Re: fcopy and /dev/null
> On Tue, 17 Sep 2019 11:20:42 +, Frank Morawietz > said: > But every time fcopy is used, it complains… > “Can’t open /dev/null: No such file or directory” I guess this comes from a redirect of some output to /dev/null. I've checked the basefiles on http://fai-project.org/download/basefiles/ and the debian, centos, Ubuntu basefiles have a /dev/null. But the opensuse basefile does not include /dev/null. Not sure if this the fault of opensuse or the tool that creates the basefiles (called rinse, also maintained by me). For now you can try to fix this in a hook using mknod. -- regards Thomas
Re: AW: fcopy and /dev/null
> On Tue, 17 Sep 2019 12:54:48 +, Frank Morawietz > said: > I added the mknod command ("mknod --mode=666 /dev/null c 1 3") to the repository hook, right before calling fcopy the first time. You have to use mknod --mode=666 $target/dev/null c 1 3 Maybe /dev/null is needed in the very beginning, so check when this device is created. Or try to add /dev/null to the basefile. -- regards Thomas
Re: Adding the option to customize grub boot arguments...
> On Wed, 18 Sep 2019 05:08:46 +0200, Robert Spiteri > said: > Any update on this please? Not yet implemented. -- regards Thomas
Using NFS v4 when mounting the nfsroot
Since a long time, I've tried using NFS v4 for the nfsroot. This did not worked since FAI uses overlayfs for making the read-only nfsroot writeable. Therefore FAI force a NFS v3 mount for the nfsroot which works nicely. But now it's possible to use NFSv4. I put some effort in debugging the setup, and asked the overlayfs maintainers for help. It's a complicated mixture of NFS and acl support (POSIX and nfs4 acl) and what overlayfs expects from the filesystems in respect to certain xattr. There may be an improvement inside overlayfs in the future. But currently we can use NFS v4 for the FAI nfsroot with a very simple change. Just disable acl for the exported filesystem on the FAI server which holds /srv/fai/nfsroot. More is described here: http://fai-project.org/fai-guide/#_a_id_nfsv4_a_how_to_use_nfs_v4_with_the_nfsroot -- regards Thomas
FAI.me service now support backports for Debian 10 (buster)
The FAI.me service for creating customized installation and cloud images now supports a backports kernel for stable release Debian 10 (aka buster). If you enable the backports option, you will get currently the 5.2 kernel. This will help you, if you have newer hardware that is not support by the default 4.19 kernel. The backports option is also still available for the images when using the old Debian 9 (stretch) release. The URL of the FAI.me service is https://fai-project.org/FAIme/ -- regards Thomas
Re: LVM/chroot: WARNING: Device /dev/vg_sys/swap not initialized in udev database even after
This bug was closed in FAI 5.8.4. There's a Debian bug number: #925247 You are using an old version of FAI, so please update. If you are still using stretch, you can use the version from backports or from the fai-project.org repository. > I'm preparing Debian Buster for use with fai, and I'm getting a very > long delay for the installation when using LVM partition layout. -- regards Thomas
FAI 5.8.7 and new ISO images using Debian 10
A new FAI release is available. It now supports apt keys in files called package_config/CLASS.gpg. Before we only supported .asc files. fai-mirror has a new option -V, which checks if variables are used in package_config/* and uses variable definitions from class/*.var. I've also created new ISO images, which now install Debian 10 by default. They are available from https://fai-project.org/fai-cd If you need a newer kernel, checkout the FAI.me service (http://fai-project.org/FAIme/) which can also build ISO images using the kernel from backports which currently is 5.2. -- regards Thomas
Re: potential patch to check-cross-arch
> On Wed, 30 Oct 2019 11:15:37 +, Matteo Guglielmi > said: > yes, > that works too for suse. Thanks for the feedback. The patch is now included -- regards Thomas
Re: AW: PXE boot with UEFI
> On Wed, 4 Dec 2019 09:40:45 +, Frank Morawietz > said: > So it looks like the UEFI tftp connection refused error happened where ldlinux should be transferred. > Is there anything else I could examine to diagnose further? I had problems with my thinkpad in the past when booting UEFI. First I needed the newest syslinux binary, in those days I was still running stretch. Then, my thinkpad T480 needs a long time and shows several transmision error (currently I cannot say which exact messages) when doing UEFI boot. But after multiple error message it succeeds to boot. -- regards Thomas
Re: Forgotten home dir
> On Thu, 5 Dec 2019 12:54:08 +, Frank Morawietz > said: > Hello all, > I came across a strange problem during installation: FAI seems to forget the home directory of the root user in the installation system. > echo $HOME My impression is, that FAI does not forget the home dir, it was never set at all. I can see a lot of processes from root on my system, that are running without having $HOME set. HOME is normally set by a login process. Since the fai script is started by dracut, it seems that the HOME is never set inside dracut. My impression is, that this is a default dehaviour. You can define variables in class/*.var files. This should also work for HOME. -- regards Thomas
Re: get-boot-info
> On Fri, 13 Dec 2019 14:40:34 +0100, Szilard Matyas-Szegedi > said: > I might be missing something, but I don't see why that wait/poll section is necessary as in the next section you poll for default route to be available which will only be available if at least > an interface other than loopback with an IP address is up. You are right. I will remove the section. -- regards Thomas
Re: get-boot-info
You issue is now fixed in https://github.com/faiproject/fai/commit/b95d56084255d8512e16724f38aab726dc7fb9d3 -- regards Thomas
Re: get-boot-info
> On Wed, 11 Dec 2019 13:57:44 +0100, Szilard Matyas-Szegedi > said: > As dev is no longer used anywhere else from now on, the section getting it could be cleaned-up as well with the exception of the dummy variable: This section is still needed. It waits until an interface (execpt lo) becomes active. Any proposal how to clean up this section? -- regards Thomas
New feature for FAI.me service, set kernel cmdline options
Several people asked if it's possible to specify additional kernel cmdline parameters in the FAI.me service. Now I've implemented this feature. After toggling to the advanced settings, you can add your options. These will replace the default "quiet" option. This feature is currently only available for the installation images, but not for the cloud images. -- regards Thomas
Re: potential patch to check-cross-arch
> On Thu, 17 Oct 2019 10:35:24 +, Matteo Guglielmi > said: > For those interested, > the final and working patch that takes care of all possible > cases is the following one: > _ls=$(realpath -m $_ls) # insert these two lines before > _ls=$target/${_ls#/target/} # line n.36 (info=$(file $_ls)) > info=$(file $_ls) Sorry, this does not work on Debian systems :-( Here's some debugging output. + '[' -f /srv/fai/nfsroot/bin/ls ']' + _ls=/srv/fai/nfsroot/bin/ls ++ realpath -m /srv/fai/nfsroot/bin/ls + _ls=/srv/fai/nfsroot/usr/bin/ls + _ls=/srv/fai/nfsroot//srv/fai/nfsroot/usr/bin/ls ++ file /srv/fai/nfsroot//srv/fai/nfsroot/usr/bin/ls + info='/srv/fai/nfsroot//srv/fai/nfsroot/usr/bin/ls: cannot open `/srv/fai/nfsroot//srv/fai/nfsroot/usr/bin/ls'\'' (No such file or directory)' I works when using this lines. Does this work for you with SUSE? _ls=$(realpath -m $_ls) _ls=$target/${_ls#$target} -- regards Thomas
Re: non zero exit codes for classes scripts
