Hi! For such situations we designed a RunOnce.Service: It runs all scripts from a hook-dir at boottime and then move it away. Or you can check the connectivity of a specific IP-address and then run the scripts.
And sorry, im not allowed to share our inhouse knowledge Reiner Von: linux-fai [mailto:linux-fai-boun...@uni-koeln.de] Im Auftrag von Pasquale Cantiello Gesendet: Donnerstag, 25. Februar 2021 11:14 Cc: fully automatic installation for Linux <linux-fai@uni-koeln.de> Betreff: Re: Driver replacement Dear John, thanks for your answer. Putting a script in the /srv/fai/config/scripts/ is the way I was trying but, if the system boots with the nouveau driver already installed, the sequence to install correctly the new one is: - blacklist the nouveau - reboot - install the proper nvidia driver - reboot again - continue with software installation And this sequence, I think, cannot be fully automated with those scripts. So I have been searching for a way to prevent the nouveau installation since the beginning. I will study the fai-scripts in order to better understand how this happens. Best regards Pasquale --- ing. PhD Pasquale Cantiello Istituto Nazionale di Geofisica e Vulcanologia Sezione di Napoli - Osservatorio Vesuviano Via Diocleziano, 328 - 80124 Napoli Tel. 081-6108.332 Il giorno mer 24 feb 2021 alle ore 14:15 John G Heim <jh...@math.wisc.edu<mailto:jh...@math.wisc.edu>> ha scritto: FAI works a lot like an installation you'd do "by hand". You boot from a live image, partition the hard drive, install a base system via debootstrap then chroot to do the rest of the installation steps (install a kernel, install packages, configure grub). So, if you want to blacklist a module, you have to do it on the new system, not on the live boot image. In other words, in your case, you'd have to fcopy the module configuration files to the new system after the base system is installed. This is done by putting a script in the scripts folder of your FAI config root. By default this would be /srv/fai/config/scripts/. On 2/24/21 4:20 AM, Pasquale Cantiello wrote: > Hi all > > I'm new to fai (only two weeks) and trying to use it to install a > cluster with a front-end and 10 compute nodes. At present stage I am > able to install all machines with related software and little > customization from the faiserver with almost no problem. > > Now, since the nodes have a GPU card I want to install the nvidia driver > and related sdk to the frontend and all nodes. I've downloaded all the > files (.run file for driver and .deb files for sdk) on faiserver and put > them in nfsroot in order to launch the installation on the nodes. > > My problems are now: > How to prevent the installation of the nouveau driver on nodes? This > will conflict with the new driver. I've also put a modeprobe.conf file > with blacklist in /nfsroot/etc/modeprobe.d/ but with no success. > > And, how to use the PRELOAD or PRELOADRM command in the package > configuration to install sdk from nfsroot? I've found no samples and > tried all combinations... > > Thanks in advance to all who can help me. > > Best regards > Pasquale > > --- > ing. PhD Pasquale Cantiello > Istituto Nazionale di Geofisica e Vulcanologia > Sezione di Napoli - Osservatorio Vesuviano > Via Diocleziano, 328 - 80124 Napoli > Tel. 081-6108.332