - ... another module was necessary (System V IPC in my case) ... - Kernel support for ELF binaries - IA32 bit emulation = are all compiled in my kernel If I install that kernel on an other 64 bit system, it boots fine. Only when it's installed in the FAI nfsroot it doesn't work. In some other threads I found - incorrect binformat. (i.e. you try to load a module which was compiled for a different architecture). (http://www.busybox.net/lists/busybox/2005-October/016713.html) - what about your modprobe format, i.e. what does file modprobe report in your nfsroot vs. file modprobe on the native 64bit system? Do you really have a 64bit nfsroot or are you using your 32bit nfsroot? Regards, Michael PS.: What about an older kernel? signature.asc Description: Digital signature
Re: cfengine error during fai-based install
Hi Martin, AFAIK nobody has replied to your post to date - sorry... i a newbie in debian, fai and cfengine (formely suse, without wonderful fai and cfengine) and need now some hint to resolve this problem. [...] Whats wrong ??? Install anyone cfengine2 with fai? the pre or post config-script gets an sed-editor error and the debconf settings will not correct apply. If got no idea about cfengine2, but it should AFAIK work. However, personally I'd suggest you simply use sed/perl instead of cfengine... But back to your question: Are you sure, squid has been installed and configured completely before your cfengine script is run for the first time? Note, that /etc/squid/squid.conf is copied by squid's postinst script and thus it might have been completely missing upon the first attempt to run cfengine!? Just a wild guess, Michael signature.asc Description: Digital signature
Re: Formatting bigger Partitions
Good Morning, did anybody tried to format partitions 3 TB partitions? We came across a problem in sfdisk/fdisk/cfdisk that it can't handle sector sizes, where the numbers are 9 chars. We tried to partition a 6.3TB Partition on an HW Raid 6 and format it later with xfs. With parted it works, but with no other partition tool. Nice to know that parted works. As you might have spotted, I've started working on a new disk partitioning script, but it will not be finished before the meeting at Linuxhotel. Regards, Michael signature.asc Description: Digital signature
Re: fai softupdate
Hi, did anyone see the same errormessage then me: when you start fai softupdate on the installation client, in task_defclass it failes with can't find /var/log/fai/FAI_CLASSES (or similar text). I can be wrong, but on my system FAI_CLASSES is in /var/log/fai/current/. Actually task_defclass seems to be very inconsistent according to the use of $LOGDIR (which points to /var/log/fai/current/) and /var/log/fai. I'll file a bug on this issue. As a workaround, you might try to do fai -N softupdate to have FAI redefine the classes. Regards, Michael signature.asc Description: Digital signature
Re: fai softupdate
Hi, did anyone see the same errormessage then me: when you start fai softupdate on the installation client, in task_defclass it failes with can't find /var/log/fai/FAI_CLASSES (or similar text). I can be wrong, but on my system FAI_CLASSES is in /var/log/fai/current/. Please ignore my previous message, except for the -N workaround. Actually the use of $LOGDIR in defclass is fine, the bug must be somewhere else. Please check your installation log for errors during task_savelog, unless you are using a hook for savelog which makes FAI skip the task completely. In the latter case the bug would be found: FAI_CLASSES is copied in task_savelog ... Regards, Michael signature.asc Description: Digital signature
Re: Script Detection of Processor Architecture
Could you be more precise on the list of processor types you'd like to distinguish? Debian, e.g., does only provide -686, -k7 kernels, so it would be pretty easy. Regards, Michael signature.asc Description: Digital signature
Re: speaking of classes and softupdate...