Hi Martin, which FAI versions are you comparing? -- regards Thomas
Re: non zero exit codes for classes scripts
Hi Martin, I can find no change that causes a change in the behaviour for that between 5.8.1 and the newest FAI version. As you pointed out an "exit X" in a .sh script causes fai-class to stop defining more classes. You can try to set an error code inside a .sh file by using a call to task_error xxx. I'm not sure if this will work, since fai-class runs in a subprocess. After looking into the code it seems that fai-class almost always ends with exit 0. -- regards Thomas
Re: Configuring uefi boot
> On Mon, 25 Nov 2019 16:53:24 -0800, CSCI Technician > said: > The related classes defined are GRUB and GRUB_EFI. GRUB_PC is not > defined. What I'm not sure how to fix is to get grub-install to realize > this is an efi and not a bios install so that it should be using the efi > path instead of the i386-pc path. I do not know how grub decides or detects when to use EFI. In FAI we check if /sys/firmware/efi exists. Is /sys mounted to /target during the installation? Otherwise this is a grub, not a FAI problem. -- regards Thomas
Re: Install RHEL8/CentOS 8
You should read the FAI guide, especially https://fai-project.org/fai-guide/#_a_id_otherdists_a_installing_other_distributions_using_a_debian_nfsroot You need a basefile for RHEL 8 but rinse does only support CentOS 7 yet. -- regards Thomas
Re: Possible bug in LAST/50-misc
> On Mon, 4 Nov 2019 09:29:00 +0100, Steffen Grunewald > said: > (It seems that, while considering all dependencies, FAI sometimes reorders or > splits installs, with a flag that doesn't consider dependencies at that There's a variable which defines the maximal amount of packages given to a apt-get (or aptitude, apt or similar) call. This is set in class/FAIBASE.var: MAXPACKAGES=800 -- regards Thomas
Re: Possible bug in LAST/50-misc
I've added -y to the apt-get call. But you should check why this happens. > 1 not fully installed or removed. I guess something went wrong before. -- regards Thomas
Re: Configuring uefi boot
> On Tue, 26 Nov 2019 09:07:00 -0800, CSCI Technician > said: > I have updated my FAI install and NFSROOT to one of the latest > versions... however my config still needs to be updated. I noticed this > when comparing some of the scripts in my config directory vs those in > the FAI repo. Is there any automated process to upgrading the config to > the latest... at least those scripts that have not been customized? or > is this a manual process? There no automatic process. If you have your config space in git and created a branch from the offical config space, you can try to rebase or merge. -- regards Thomas
Re: potential patch to check-cross-arch
> On Wed, 9 Oct 2019 15:40:02 +, Matteo Guglielmi > said: > This script: > /srv/fai/nfsroot/usr/lib/fai/check-cross-arch > fails (line 36: info=$(file $_ls)) on openSUSE > and Suse Linux Enterprise systems because: Which openSUSE version are you using? Does file -L help? -- regards Thomas
Re: /etc/resolv.conf contents reversed after rebuilding nfsroot
Hi Merlin, if I understand your correct, the resolv.conf inside the nfsroot is always correct. Upgrading the nfsroot using -kvp did not rebuild the whole nfsroot, it just upgraded it without changing resolv.conf Using fai-make-nfsroot -fvp rebuilds the whole nfsroot. But building the nfsroot just copies the /etc/resolv.conf from the FAI install server into the nfsroot. This should/can not reverse the order of the entries in resolv.conf. When the client does a network installation using PXE it uses the resolv.conf from the config space or from inside the nfsroot. See lib/subroutines line 1066ff We also use lib/create_resolv_conf which copies the resolv.conf from the initrd (created by dracut) or loops through the list defined by $DNSSRVS (which comes from the DHCP server and is created by ./lib/dhclient-perl). The difference between 5.3.6 and 5.8.x may be that we copy the resolv.conf from the config space nowadays. IIRC we did not that before. But I'm not sure. Maybe the order of the nameserver in $DNSSRVS was always reversed but noone had problems with it. So, please check all resolv.conf versions (on the FAI server itself, in the nfsroot, inside the initrd, on the client) and tell which are the wrong version. Maybe change all cp commands of a resolv.conf to cp -v. Also check the order of DNS servers in dhcpd.conf on your dhcp server. -- regards Thomas
FAI 5.8.9 released
FAI 5.8.9 was released which fixes a small bug in install_packages that created package lists with additional packages when using different PACAKGES lines. -- regards Thomas
Re: 20 years of FAI
Hi Denny, thanks a lot for your kind feeback. I'm still very interested in getting feeback via the FAI http://fai-project.org/questionnaire. I do not have a wishlist on amazon, but if you meet me at some open source event, you can buy me a beer. -- regards Thomas
Re: FAI_BASEFILEURL and trailing slash
> On Tue, 11 Feb 2020 11:28:15 -0800, andrew bezella > said: > in most places the documentation includes a trailing slash but it could > be updated for consistency in e.g., https://wiki.fai-project.org/index. > php/Variables and https://fai-project.org/fai-guide.html#_a_id_c3_a_the > _configuration_space_and_its_subdirectories Thanks for your debugging. This is now both fixed. -- regards Thomas
Re: Upgrade error in nfsroot
> On Wed, 26 Feb 2020 09:20:47 +0100, Andreas Heinlein > said: > Hello, > I have to correct myself: upgrading the nfsroot worked, but the NFS root is not functional now. When booting, it says "cannot execute /etc/init.d/rcS". Indeed, /etc/init.d/rcS does not exist. > Is rcS supposed to be a real file or a link somewhere? I must be a link to fai rcS -> /usr/sbin/fai -- regards Thomas
downtime of fai-project.org on saturday feb, 29th
Due to a power shutdown the whole fai-project.org domain will be offline on saturday feb, 29th from 8:00 to about 15:00 CET. -- regards Thomas
Re: /etc/hosts in scripts/FAIBASE/10-misc
You are right, I've moved the code in this commit. Thanks for reporting this issue. https://github.com/faiproject/fai-config/commit/18abec119171380a78b87b81bad392ce8527129e -- regards Thomas
Re: FAI build using a small disk image with LACP support
> On Fri, 28 Feb 2020 16:21:38 +1300, Andrew Ruthven > said: > All the servers support Redfish, so I can script uploading a disk image > to be booted from. > What I'm thinking is prepare a FAI boot image based on an nfsroot, with > minimal configuration to bring up a degraded LACP bond. > Has anyone already done this? No. But using fai-cd you can create different flavours of a boot CD. A boot CD with just the kernel and the initrd, or a CD including additionally the nfsroot. In your case a boot only CD (kernel+initrd) should be fine. Have a look at the man page dracut.cmdline(7) (which we are using to build the initrd). This includes already options for bonding: bond=[::[:[:]]] Setup bonding device on top of . is a comma-separated list of physical (ethernet) interfaces. is a comma-separated list on bonding options (modinfo bonding for details) in format compatible with initscripts. If includes multi-valued arp_ip_target option, then its values should be separated by semicolon. if the mtu is specified, it will be set on the bond master. Bond without parameters assumes bond=bond0:eth0,eth1:mode=balance-rr Also read man fai-cd(8). -- regards Thomas
Re: commit c817ba1 - test for systemd-machine-id-setup
