On Thu, Apr 07, 2005 at 08:09:16PM +0200, Torsten Landschoff wrote: > > > > That's not true. If you can script it, FAI can do it. It just becomes a > > > > post-installation task. Using packages.d.o as an example, it's just > > > > going to > > > > be a predominantly an Apache configuration and some scripts, right? So > > > > you > > > > restore the scripts from backup, and dump the Apache config into the > > > > appropriate directory, all from within FAI. > > > > > > ... and adding users > > > ... and adding groups > > > ... and permissions > > > ... fixing/checking paths > > > ... and adding links > > > ... and initialising > > > ... and fixing scripts > > > ... and fixing the installation > > > ... and adding more permissions > > > ... and monitoring the initialisation > > > > All doable. > > Yep. The problem is that you need umpteen iterations for the > installation and you'll end up changing the FAI setup, installing, > waiting half an hour, testing, doesn't work, change FAI, repeat. > > Or you'll fix it up by hand and forget to tell FAI of your manual > changes to your next iteration will stop at the same position. And yes, > I have experience with that. Currently I am working on a FAI setup for > a lab of 10 Linux workstations which used to be manually maintained > which just eats too much time...
I patched FAI to do on-line updates, i.e. system updates without reinstallation. See http://www.physik.fu-berlin.de/~glaweh/pfai for details. In the near future, these patches will be merged into mainstream FAI, so 'softupdate' support will be available there. Thus, you can shorten your modify-test cycles quite a lot. the only things not handled by softupdate are (re-)partitioning (doesn't make much sense) and automatic software _removal_ when its not selected to be installed (I will work on this when the merge is complete). -- c u henning -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]