On Tue, May 23, 2006 at 11:47:45AM +0200, Adrian von Bidder wrote: I'm a bit irritated about fai's use of /var/log/fai/FAI_CLASSES in softupdate. Shouldn't this be stored in something like /var/lib/fai/FAI_CLASSES? Just a minor issue, but I usually expect to be able to wipe old files on /var/log without any conseuquences for the system. IMO this should not be stored anywhere... you should always be able to obtain the classes using the $FAI/class/ scripts. This is at least how I developed the softupdate code... I don't even know about the goal of the FAI_CLASSES file - I'm always using the -N switch... Regards, Michael signature.asc Description: Digital signature
Re: Script Detection of Processor Architecture
I have kernels compiled for p3, p3-smp, p4, p4-smp, 586mmx, k8, and k8-smp. Essentially, I'd like to be able to detect which selection under Processor Type in the kernel source config is best compatible with the running processor. It'd be nice to detect SMP, as well, and I'm not sure if cpuinfo can detect multiple processors if the kernel is a uniprocessor build. I'm wondering if there might be a way to make the distinction based on the cpu flags, like sse2 or mmxext and so forth. What would you suggest? http://www.cs.helsinki.fi/linux/linux-kernel/2001-30/0026.html seems to be the only way to detect the SMP/UP-thing, if it still works. Thereby you should be able to distinguish UP vs. -smp variants. The next thing to find out is probably Intel vs. AMD, which should be easy (vendor_id). I haven't got any P3 here, so I cannot tell you how they announce themselves, but the P4s here either call themselves Intel(R) Pentium(R) 4 CPU or simply Intel(R) Xeon(TM). Only the flags are somewhat different... Good luck, Michael signature.asc Description: Digital signature
Re: client machine
Hi, I'm from Brazil (sorry for my english) and I'm using FAI 2.10 I'm having a problem with the boot of the client machine... It's the dmesg. [...] Are you trying to install the client, or has it been installed (using FAI) and should now boot and bring up the normal system? If you are trying to install it, which kernel are you using? Are you booting using PXE or a FAI-CD? Thanks, Michael signature.asc Description: Digital signature
Re: client machine
So - you are trying to install the system!? Yes. What about a newer kernel, some 2.6.x? I think it's not the problem, because the client machine have an old hardware... like K6-2 processor... Do you think I have to try it? My concern was missing support for tmpfs or the like, but this is just a guess... Regards, Michael PS.: Please CC the list in all replies, thanks! signature.asc Description: Digital signature
Re: xen-vm install with FAI - dhcp requests being ignored as not authoritative
Hi, On my way to install a xen vm with FAI over the network. I have a ip autoconf-enabled kernel, but am experiencing the following problem. On the xen-users list nobody had an idea, it seems the xen kernel makes wrong dhcp requests, as my dhcp server works perfect with all other machines amd setups. Maybe some other FAI user has an idea why this happens with xen 3.0.2: When trying to get an IP adress, nfs root host and directory information for the xen vm from dhcp, the dhcp server's syslog reads: May 4 05:13:21 amalgam dhcpd: DHCPREQUEST for 255.255.255.255 (1.2.3.4) from 00:16:3e:00:00:11 via eth0: ignored (not authoritative). Is your dhcp server authoritative? Maybe one of these http://www.mail-archive.com/linux-fai@rrz.uni-koeln.de/msg01610.html http://www.bsdforums.org/forums/archive/index.php/t-29534.html solves your problem!? [...] HTH, Michael signature.asc Description: Digital signature
Re: Proposal for a new disk configuration utility
For those of you who are already using LVM/RAID: Please check, whether the format described below allows for configuring your exact setup using FAI - thanks! Looks like it supports everything the old one supported, except: 1. RAID - name metadevices and support nesting Previously you could set up a raid 10 device with just RAID 0 + 1 modules. Just define your raid 0's, then use those metadevices as components of a raid 1. Would that be possible? Yes, I think we should allow for that, but the implementation must take care of possible recursion... Also you used to be able to specify which metadevice each one is. Normally you only care for consistency. These problems are perhaps peripherally related. I'm not entirely sure, whether I understood that correctly: Are you talking about explicit numberings of the devices? My idea was to make the way the RAID config works as similar as possible to the configuration of physical disks, where the numbers are set implicitly by the ordering as well. 2. Support optional specification of a PV to put a volume on. The script I wrote allowed you to do this. For instance you might use one bug VG for flexibility, but specify that certain partitions end up on one physical volume seperate. For example database journal volumes are often moved onto seperate