> On Thu, 23 Jan 2020 16:58:27 -0800, andrew bezella > said: > and i was wondering if the removal of "$target" in the "-f" test for > `systemd-machine-id-setup` was intentional? since the next line is > "$ROOTCMD systemd-machine-id-setup" it would seem that the test should > still be for "-f $target/bin/systemd-machine-id-setup"? Hi andy, you are right. It seems that $target was lost. I see no reason to remove it. Fixed in 2ac9f6afa62cb0446db9f622036e28e4c540d901 -- regards Thomas
FAI 5.9 hash sum mismatch now fixed
Hi, you may have seen some warnings about Hash Sum mismatch for the deb files of version 5.9 is you were using both the fai-project.org package repository and some official Debian mirror. This looks like this Err:1 http://deb.debian.org/debian testing/main amd64 fai-server all 5.9 Hash Sum mismatch Hashes of expected file: - W: Sources disagree on hashes for supposely identical version '5.9' of 'fai-server:amd64'. E: Failed to fetch This was because I've rebuild the deb packages which I copied to the fai-project.org repository which lead to other checksums. The content of the packages were identical. This is now fixed. I'm sorry if I caused some troubles for you. -- regards Thomas
FAI 5.9.1 with CentOS 8 support
FAI 5.9.1 is now released including CentOS 8 support. New ISO images are now available on the web page https://fai-project.org -- regards Thomas
FAI 5.9.2
The new FAI version 5.9.2 now includes the nvme cli tool for better listing of nvme devices. It also fixes a bug that btrfs file systems were not preserved. setup-storage itself now stops syncing software raid devices, which was formerly done in a hook. fai-monitor can now use custom formats for the time. This release add supports for the next dracut release, which adds two more packages. I plan to release the next dracut version in a couple of weeks. New ISO images will be released during the next days. -- regards Thomas
Re: FAI 5.x - Roadmap FAI-CD with UEFI
Hi Marco, wollte dir nur sagen das UEFI support seit FAI 5.9 seit Januar in fai-cd drin ist. Viel Spass beim ausprobieren und ueber Rueckmeldungen freu ich mich. -- viele Gruesse Thomas
Re: Package dependency problems
You can use dpkg-divert (also used by fai-divert) for diverting a file, even before it's installed I guess. > On Fri, 10 Jan 2020 14:11:43 -0600, John G Heim > said: > Geez, what a load of bullstuff. Maybe I can add a hook to create that > file before fai starts installing packages. -- regards Thomas
Re: FAI 5.9 released, adding UEFI support
> On Wed, 8 Jan 2020 18:26:28 +0100, Martin Krämer > said: > After a short check I found that the line: > "search --set=root --file /FAI-CD" is missing within my grub.cfg and within your "FAI server installation" entries. > So as soon as I add this line as it is within "Client standalone installation" and "Rescue System" all entries are working. Thanks for the feedback and debugging help. I've created new ISO images and will include the fix into the next FAI release. -- regards Thomas
20 years of FAI
I wrote a small blog post[1] about what has happened since the last 20 years since the release of FAI 1.0. I would be happy to collect more user reports of FAI. If you never send the FAI questionnaire [2], please do so now. If you already send a report but this was long time ago, please update your report. If you agree, your feedback will be listed on [3] and the logo of your project or company will be listed on this nice overview [4] of FAI users. [1] https://blog.fai-project.org/posts/fai-5.9/ [2] https://fai-project.org/questionnaire/ [3] https://fai-project.org/reports/ [4] https://fai-project.org/references/ -- regards Thomas
Re: fai-sed exit code
>>>>> On Fri, 10 Jan 2020 16:16:19 +0100, Andreas Heinlein >>>>> said: > Am 10.01.20 um 16:02 schrieb Thomas Lange: >> I want to distinguish if fai-sed has nothing to change or changed the >> file. Therefore in one the cases it has to return and exit code != 0. >> >> Do you have any better idea? > If you want to do this, no, I have no better idea. You would have to exit any script using fai-sed with "exit 0" at the end, or the fai master process will treat the script as failed. > Maybe you could add a "-q" switch to fai-sed that will turn off that behaviour for those who don't need it? I can add such an option, which I already have with fcopy. I a shell script, you can use this code to always get exit 0 from a command. fai-sed .. || true -- regards Thomas
Re: fai-sed exit code
> On Fri, 10 Jan 2020 14:47:15 +0100, Andreas Heinlein > said: > can you explain the purpose of fai-sed exiting with '1' if the file was changed? By default, this would mean a script *fails* with this exit code if a file was actually changed. > Would also be nice to update the example scripts to use fai-sed instead of sed. I want to distinguish if fai-sed has nothing to change or changed the file. Therefore in one the cases it has to return and exit code != 0. Do you have any better idea? -- regards Thomas
Re: fai-cd 5.9 grub-mkstandalone error
> On Fri, 10 Jan 2020 16:08:03 +0100, Andreas Heinlein > said: > I just tried the fai-cd command from FAI 5.9. After creating the squashfs, I get: > grub-mkstandalone: error: cannot make temporary directory: No such file or directory. It seems that grub-mkstandalone uses TMPDIR. Did you set this variable? -- regards Thomas
Re: Package dependency problems
apt-rdepends -r chromium-browser will list the reverse dependencies. You can set the variable MAXPACKAGES to 1 so only one package and its dependencies are installed at a time. By default we have this set in class/FAIBASE.var -- regards Thomas
Re: Update for Trusty PPA?
> On Fri, 03 Jan 2020 11:49:52 +1300, Andrew Ruthven > said: > Would it be possible to get the fai-server available for Trusty from > http://ppa.launchpad.net/fai/ppa/ubuntu updated so that at lase fai- > make-nfsroot includes the fix to run mkdir -p $NFSROOT/dev/pts in the > upgrade_nfsroot function? Hi Andrew, I do not know how maintains the PPA. I only work on the Debian versions. -- regards Thomas
Re: dracut: "No carrier detected on interface eth0", but interface later gets up
This is your kernel cmdline: Command line: BOOT_IMAGE=vmlinuz-4.19.0-6-amd64 initrd=initrd.img-4.19.0-6-amd64 ip=dhcp root=/srv/fai/nfsroot:vers=3 rootovl FAI_FLAGS=verbose,sshd,createvt,reboot FAI_CONFIG_SRC=nfs://faiserver/srv/fai/config FAI_ACTION=install rd.debug I guess the problem is root=/srv/fai/nfsroot. You hit this bug because your dhcp server is not the FAI install server. https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=731244 The fix is easy, just add the IP of your install server like this root=11.22.33.44:/srv/fai/nfsroot -- regards Thomas
Re: dracut: "No carrier detected on interface eth0", but interface later gets up
> On Sun, 5 Jan 2020 19:35:45 +0100, Markus Weich > said: > I set up a client using > fai-chboot -Iv -f verbose,sshd,createvt,reboot -u nfs://faiserver/srv/fai/config 90:1B:0E:C5:4E:48 > The client is booting fine, entering dracut. However, eth0 won't reach state UP. From rdsosreport.txt (see attachment): I guess your are booting via network card via PXE. Does this client has more than one network card? > After some time, dracut is dropping to debug shell. > However, from there, I can mount nfsroot as follows ("ip -o link show up dev eth0" now (i.e. after about 205 sec.) gives state "UP"): Maybe your network need too long to come up. Sometimes a switch can cause problems if the spinning tree algorithm takes too long. Some switches have the option fast port. There's the option rd.timeout you can use. There are more option called rd.net.timeout.*** See dracut.cmdline(7) for more info. > dracut:/# dhclient eth0 > dracut:/# mkdir -p /mnt/nfs > dracut:/# mount.nfs 172.16.7.152:/srv/fai/config /mnt/nfs > Thus, configuration seems to be fine, I guess. Yes, that looks good. > My fai-server is in a kvm running only tftp-server. dhcp-server is on another machine, setting "next-server" and "filename" for pxe-client. But this should be irrelevant, I guess. That shouldn't be a problem. -- regards Thomas
