Automating setup of disk_config files

[Ryan Steele]

Thomas Lange wrote:
>>>>>>     >> system, should it generate random disk configs, or do you mean to 
>>>>>>     >> create a disk_config from an existing, running system?
>>>>>>     >> 
>>>>>>
>>>>>>     > Exactly, the last option above. 
>>>>>> IMO this makes no sense. If you generate a disk_config file from an
>>>>>> existing partitioning scheme, this will not be that flexible as if you
>>>>>> write it manually. For e.g. if you have to change your disk, and the
>>>>>> new disk is smaller, than the generated partitioning scheme will
>>>>>> not fit to the new disk.
>>>>>>             

I am specifying specific sizes for my disk config, because I need my 
partition sizes to be known quantities prior to the installation.  I 
don't want FAI to 'guess' about how big to make them.  E.g., in the 
following configuration:

disk_config disk1
primary  /             150-300      rw,errors=remount-ro ; -c -j ext3
logical  swap          40-500      rw
logical  /var          90-1000     rw                   ; -m 5  -j ext3
logical  /tmp          50-1000     rw                   ; -m 0 -j ext3
logical  /usr          200-4000    rw                   ; -j ext3
logical  /home         50-         rw,nosuid            ; -m 1 -j ext3
# logical /home        preserve9   rw,nosuid            ; -m 1 -j ext3

...how big does the root partition end up being?  And what of /home?  I 
need to know that / is going to be of size X, and /home is going to be 
of size Y.

>>>>>> I guess you should think again about "Plan your installation and FAI
>>>>>> installs your plan". This is what is missing in your setup.
>>>>>>
>>>>>> I hope that the next step is not that you like to generate the
>>>>>> package_config files from an existing installation. This will also not
>>>>>> fit into the FAI method of "plan our installation".
>>>>>>             

I have several goals here, but one of them is to provide an easy way to 
reinstall hosed machines to the exact same state as they were before the 
catastrophe.  The way that the existing nodes are is exactly how I want 
future nodes to be installed, so I have consistency in my clusters.  
What is wrong with my existing nodes being reflective of my plans, if 
they fit my plan's model?