disksets for busy OLTP servers. However, you don't want to set up seperate VGs for them because then you can't move LVs between them. Actually I didn't even know that this is possible... I'll put this in my notes. Otherwise, I like the fact you're using EBNF to specify, makes implementing it look more tempting :). Actually reading the EBNF I think you've got the latter concern covered; is that what is intended with this rule? type ::= primary /* for physical disks only */ | logical /* for physical disks only */ | raid[015] /* raid level */ | [^/[:space:]]+:[^/[:space:]]+ /* lvm logical volume: pv name and lv name*/ What do you mean by the latter concern? I'm not really sure what I am to tell you here!? Thanks a lot for your input, Michael signature.asc Description: Digital signature
Re: [PATCH 00/12] A mechanism for facilitating vserver builds
[...] Once I can figure out how to get past the subversion red tape then I'll publish these changes on the FAI subversion repository as a branch. AFAIK you need to get yourself an alioth account and then ask Thomas to grant this user access permissions, see also http://faiwiki.debian.net/index.php/Fai_subversion_howto Regards, Michael signature.asc Description: Digital signature
Proposal for a new disk configuration utility
Hi, on linux-fai-devel I've initiated a discussion of a new new disk configuration utility. You can find the thread at http://www.uni-koeln.de/bin2/maillist/linux-fai-devel/20060429.151232/171697 if you aren't subscribed to linux-fai-devel. Any input is highly appreciated! Cheers, Michael signature.asc Description: Digital signature
Re: /dev, /etc, /var- and mkrwsize()
[...] Could you please elaborate on the use of these directories? Usually I wouldn't expect to much to be written to the NFSROOT ... One can learn about the use of those directories from the Filesystem Hierarchy Standard (FHS). Copied from http://www.pathname.com/fhs/pub/fhs-2.3.html: [...] Sorry, my question was not clear enough, I do know about the general use of those directories, but I do not know why one would want them to be writeable. PS.: Or is it just the LVM things you mentionend lately? Indeed I am looking at it because of lvm. If I had the whole of /etc rw I wouldn't have to give special care to the lvmtab file. And mind you that mkrwsize handles only directories, not files. I also believe that others see here a problem too. When announcing mkrw(), Thomas Lange wrote in the changelog that he wish the unuions were already here. And, since we are talking about automation, we have to try and plan for all sorts of requirments, as well as scalability. For example, I don't know what will be required to install a machine that has raid. Ok, now I can see your point. I'd suggest you file a (wishlist) bug for this thing not to be forgotten! Thanks, Michael signature.asc Description: Digital signature
Re: /dev, /etc, /var- and mkrwsize()
With 2.10 and hardly anything added to the default nfsroot, du shows: 16K for /dev, 1.8M for /etc and 78M for /var. The large consumers of those 78M are: 17M nfsroot/var/lib/apt 5.8Mnfsroot/var/lib/dpkg 15M nfsroot/var/cache 41M nfsroot/var/tmp Perhaps mkrwsize() from subroutines-linux should create a 5m for /etc another 10m for /var, where for /var the large consumers from above will be excluded and another 100k for /dev Could you please elaborate on the use of these directories? Usually I wouldn't expect to much to be written to the NFSROOT ... Thanks, Michael PS.: Or is it just the LVM things you mentionend lately? signature.asc Description: Digital signature
Re: changing the default base.tgz
Hello list, i'm trying to install our heavily customized Ubuntu on a bunch of laptops using FAI and the extra fai- distributions package. I've successfully created a customized base.tgz but since it's named after the disatribution (base-flap.tgz) i doesn't get used by the installation process. Is there a way to customize the name of the base image? It _seems_ to be hardcoded in subroutines-linux, function 'task_extrbase'. Ok, IMHO there are a few things to note here: - Why do you need to customize base.tgz at all? - If it is really required, why don't you just manually replace the base.tgz file? - It is hardcoded there as it doesn't make sense anywhere else, but you could simply replace extrbase by a hook that does the same things apart from the different filename and an additional skiptask extrbase HTH, Michael signature.asc Description: Digital signature
Re: Grub and XFS Problem during installation with FAI
[...] Thats the error I get using xfs on / (without /boot with ext2/3). Maybe I missed something, but where do I find bug #246111? Bugtracker or..? http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=246111 [...] is one of the possible ways :-) Regards, Michael signature.asc Description: Digital signature
Re: Kernelversion is wrong