FAI 5.9 released, adding UEFI support
Hi folks, a new FAI version is available. FAI 5.9 includes following enhancements: * add UEFI support to fai-cd, fai-kvm * add disk configs for UEFI * fai-sed - call sed on a file but check for changes before writing * fai-link - create symlink idempotent UEFI support in fai-cd only used grub, no syslinux or isolinux is needed. New FAI installation images are also available from https:/fai-project.org/fai-cd The FAI.me build service is also using the newest FAI version and the customized ISO images can now be booted in an UEFI environment. Homepage: https://fai-project.org -- regards Thomas
Re: FAI-CD / EFI
> I've got a blocking problem with FAI-CD, as you know, one can't use FAI-CD for booting EFI machines. But I'm working with several developers in my company pretty much everywhere in the > world and since the new workstations work only with EFI, I can't provide a functional ISO file. I'm currently working on UEFI support for fai-cd. It's working now for me, but I still need to do some cleanup and adjust the grub.cfg. I think a new FAI version will be available until mid of january. -- regards Thomas
Re: minor inconsistencies in the example fai-config space
Hi Andrew, thanks a lot for your hints. > On Thu, 05 Mar 2020 15:10:19 -0800, andrew bezella > said: > * in `scripts/FAIBASE/10-misc` there is a modification of `/etc/fstab`: > # use tmpfs for /tmp > ainsl /etc/fstab "tmpfs /tmp tmpfs nodev,nosuid,size=50%,mode=1777 00" I have added a test. https://github.com/faiproject/fai-config/commit/c7465c21495a8d8e9d98b644ee8cfd3821e7bbab > * `scripts/UBUNTU/90-apt` contains a number of locale-related lines, I now use the hook proposed in https://lists.uni-koeln.de/pipermail/linux-fai/2016-July/011398.html https://github.com/faiproject/fai-config/commit/f2a575c7e584709d014902e2fbe9c3dee6d8d148 > * `scripts/UBUNTU/91-misc` is all gnome-related. perhaps it would be > better positioned as `scripts/GNOME/91-misc` with an `ifclass -o DEBIAN > UBUNTU` wrapper? I didn't changed this. -- regards Thomas
Re: 5.9.2 - fai-divert problem
> On Tue, 10 Mar 2020 12:53:41 -0700, andrew bezella > said: > i believe commitid 84eda68 in the 5.9.2 release introduced a bug in > fai-divert: `command` is a bash builtin and fails when used with > `chroot`. Hi Andrew, you are right, it's a bug. Luckily it does not affect services started via systemd. It seems that systemd does not consider policy-rc.d. I've added you patch and will try to do a new release soon. -- regards Thomas
Re: Remote touchless install with a static IP?
> On Thu, 1 Oct 2020 23:20:13 +, "Holland, Jake via linux-fai" > said: > I tried that iso as the CD on a local vm as a sort of "getting started" > step, but it seemed to go into an infinite loop of installing. > I tried You need to set the boot order of your VM to first local disk, then CD. Otherwise you have this infinite loop. > After a few tries, this menuentry gets me to where in my vm it looks like > it's trying to do something with the .iso I've put into the hard drive: > menuentry "jake fai" --class gnu-linux --id jake-fai4 { > loopback loop (hd1,gpt2)/faime-PR5UDQU6.iso > search --set=root --file /FAI-CD You must boot the VM from CD. Do not put the ISO onto the hard drive and try to boot from it. This will not work. > As a guess after looking through options, I tried adding this: > rd.live.overlay=/dev/sdb2:faime-PR5UDQU6.iso I'm pretty sure it cannot work this way. Let's see why your VM installation does not work. Did you try to boot this ISO? https://fai-project.org/fai-cd/faicd64-large_5.9.4.iso Does it boot and does the installation start and finishes? If not, please put the fai.log onto paste.debian.net so I can look at the errors. And maybe join the IRC channel #fai, then we can better debug further things. -- regards Thomas
Re: Fai.me support for i386 and lxqt desktop
> On Sun, 18 Oct 2020 16:41:23 +0200 (CEST), Jean-mathieu CHANTREIN > said: > Can fai.me provide iso for 32 bit processors ? Currently fai.me does not provide it, but the tools behind this service can create cross-architectur images. If you set up your own FAI install server you can build your own 32bit ISO image. > Also would it be possible to add the lxqt environment in the desktop choices ? You can easiyl create an ISO image with lxqt environment if you choose "Plain text console, no X11" and then add the package lxqt-desktop in the web form. -- regards Thomas
RE: Remote touchless install with a static IP?
Hi > On Wed, 30 Sep 2020 03:20:03 +, "Holland, Jake via linux-fai" said: > But it looks like the FAI docs really want me to use dhcp, and I > cannot do this thing. I don't have control of the environment there > where the machine is located. If that's truly a requirement for FAI, > I'll have to find another way. Hi Jake, you can build a FAI CD (or USB stick) which does not need any internet connection. Have you already tried the FAIme service at https://fai-project.org/FAIme/ ? If you remove rd.neednet from the grub cmdline, the system will not try to use dhcp. You also set an IP via ip=... > I get the impression it's possible to do this from the grub menuentry, > but I haven't figured out how yet (Perhaps export? But I'm not sure, > this seems a different use than the examples the grub docs give.) During the FAI boot process we use dracut, which supports setting a static IP. Have a look at the dracut.cmdline manual https://manpages.debian.org/testing/dracut-core/dracut.cmdline.7.en.html -- regards Thomas
Re: Autodiscover.iso - No servers found
Hi Jules, Are the network parameters shown by the autodiscover ISO correct? It executes this line: echo "Scanning $iprange for FAI server (port $FAI_MONITOR_PORT)" Please check the IP range and the port. Then the autodiscover scripts do a nmap to get the list of possible FAI servers. You can also check if the fai-monitor is problery running from another host. echo "VARIABLE FAI_VERSION" | nc ip_of_server 4711 This should output the FAI version number. echo "VARIABLE NFSROOT" | nc ip_of_server 4711 Should output the path of the nfsroot. echo "VARIABLE FAI_CONFIG_SRC" | nc ip_of_server 4711 Should output the URL for the config space. -- regards Thomas
Re: Fai.me support for i386 and lxqt desktop
> On Wed, 21 Oct 2020 19:35:11 +0200 (CEST), Jean-mathieu CHANTREIN > said: > I tried in plain text consol (no x11) mode and just putting the package task-lxqt-desktop, unfortunately there is a configuration problem (no icon is displayed in the interface). Mmm, I do not know what the reason is. Maybe the order of installing the packages is different. > I tried again by creating the iso myself on my own FAI server (very great > feature by the way), but it didn't change anything. > It's curious, because: > - if i just take a netinstall and install, after basic install, the task-lxqt-desktop package by hand, it works fine. > - if I start from an iso of fai.me, without additional package and I install, after basic installation, the package task-lxqt-desktop by hand, that also works well Did you already compared the list of packages that are installed at the end of the different installations? -- regards Thomas
Re: FAI with pre-release bullseye yielding inconsistent network names
