Half a year ago, when I was exhausted in a Doc mace, I found myself lost between interfaces like that virt-install,qemu ,virtmanager,virsh ,presseed, boot prompt. Now after analysing the situation I can present this script as a tutorial and program https://github.com/qemu-buro-point-dpkg/qemu-buro-point-dpkg/blob/master/Testsscripts/qemuburodpkg/preseediso2ssh.sh
You find everything self-explanatory, including pointers to relevant howtos and even usable. Things one needed to understand where:
outside the scope ___remaster the ISO image ___developers‘ documentation for debian-installer
Note that it makes ___no sense when using __file or __network preseeding to preseed values for questions__(Which??) that are asked ___before the preconfiguration file is parsed.
There is one_ universal method for install preseeding: hacking the iso. So you hack the iso file, then you hack the initrd file in it, there you put the preseed file. You find the 71 Scripts documented (described prettily) in the internals doc. That is where you get into touch with it, understanding the execution flow and debug logging. Every other preseed interface (such as file or network) are limited, for not covering complete non-interactivity, to say for being useless, meanwhile: initrd injection can be used for any use case I can imagine. Lets say: You are invited in a house of a friend, they hand you out the CD. Only in case you really, really do not find a free software internet computer for manipulating you CD on the fly and by the way you prepare the preseed file anyway, you will opt learning the file interface. Is it functioning? – In case I had a server farm, opted pxe, which is bios, me? Never.
Hey, one is really comfortable with. One can even trace or track the preseed commands, to find out what they are. And even you are even tempted to by pass them. Especially, if for some magic, they leaved you helpless. http://stackoverflow.com/questions/33113363/debian-preseed-late-command-not-executed
Problems like these?
So what is the preseed technology then in a few words? Debian installer is a program, which offers a script interface. This script is read at – the beginning – and: there is a data base, you know, called debconf. And for my superficial get-into-touch with it it is messy as hell, but it is smart enough the installer itself can read it. So if you got a working configuration, shift it in, but without shifting a bit in it. Right? Then, well, after your preseed is registered in data base, in the course of installation it is worked out by the namely 70 scripts, among so called „categories“, „components“, „templates“ from fort nox „ramdisk“ to „target“. I did not get to early logs, but I get to late logs as noisy as necessary in the end.
There are still things to find out, fine grain monitor the process and fine grain controlling it: issuing a 7 seconds lasting installation step lets say 7, between 6 and 8. Do not know, if that is already implemented, among the present methods. This is to find out. I did not came to any insights in #debian-boot chat (; , Maybe if they see, that someone is actually checking out d-i source and compiling, people get more helpful recently. That is what my bet is. Even though there is not a feature like this in the end, the information were valuable, because then one could file/communicate a feature request for it. Fine grained process steps are needed, to fit into a unit test frame work, for development of free technology to be eased sufficiently. To say to lower the hurdle enough. Says I.
Yes, and what happened with virt-install? Is a part of the virtlib garden. For command line use they, redhead, developed a xml format, while still relying on VMs like qemu as a backend or frontend. Here is where you get into mace. One gets the impression that effects take place where things are uneased on purpose to sell them later better, expertism that drives itself important, in a scale – that someone likes to express that, here. To say to get into the – uncomfortable – situation of needing to criticize that as a by passer. Might be that things in modern industry got down to that normal level, I severely disapprove it. It is about betraying the humans where they might not and should not expect it. Well this is the tale about the holy working places behind it and the success behind it, that is a lie. With it I do not say, that there were not gradual trade offs in business practices rather than a subjective good-evil borderline, neither I should deem the problem to solved by superficial anti-discrimination, that is strictly maintaining people on earth without money at the hour of saying no to business whatsoever, that is without dignity, which is status quo to be fought against in the end of the day. Lets address that.
In order, namely, to issue the basic functionality such as network access from host to guest, qemu offers a interface that is recommended as well in debian howtos:
-redir option. Recently libvirt users 2011 came to that rabbit hole of connecting to their vms easily. There is to be setup a xml file, which is something like the representation, the setup of the virtual machine, to be used by one the interfaces virsh, virtlib or virtmanager.
Each time a new KVM guest is created using the virt-install tool, an XML file is created in the /etc/libvirt/qemu/guest_name.xml
Well, it did not surprise me, when I did not find a file created there or elsewhere. There is a function called –print-xml, so I found some xml code out of it. The way should then with virt-install to install a vm, without using this xml domain code but producing it, or lets say, producing a template of it. Hacking it by recently „leaked“ „secret backdoors“ and restarting it by virsh interface, which is supposed to have limitation given the late date of the „-redir“ access, right? How unfree is that?
The single user’s profit in domain representation should be the exchanging with the GUI virt-manager and with the management of VM parks. These are usecase designs that could_ come across. virt-manager at least seems, if GUI is questioned and it is, the better alternative than crippleware virtualbox. And I believe_ xml between it and qemu. Well, one sees, its a side street, luckily because it is -kept- uncertain. It was a bad idea to start learning hacking out of a tool like virt-install, even though it gave quick results, later there is no way further designed: cul de sac.
Qemu, however, is quite community driven, because community relevant, the kernel linux maintains it for their own purposes. So there was my mistake and there was the point where I found at last the way out.
Virtmanager keeps sexy. Hm?