Hallo list, I have a new problem with the kernelversion. The log said: Error. Kernel package /usr/lib/fai/kernel/kernel-image-2.6.14-fai-kernels-1-i386.deb not found. It's right, the kernel is not in the directory, but the kernel 2.6.15 is available. I'm changed in the file /etc/fai/make-fai-nfsroot.conf the entry to version 2.6.15. and now is the errormessage FATAL: Could not open '/boot/System.map-':No much file or directory What's wrong? What can I do? Did the kernel-image-2.6.15-fai-... get installed? Did you run make-fai-nfsroot with debug=1 make-fai-nfsroot -v This should tell you as much as possible... Regards, Michael signature.asc Description: Digital signature
Re: auto-softupdate
hi, I'm thinking to put /usr/sbin/fai softupdate in the cron of every client. is there a best practice for doing this. let's say that this softupdate will happen every night. What happen If I'm working on a the configuration space, and that this configuration is not yet working well. To avoid such situation, I was thinking to give a version to the configuration space, and to do a script on any client which will do a softupdate only if the version of the configuration space is higher than the last softupdate done. Does some one of you has already think of this; is this a good idea ? We are using svn to manage our configuration since a few weeks, before the whole thing was based on CVS, which is completely supported by FAI already and works very well. Usings tags you can even commit to the repository, as long as you don't mark the changes as stable. However, I do not suggest automated updates in a mission-critical environment, not even on Debian-stable. We do not even run manually invoked concurrent updates on all hosts anymore as this, e.g., might break both of your DNS servers at once, which happened here :-) HTH, Michael signature.asc Description: Digital signature
Re: MENU question/problem
So, the point is, what I am trying to do is ask all of the questions at the very first (menu.lst) screen so that it is a FAI.. Is that possible? AFAIK you cannot ask questions at that point, but you can set some variables to provide the relevant data - but that makes me ask, which questions you want to ask. Thanks, Michael signature.asc Description: Digital signature
Re: 2.6 kernels, initrd and /sys
Yo! (Using fai 2.8.4 from Debian sarge) Using the 2.6.15 kernel from backports.org as post-installation kernel with the same kernel (plus non-modular network drivers and NFS root support compiled in) as install kernel, I see that mkinitrd.yaird fails, presumably because /sys is not mounted during installation. There certainly is an official fix described somewhere, isn't there? Sorry, I don't really understand what you are doing and at what point mkinitrd.yaird is involved!? Are you trying to build a new kernel image? If that is true, at what stage is it FAI's fault that it doesn't work? Thanks, Michael signature.asc Description: Digital signature
Re: MENU question/problem
Ok, I now have: 30-menu.source in my class directory. The 30-menu.source looks like: [...] (didn't check whether the code is correct) Don't I need to be calling this file from somewhere? Or is that all I need to do? That's FAI's job :-) I also see that some places the script is called as: 30-menu.source and some call it S30-menu Does that make a difference? The difference is that the latter (S30...) is deprecated. Regards, Michael signature.asc Description: Digital signature
Re: cant open /dev/hda
Hi all, I'm new to FAI so please bear with me. I'm trying to install etch with FAI with nfsroot and PXE booting. the kernel boots but stops at the partitioning of disks, the following appears in fai.log: sfdisk: cannot open /dev/hda for reading /dev/hda: no such file or directory But I get lines like this in the kernel.log 4hda: attached ide-disk driver. 6hda: 160086528 sectors (81964 MB) w/2048KiB Cache, CHS=158816/16/63, UDMA(100) 6Partition check: 6 hda: [PTBL] [9964/255/63] hda1 hda2 hda3 I've tried a number of the fai kernel versions with similar results. any suggestions ? if more logs are needed please tell me, didnt want to spam the list Which fai-version are you using? Most likely you are missing the udev-Package in your NFSROOT (and thus there is no such file/device as hda in your /dev directory). HTH, Michael signature.asc Description: Digital signature
Re: nfsroot boot problems
Hi, VFS: Cannot open root device nfs or unknown-block (0,255) Please append a correct root= boot option Kernel panic - not syncing: VFS: Unable to mount root fs on unknown block (0,255) It couldn't mount your client to FAI -server, i think that you haven't chosen the NFS root option in kernel configuration. Check it as Y (build into kernel). If you build config kernel from scratch without FAI template configuration, you might not see that option. Try to use the FAI template config. There're config file for kernel 2.6.8 , 2.6.14 on FAI download repository. IMHO you need to enable one of the IP autoconfiguration options as ROOT_NFS depends on IP_PNP in the kernel configuration. HTH, Michael signature.asc Description: Digital signature
Re: FAI for AMD64, current status?