> On Tue, 19 Jan 2021 06:57:26 -0800, fai1...@macrotex.net said: > I am already using that patch. The problem seems to be that the altname > returned by "ip link show eth0" in 30-interface is "enp11s0" and hence > that is the filename what 30-network puts into > /etc/network/interfaces.d, but when the new system boots the name in > /sys/class/net is "ens192". > Why is there this mismatch? Is it an issue with bullseye itself? udeadm info sometimes show ID_NET_NAME_SLOT and ID_NET_NAME_PATH. On my bullseye kvm machine is shows E: ID_NET_NAME_PATH=enp0s3 E: ID_NET_NAME_SLOT=ens3 In the 30-interface script we loop over all fields fields="ID_NET_NAME_FROM_DATABASE ID_NET_NAME_ONBOARD ID_NET_NAME_SLOT ID_NET_NAME_PATH" In /lib/systemd/network/99-default.link there's the order of names defined: NamePolicy=keep kernel database onboard slot path It's correct to use slot before path. I guess in your case you should add udevadm trigger, so FAI does not need the fallback to altname from ip link show which seem to have only the path name. -- regards Thomas
Re: RAID + UEFI
Hi Alexander, my short search found some links that all say, UEFI cannot boot software AID. This makes much sense to me. The solution seems to have a /boot/efi partitions on each disk and have two UEFI boot entries. https://askubuntu.com/questions/66637/can-the-efi-system-partition-be-raided https://outflux.net/blog/archives/2018/04/19/uefi-booting-and-raid1/ -- regards Thomas
Re: RAID + UEFI
> On Thu, 21 Jan 2021 18:13:25 +0100, Alexander Thomas > said: > I would prefer the second method because it is not a hack. I have > successfully tried it with only 1 disk in the RAID during install, but > with 2 disks there is a problem. The disk_config for this 2-disk case > currently looks like: > disk_config disk1 disklabel:gpt bootable:1 > primary /boot/efi 512M vfat rw > primary - 512M - - > primary - 1G-4G- - > primary - 30G-36G - - > primary - 4G- - - > disk_config disk2 sameas:disk1 bootable:1 > Obvious problem: the `sameas` also clones the /boot/efi mount point, > and install fails with “mount point /boot/efi used twice”. > Is there an easy way around this, like some mountdisks hook script, or > is the ESP-inside-RAID a better option after all? The only workaround I know is not to use samas, but to copy the config from disk1 to disk2 and then remove the mount point from disk2. -- regards Thomas
Re: FAI with pre-release bullseye yielding inconsistent network names
> On Fri, 15 Jan 2021 09:38:41 -0800, fai1...@macrotex.net said: > I do the initial FAI imaging using a buster NFS root and install buster > in /target. When I then run "chroot /target udevadm info > /sys/class/net/eth0" I see about 27 lines including a bunch of > ID_NET_NAME_* values. > I do the above again, but this time with a bullseye NFS root and install > bullseye in /target. When I then run "chroot /target udevadm info > /sys/class/net/eth0" I see only about 6 lines and, in particular, no > ID_NET_NAME_* values. I booted a KVM machine with a bullseye nfsroot. I see 6 lines when FAI_ACTION=install but 27 lines when FAI_ACTION=inventory. Very strange. I found out that "udevadm trigger" is executed when doing the inventory but not when doing the installation. I will think about where to put the missing udevadm trigger call. -- regards Thomas
Re: FAI with pre-release bullseye yielding inconsistent network names
> On Tue, 19 Jan 2021 21:00:28 +0100, Stefan Möding > said: > It seems the path based name is used when booting the NFSROOT while the slot > based name is used when the machine is rebooted after installation. To me > that looks like a problem with the NFSROOT like missing kernel modules or > libraries. No, the only thing that is missing currently in FAI 5.9.4 is the call to "udevadm trigger". If this is added, then udevadm info will show all needed information. This works in my environment will bullseye in a VM. I didn't had the chance yet to test real hardware. -- regards Thomas
Re: FAI with pre-release bullseye yielding inconsistent network names
>>>>> On Mon, 18 Jan 2021 18:23:05 +0100, Thomas Lange >>>>> said: > I found out that "udevadm trigger" is executed when doing the > inventory but not when doing the installation. I found out, that I've added a patch after the release of 5.9.4 which tries to get the interface name if udevadm info does not show the name. It's this change: https://github.com/faiproject/fai-config/commit/45a802ca5f0a7cc229edb4cedd1b22ab4f790ba4 Now we have two solutions for this problem. Add the patch above to your config space or add a call to udevadm trigger for e.g. in the beginning of scripts/DEBIAN/30-interface. -- regards Thomas
Re: FAI with pre-release bullseye yielding inconsistent network names
> On Thu, 14 Jan 2021 12:19:30 -0800, fai1...@macrotex.net said: > I am attempting to build Debian "bullseye/sid" (pre-release bullseye) > using FAI 5.9.4. The network interface names come out being inconsistet, Maybe you are missing this patch which is not in 5.9.4 included: https://github.com/faiproject/fai-config/commit/45a802ca5f0a7cc229edb4cedd1b22ab4f790ba4 -- regards Thomas
Re: Remove a defined class
> On Thu, 10 Dec 2020 15:11:14 +, Markus Rexhepi-Lindberg > said: > Is it possible to remove a defined class during a install? Uuuuh. It may be possible but in 2006 we already had this questions https://lists.uni-koeln.de/pipermail/linux-fai/2006-November/004613.html The clean way would be to write a hook with suffix .sh (then it is sourced). This hook must remove the class from $LOGDIR/FAI_CLASSES and also redefine the shell variable $classes. We use this command in FAI for setting $classes classes=$(< $LOGDIR/FAI_CLASSES) -- regards Thomas
Re: Bootable FAI installer USB stick without ISOHybrid?
> On Mon, 1 Feb 2021 18:08:22 +0100, Alexander Thomas > said: > In short: would it be possible to make a FAI installer boot and run > from a USB disk that has a normal MBR partition table, instead of the > ISOHybrid hocus-pocus? Sure. FAI does not need an ISO at all. We can already boot from network and it will be possible to create a boot device using a different partition layout. > contains both a FAI installer and also a partition that is easily > accessible and writeable under Windows. It must also be possible to > add or at least expand that partition using Windows tools without > breaking the ability to boot from it. Do you want to have one or two partitions the the USB stick? Since we nee another partition for UEFI maybe you will have three partitions. Which filesystems are writable by Windows? I guess an ext3/4 partition is not writeable by default. > We don't need the ability to boot from a CD-ROM drive and we don't > need to support all kinds of platforms like Macs, we only care about > booting a typical modern PC architecture server from USB and running a > standalone FAI installer from it. Although legacy boot support with > the same disk image would be nice, it would be OK if it only works > with UEFI. For UEFI we need the /boot/efi partition which must be a vfat filesystem. This contains the boot loader. MBR does not need this partition. Then we need a partition for FAI which contains the kernel and initrd and the nfsroot (not sure if we could use the UEFI partition. This can be a plain filesystem on a seperate partition or a squashfs (maybe on a vfat). > The biggest questions are > probably how to do it with minimal changes, and also whether there's > anything inside FAI that assumes an ISOHybrid environment and would > break when booted differently. As I said before, ISOHybrid is not needed. Using fai-cd -S it will only create a squashfs filesystem. This is normally booted via network, but maybe a start to create a bootable USB stick. Currently I have no time to work on this, but I'm happy to help you. -- regards Thomas
Re: Bootable FAI installer USB stick without ISOHybrid?
