Old style fixed partitions in KS-- Anaconda bug??

[Cris Rhea]

Michael DeHaan <mdehaan at redhat.com> wrote:
> Can't help you debug your sfdisk, but the magic cobbler
> (http://cobbler.et.redhat.com) uses by default is the following...
> 
> %include /tmp/partinfo
> 
> %pre
> # Determine how many drives we have
> set \$(list-harddrives)
> let numd=\$#/2
> d1=\$1
> d2=\$3
> 
> cat << EOF > /tmp/partinfo
> part / --fstype ext3 --size=1024 --grow --ondisk=\$d1 --asprimary
> part swap --size=1024 --ondisk=\$d1 --asprimary
> EOF
> 
> This has been working very well.
> 
> There shouldn't be any reason to call sfdisk that I'm aware of.

The difference is that in my case, I have an external program (sfdisk)
writing the partition table-- in the above, Anaconda is writing the
partition table with its internal routines.

Sfdisk writes the partition table correctly (as viewed by a shell window), 
but Anaconda gets confused using it.

I don't want to use something like the above because:

-- I want to have more granular file systems (separating /, /usr, /var,
   /tmp, etc.). 

-- I want to have the same layout across many systems (e.g., cluster nodes).
   From what I've seen, having Anaconda/Disk Druid lay out partitions 
   can result in them being in a different order (e.g., Disk Druid
   arranges them by size).

Controlling disk layout compared to having everything under "/" is a
subjective, administrative decision. I'm not trying to convince anybody
that "my way is better", just trying to get it to work (and it has 
always worked fine in everything pre-RHEL4u5).

Thanks--

--- Cris



-- 
 Cristopher J. Rhea                     
 Mayo Clinic - Research Computing Facility
 200 First St SW, Rochester, MN 55905
 crhea at Mayo.EDU
 (507) 284-0587