Hi, I'm running etch on a 2*2 Opteron box, and tried to set up FAI (mainly to create a nfsroot to be able to rebuild this box afterwards) - but fai seems to be uninstallable at the moment. Should I go for non-Debian FAI packages? ... Which is most probably not FAI's fault - which package makes it uninstallable? As far as I can remember, libapt-pkg-perl caused trouble at our site. Regards, Michael signature.asc Description: Digital signature
Re: debootstrap: security updates
[...] I think the delete_base_packages function should be reworked so it does not remove security updates. At the moment it removes packages with that command, where $p is a name from base-pkgs.lis (e.g. perl-base) # rm $archivedir/${p}_*.deb [...] In my opinion it would be better to have base-pkgs.lis to include the exact version of the package, if that is possible - Thomas? Regards, Michael signature.asc Description: Digital signature
Re: experimental fai-multi-distribution packages
Perhaps the normal fai has reached the point that a contrib directory is in place. By that I mean a directory which will include scripts and patches that enhance the normal fai and which the upstream author, which is the debian maintainer in this case, is not willing to support. Note that I didn't look whether fai-multi-distribution is not too big to be distributed as patches to the normal fai. I do believe that distributing patches to convert the normal fai to fai-multi-distribution can help fai* users and developers. What about a separate Debian package fai-md? Did Henning change anything in the base package or are the changes just add-ons? In the latter case, there is not much to do, else we'd need to reconsider... Just an idea, Michael signature.asc Description: Digital signature
Re: fai class hierarchy brainstorming
[...] CLASSNAME/ class_scripts/ - (1) NR[-NAME] variables - (2) package_config - (3) disk_config (4) debconf (5) a debconf file for that class README (6) hooks/ (7) TASKNAME TASKNAME depends(8) scripts/ (9) NR[-NAME] NR[-NAME] version And here the explanations: (1) the classic class directory Do we need it in that kind of hierarchy or even more, does it make sense? Note, that for that kind of structuring you need to have classes defined already! (2) a file which was before class/CLASSNAME.var (3) a file which defines all packages needed for that class. Should we make a directory as before? maybe we want some packages be only installed if another class is defined? On the other hand, the proposed way we don't need to name the file CLASSNAME in a directory package_config. I think a class name should appear in as little locations as possible, to make class renaming easy. IMHO this should be a directory and the files should be named NR[-NAME] such that the tasks could read them one by one and things like Jürgen's troubles of installing alsa before kernel-image could be solved by putting kernel-image into some file with a lower number. (4) a file that was before: disk_config/CLASSNAME (5) a debconf file for that class (6) a file containing some info what this class is about Should not be mandatory, although it is pretty useful. (7) the classic hooks directory, only that the files inside don't need the classname added again. (8) a file with a list of classes this class depends on A great idea, but IMHO hard to implement - how would you deal with those dependencies, what would they mean to FAI? We'd need to take care of recursive dependencies! (9) the classic scripts directory, only there are no extra subdirectories and the CLASSNAME doesn't appear again here. (19) file with a version string inside The same as with the README file... What about the files/ directory? Wasn't that a major concern? This is just a rough idea, I think will try to make an example implementation some day 'cause I like to see how it feels :) What do you think? Anything wrong or good with that? I think this is a really great idea and well thought too, but implementing it won't be as easy. We need to try, though. Best regards, Michael signature.asc Description: Digital signature
Re: Install nodes with re-configured kernel image
Hi, On Wed, 18. Jan. 2006 at 05:29:33 -0800, Brian Showalter wrote: You can see the updated tutorial on the FAI wiki at http://faiwiki.informatik.uni-koeln.de/index.php/User:BrianShowalter/Using_customized_kernels_with_FAI. and what about modules for the custom kernel? For example alsa? If those alsa package is installed before the kernel is installed, than the kernel won't like to be installed anymore because there's already an existing /lib/modules/2.x.x-xxx directory. Add silent_modules = yes to /etc/kernel-img.conf, which thus needs to be copied before a kernel might get installed; to do so, I added the hook updatebase.DEFAULT with the following content: #!/bin/bash fcopy -Bi /etc/apt/sources.list fcopy -Bi /etc/kernel-img.conf Regards, Michael signature.asc Description: Digital signature
Re: Install nodes with re-configured kernel image
Hi, On Wed, 18. Jan. 2006 at 17:35:10 +0100, Michael Tautschnig wrote: Add silent_modules = yes to /etc/kernel-img.conf, which thus needs to be copied before a kernel might get installed; but this is an general ignore of this warning. If everybody in every case could safely ignore this warning, why is it ever displayed? ;-) The above line is meant for the I pretend I know what I'm doing case... Is there a way to add dependencies with make-kpkg? Or modify the package after build to add dependencies? In what way would dependencies help you? Thanks, Michael signature.asc Description: Digital signature