Another hint for creating the USB stick: https://grml.org/grml2usb/ They are not using dracut, but I guess a similar partition layout should work with dracut. -- viele Grüße Thomas
Re: Partition type of USB live image is now "empty"
Hi Alexander, I've checked several other linux ISO images (ubuntu,grml, Centos, Debian, archlinux) and only archlinux does not use type 0 for the first partition on their ISOs. But archlinux set a flag called hidden. Fedora use type 0 but with parted I canot set the partition, cfdisk shows the partition. So, there's some magic in all this. In fai-cd we call xorriso with -iso_mbr_part_type 00. I do not like to change this without knowing it will not cause other problems. Can you find some information that convice me that a different type is correct? -- regards Thomas
Re: Bootable FAI installer USB stick without ISOHybrid?
>>>>> On Wed, 3 Feb 2021 15:25:44 +0100, Alexander Thomas >>>>> said: > On Mon, 1 Feb 2021 at 20:26, Thomas Lange wrote: > We have already started to try a few things and the main problem now > seems to be to get dracut to play nice with an installer medium > formatted as FAT32 or NTFS. It lacks these modules and so far we > haven't been able to provide them. If you have any pointers as to how > to tackle this, that would be great. There is some ntfs support, see below. Otherwise ask on the dracut IRC channel. The relevant man page for booting with dracut is this https://man7.org/linux/man-pages/man7/dracut.cmdline.7.html Ther's a dracut module called 90dmsquash-live-ntfs which uses ntfs-3g. I couldn't find more info, but maybe this would provide dracut access to ntfs filesystems. Another option is iso-scan/filename, which scans for an ISO file. Or maybe rd.live.squashimg= In the end we want to boot a kernel and initrd which is on a vfat or ntfs filessytem and then boot the FAI live system which can be an ISO or a squashfs. This may also help. https://fedoraproject.org/wiki/Livecd-iso-to-disk -- regards Thomas
Re: Bootable FAI installer USB stick without ISOHybrid?
Another hint: https://access.redhat.com/discussions/5706091 I guess you can add vfat support easily into dracuts initrd. Then set these kernel commandline paramters to point dracut to the squashfs file which should also be accessible insode a vfat fielsystem. rd.live.dir= rd.live.squashimg= -- regards Thomas
Re: Weird generated hostname
> On Fri, 18 Jun 2021 08:37:52 +0200, Andreas Heinlein > said: > Now it is '127-0-0-1/8 192-168-10-123'. Yes, including the slash and the space. Didd you upgrade to a newer FAI version? In FAI 5.10 we changed some IP detection in get-boot-info. Maybe this leads to this weird hostname. There's a fix in FAI 5.10.3 which may help you: https://github.com/faiproject/fai/commit/e85717d2efc94e221c16502c30fdc9fda769aa27 -- viele Grüße Thomas
Re: unshare causes breakage in docker
> On Tue, 27 Apr 2021 21:05:37 +0100, Justin Cattle said: > We build FAI images in docker. Have a look at https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=989547 There's also a workaround how to disable the use of unshare. -- viele Grüße Thomas
Re: Grub install fails because of missing files
> On Mon, 10 May 2021 16:28:56 +, "Moser, Marcus" > said: > I want to install CentOS 7 with FAI, but the grub install always fails. In the log file was the following massage: > „grub2-install: error: cannot find EFI directory.“ I seems that you want to do a CentOS installation using UEFI bios. Is that right? I guess CentOS with UEFI is not tested or you are using a wrong partition layout. Please provide more info from fai.log. -- viele Grüße Thomas
FAI 5.10.3, feature added to FAIme service
Hi all, I've released FAI 5.10.3, including a minor fix which only affects raid setup in virtual environments. See #988987 for more info. The FAIme service was extended. I've added the feature to upload a custom shell script, which is added to the installation ISO and gets executed during boot on request. -- regards Thomas
Re: AW: fai-diskimage variable "classes" not inherited to fai
> On Thu, 1 Jul 2021 09:24:37 +, "Schulz, Reiner" > said: > # fai-diskimage -v -N -u dvzsn-ra0317 image.raw > ERROR: No classes are defined >> From /usr/sbin/fai-diskimage: > 182 if [ -z "$classes" ]; then > 183 die 7 "No classes are defined" > 184 fi Ooops. You can work around this y definin a class, but not use it in your config space. Just add -c Z. -- viele Grüße Thomas
Re: fai-diskimage variable "classes" not inherited to fai
> On Wed, 30 Jun 2021 08:09:43 +, "Schulz, Reiner" > said: > How is "-c" concatenate to the classes from config_space/class ? > We set a couple of default classes (class/49-always_set_classes) and add only the invidual, server specific, classes via class/50-host-classes > # fai-diskimage -v -N -cDEBIAN_10,S2APP image.raw I guess the man page of fai(8) explains it better, since -c and -N from fai-diskimage are passed to the fai call itself. -N executes the scripts in the class/ subdirectory. -c just sets the variable $cmdlineclasses which is then used by class/01-classes. So if you do not have the script class/01-classes the classes from -c are not used. If you rename 01-classes to 99-classes you can change the order of the classes defined by -c. BTW, I never use -c and -N at the same time. -- viele Grüße Thomas
Re: VMs end up without grub installed
> On Wed, 21 Apr 2021 00:12:02 -0700, Bill MacAllister > said: > What I am seeing is the that everything on the build goes fine until > GRUP_PC/10_setup. 10_setup exists with error status of 1. The > resulting system does not boot, well, boots to a blank screen. Looking > at the system I see that /boot/grub/ does not contain a grub.cfg. You should add set -x info the GRUB_PC/10-setup script. Then you will see more information in shell.log -- viele Grüße Thomas
Re: unshare causes breakage in docker
> On Tue, 27 Apr 2021 21:05:37 +0100, Justin Cattle said: > Can we somehow make unshare vs chroot an option ? Or, even better perhaps, detect docker > and don't use unshare in that case ? Do you know how to detect that the process is running inside docker? -- viele Grüße Thomas
Re: CD, NFSROOT and expiring DHCP lease
> On Thu, 29 Apr 2021 21:21:06 +1200, Andrew Ruthven > said: > Dracut acts as a DHCP client, but is a single shot only. Once the > lease expires nothing on the machine building rebuilt renews > it. Hi Andrew, you are right, in FAI the kernel (or maybe the initrd) configures the IP addresses and sets valid_lft and preferred_lft to forever. Then a one-shot dhcp request is done, but with a different config and dhcp script to write the file boot.log. FAI does not start the dhclient to automatic renew the ip leases. This is a bug. > Is there an approach so that ISO images created with fai-cd perform > DHCP lease renewals? As a workaround, you could write a hook, that starts dhclient. Since the hooks are on the ISO and will be executed, this should work until I release a new FAI version with a fix. It would be nice if you could file a bug report for this. -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
> On Mon, 8 Feb 2021 07:24:33 +0100, Carsten Aulbert > said: > grub-pc/grub2 in buster's 10.8 point release "sacrifices" > non-interactive upgrade-ability[1]. :-( I also have some machines that are stuck because of this. I wonder why not all machines are affected. -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
In a KVM VM, we cannot detect the disk by looking into /dev/disk/by-id :-( We need a fallback for this environment. Maye we then just use /dev/vda? buster[~]$ ls -l /dev/disk/* /dev/disk/by-id: total 0 lrwxrwxrwx 1 root root 9 Feb 2 11:32 ata-QEMU_DVD-ROM_QM3 -> ../../sr0 /dev/disk/by-partuuid: total 0 lrwxrwxrwx 1 root root 10 Feb 2 11:32 bfea682b-01 -> ../../vda1 /dev/disk/by-path: total 0 lrwxrwxrwx 1 root root 9 Feb 2 11:32 pci-:00:01.1-ata-2 -> ../../sr0 lrwxrwxrwx 1 root root 9 Feb 2 11:32 pci-:00:05.0 -> ../../vda lrwxrwxrwx 1 root root 10 Feb 2 11:32 pci-:00:05.0-part1 -> ../../vda1 lrwxrwxrwx 1 root root 9 Feb 2 11:32 virtio-pci-:00:05.0 -> ../../vda lrwxrwxrwx 1 root root 10 Feb 2 11:32 virtio-pci-:00:05.0-part1 -> ../../vda1 /dev/disk/by-uuid: total 0 lrwxrwxrwx 1 root root 10 Feb 2 11:32 81de4336-a35e-4615-b5e3-8d3c6dfb73b7 -> ../../vda1 -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
I've updated your script. I now scans MBR, skips cdrom devices, and checks devices in a VM, which does not use /dev/disk/by-id. I'm not sure if still some edge cases are not covered. I'm only putting it here for a day, since I guess it will get udpates soon http://paste.debian.net/1184540/ -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
I've created a patch for scripts/GRUB_PC/10-setup https://github.com/faiproject/fai-config/commit/bf90f3048f552f2dc1a0f50766646dff9f67aef9 This should fix the issue for new installations. Mmm, I forgot the VM case. So this does not work yet on virtual machines. -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
A simple fallback was added, so it now works in a VM. The newest version of the grub pc setup script is https://github.com/faiproject/fai-config/blob/master/scripts/GRUB_PC/10-setup I'm very happy to get feedback -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
I like to summarized the grub-pc problem the new point release 10.8 introduced to FAI users. If you have installed a computer with grub-pc (grub EFI is not affected), FAI installed grub into the MBR of the boot disk(s), but it did not wrote the debconf variable grub-pc/install_devices. After the last point release the grub-pc package complains about this and you have to fix it interactively, so no unattended installations are possible any more. There are two fixes, one for new installation, one how to fix exisiting installation. If you want to install a new machine from scratch with FAI just get the new version of the GRUB_PC/10-setup script. https://raw.githubusercontent.com/faiproject/fai-config/master/scripts/GRUB_PC/10-setup To fix exisiting FAI installation you can use the script grub_debconf.sh from https://fai-project.org/download/misc/ This will detect which device(s) hold a grub MBR and add the list of devices to the debconf variable like this: * grub-pc/install_devices: /dev/disk/by-id/ata-INTEL_SSDSC2CW120A3_CVCV249103KG120BGN You can check if this variable is set by $ debconf-show grub-pc| grep install_devices: Thanks to Carsten for seeing this problem and writing the first version of the script. -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
> On Wed, 10 Feb 2021 12:13:28 +0100, Stefan Möding > said: > mbrdevices=$($ROOTCMD udevadm info -r --query=symlink $BOOT_DEVICE | grep -Po '/dev/disk/by-id/ata\S+') > will emit an exit status 1 if it does not find a match. This is > propagated as the script error status (due to the trap). Maybe an > additional "| cat" at the end could be used to gobble the status of > "grep"... Thanks for the hint. It's now fixed. -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
Ooh, it seems more complicated. Your code does not work on a machine with one SATA disk: $ udevadm info -r --query=symlink /dev/sda /dev/disk/by-id/wwn-0x5001517803d00e48 /dev/disk/by-id/ata-INTEL_SSDSC2CW120A3_CVCV249103KG120BGN /dev/disk/by-path/pci-:00:1f.2-ata-1 $ udevadm info -r --query=symlink /dev/sda | sed -n 's;.*\(/dev/disk/by-id/\S\+\).*;\1;p' /dev/disk/by-id/wwn-0x5001517803d00e48 Before changing the code, I like to know if there are more unknows cases we should keep an eye on. -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
> On Wed, 10 Feb 2021 19:53:39 +0100, Stefan Möding > said: > Oh... So "sed" might not be a good idea if you want all entries. I don't want all entries, I just want the one that's better readable. > # udevadm info -r --query=symlink /dev/sda > /dev/disk/by-path/pci-:00:10.0-scsi-0:0:0:0 I think we can use the fallback to /dev/sda for such a case. I use a similar name (/dev/vda) on my KVM machines. I think we want to match all real disks which can be identified by a name, model, serial number. Maybe also virtualized disks which have such an entry at by-id. -- viele Grüße Thomas
Re: Some troubles with grub with buster point release 10.8
> On Thu, 11 Feb 2021 08:23:07 +0100, Carsten Aulbert > said: > I assume udevadm would return any of the valid device links and by-path > is always set, thus: > get_stable_devname() { > _DEV="$1" This does not work, because the order of links returned by udevadm changes with every call. [~]$ udevadm info -r --query=symlink /dev/sda /dev/disk/by-path/pci-:01:00.0-sas-exp0x500304800096b4bf-phy17-lun-0 /dev/disk/by-id/wwn-0x50014ee35570a5e8 /dev/disk/by-id/ata-WDC_WD3000FYYZ-01UL1B0_WD-WMC1F0499068 [~]$ udevadm info -r --query=symlink /dev/sda /dev/disk/by-id/wwn-0x50014ee35570a5e8 /dev/disk/by-path/pci-:01:00.0-sas-exp0x500304800096b4bf-phy17-lun-0 /dev/disk/by-id/ata-WDC_WD3000FYYZ-01UL1B0_WD-WMC1F0499068 [~]$ udevadm info -r --query=symlink /dev/sda /dev/disk/by-path/pci-:01:00.0-sas-exp0x500304800096b4bf-phy17-lun-0 /dev/disk/by-id/ata-WDC_WD3000FYYZ-01UL1B0_WD-WMC1F0499068 /dev/disk/by-id/wwn-0x50014ee35570a5e8 [~]$ udevadm info -r --query=symlink /dev/sda /dev/disk/by-id/wwn-0x50014ee35570a5e8 /dev/disk/by-id/ata-WDC_WD3000FYYZ-01UL1B0_WD-WMC1F0499068 /dev/disk/by-path/pci-:01:00.0-sas-exp0x500304800096b4bf-phy17-lun-0 [~]$ udevadm info -r --query=symlink /dev/sda /dev/disk/by-path/pci-:01:00.0-sas-exp0x500304800096b4bf-phy17-lun-0 /dev/disk/by-id/ata-WDC_WD3000FYYZ-01UL1B0_WD-WMC1F0499068 /dev/disk/by-id/wwn-0x50014ee35570a5e8 [~]$ udevadm info -r --query=symlink /dev/sda /dev/disk/by-path/pci-:01:00.0-sas-exp0x500304800096b4bf-phy17-lun-0 /dev/disk/by-id/wwn-0x50014ee35570a5e8 /dev/disk/by-id/ata-WDC_WD3000FYYZ-01UL1B0_WD-WMC1F0499068 [~]$ udevadm info -r --query=symlink /dev/sda /dev/disk/by-id/ata-WDC_WD3000FYYZ-01UL1B0_WD-WMC1F0499068 /dev/disk/by-path/pci-:01:00.0-sas-exp0x500304800096b4bf-phy17-lun-0 /dev/disk/by-id/wwn-0x50014ee35570a5e8 [~]$ udevadm info -r --query=symlink /dev/sda /dev/disk/by-path/pci-:01:00.0-sas-exp0x500304800096b4bf-phy17-lun-0 /dev/disk/by-id/ata-WDC_WD3000FYYZ-01UL1B0_WD-WMC1F0499068 /dev/disk/by-id/wwn-0x50014ee35570a5e8 [~]$ udevadm info -r --query=symlink /dev/sda /dev/disk/by-id/ata-WDC_WD3000FYYZ-01UL1B0_WD-WMC1F0499068 /dev/disk/by-path/pci-:01:00.0-sas-exp0x500304800096b4bf-phy17-lun-0 /dev/disk/by-id/wwn-0x50014ee35570a5e8 -- viele Grüße Thomas
Re: Parmeter DEFCONSOLE=ttyS0,115200n8 / fai.me
> On Wed, 27 Jan 2021 21:13:02 +0100, Jürgen Fricke > said: > Hi folks, is it possible to set the parameter DEFCONSOLE=ttyS0,115200n8 > via grub options? As a beginner I'm trying to get SVGA etc to work with > FAI on a small target system without display, only ttyS0 is possible. Do > I maybe need to set up a regular FAI install server or not? I do not know this DEFCONSOLE parameter. Is this something special? Most people use console=ttyS0,115200n8 If you enter this to the grub option in the FAIme web page, it will get added as a debconf question in this way grub-pc grub2/linux_cmdline_default string x This should enable the console on the serial port. -- regards Thomas
Re: fai-setup: ECDSA-key missing in known_hosts
> On Fri, 22 Oct 2021 23:21:07 +0200, Christian Meyer said: > This file is created by /usr/sbin/fai-setup and I found that the rsa- > key and the ed25519-key are recognised and handled well (except some > obfuscation), but the ecdsa-key still is missing. You are right, fai-setup has no code for copying the ecdsa-key. This is a bug, it's not made on purpose. -- viele Grüße Thomas
Re: Newbie question about image generation
Hi Carlos, that is strange. I see this in my logs: Copying the nfsroot to CD image Copying the config space to CD image Parallel mksquashfs: Using 2 processors . . . Number of gids 1 root (0) mkfs.fat 4.2 (2021-01-31) Writing FAI CD-ROM image to /home/lange/a.iso. This may need some time. xorriso 1.5.2 : RockRidge filesystem manipulator, libburnia project. ISO image size and filename: 338M /home/lange/a.iso So it seems that some commands are not exexcuted. Please call: bash -xv fai-cd so we can debug why fai-cd stops without creating an ISO file. > 3. Generating Image > fai-cd -C /root/fai-bullseye -g $PWD/grub.cfg.install-only -m/home/mirror /home/fai.iso > All commands are executed flawlessly, but the ISO is not generated. > Complete log: https://controlc.com/19c9cc6e -- viele Grüße Thomas
Re: Newbie question about image generation
> The complete log was stored in: https://controlc.com/2cfb50fd + dd if=/dev/zero of=/tmp/scratch/efiboot.img bs=1M count=3 cleanup_liveos_mounts + cleanup_liveos_mounts This is the part where it failes. I guess there's no disk space left in /tmp/scratch when creating this small image. -- viele Grüße Thomas
Re: updates to fai-cd
> On Fri, 23 Jul 2021 11:12:18 +0200, Sinh Lam > said: > Some of these options are not valid in the version of xorriso that’s available in 16.04 (or 20.04 for that matter). I’ve even tried looking for this version on a Debian server and they > aren’t there. Can you please provide some error output. Which options are not supported? On when Debian version did you also had these errors? Which xorriso version on Debian? -- viele Grüße Thomas
Re: no operating system found
Hi Matthias, auf jeden Fall mal die Log Dateien der FAI Installion ansehen. Irgendetwas ist das nicht richtig gelaufen. Ist das ein UEFI System? > On Tue, 15 Mar 2022 09:05:21 +, "Weber, Matthias - LASUB" > said: > Hallo, > ich habe FAI Ubuntu 20.04 LTS auf einem Lenovo ThinkCentre M72e installiert. > Beim 1. Neustart nach der Installation erscheint "error: file boot grub fonts unicode.pf2 not found", was aber nicht schlimm scheint. > Beim 2. Neustart aber kommt "error 1962: no operating system found" und Ubuntu wird nicht gebootet. > Was kann ich tun? > -- > Vielen Dank und beste Grüße > Matthias -- viele Grüße Thomas
booting FAI without PXE, tftp using kexec
Hi all, I just installed a machine in a different subnet where I could no use network booting using PXE. But on this machine linux was already running. This is how I managed to boot the FAI system and to start the installation. Load the FAI kernel and initrd from the nfsroot and add FAI specific boot parameters: # kexec -l vmlinuz-5.10.0-13-amd64 --initrd=initrd.img-5.10.0-13-amd64 --append="FAI_FLAGS=sshd,verbose,createvt FAI_ACTION=install rootovl rd.neednet ip=dhcp root=1.2.3.4:/srv/fai/nfsroot,nfsvers=3 FAI_CONFIG_SRC=nfs://faiserver/srv/fai/config" Then reboot the computer using kexec. This will shutdown the OS and boot the FAI kernel and initrd without going through the BIOS. # systemctl kexec -- best regards Thomas
Re: LUKS, "no device found"
> On Mon, 24 Jan 2022 10:31:19 +0100, Juri Grabowski said: > I was looking to my configuration right now and it looks like yours, but with "disk_config end" on the end. "disk_config end" ? I never used this, but indeed it's in the man page. Something new for me ;-) -- viele Grüße Thomas
Re: LUKS, "no device found"
> On Tue, 25 Jan 2022 16:06:52 -0800, Ian Eure said: > Is it trying to tear down the existent (broken) disk structure > instead of overwriting it? Otherwise, I don’t understand what If you like to overwrite all old lvm softraid stsructures, you can use a hook for that. I use this script https://fai-project.org/download/misc/wipe-all > Sorry for the dumb question, but where would I add the "-d" flag? > I’m creating a CD image and booting it on the target machine using > Ventoy on a USB stick. You have to edit the fai source code (lib/subroutines) which calls setup-storage. -- viele Grüße Thomas
Re: booting FAI without PXE, tftp using kexec
> On Wed, 6 Apr 2022 09:37:08 +0100, Justin Cattle said: > Hi Thomas, > That's very cool. Thanks for sharing. > I guess you are copying the kernel and initrd over manually and putting them in your current working > dir ? Yes, excatly. -- viele Grüße Thomas
Re: Squashfs Image No space left on device
> On Thu, 19 May 2022 15:11:24 +0200, Karsten Nordsiek > said: > Hello, > on one FAI-Server (we are using multiple FAI-Server) I got an error > message no space left on device. fai-cd uses /tmp for creating temporary files which may big a huge amount. Check if this partition has enough free space. -- viele Grüße Thomas
Re: dhcp issue...
Hi Andreas, you should check if the dhcp request from the client (which are technical bootpc packets are received on the dhcp server. Does the dhcp server have more than one network card? Then check /etc/default/isc-dhcp-server if the dhcp server is listening on all interfaces. # tcpdump -pnlqte udp and src port bootpc and broadcast You should also see some messages in the log of the dhcp server, for e.g. if the server receives the request but has no free IP addresses. In addition to tcpdump there's also dhcpdump for debugging. -- best regards Thomas
Re: Ubuntu-22.04
> On Wed, 20 Apr 2022 09:40:17 +0200, Stefan Möding said: > Ubuntu-22.04 now uses the zstd(1) compression program for packages. FAI > can't install these packages out of the box because debootstrap on > Debian-11 doesn't know about this compression format. Here are the things > I changed. I would just install debootstrap >= 1.0124 on my machine, because this version added zstd support. Normally it's easy to install a newer version of debootstrap because it has no special dependencies. But I didn't tested this yet. -- regards Thomas