Partition Sizes, AGAIN!

Dominic Ringuet dringuet at gmail.com
Fri Mar 5 14:46:20 PST 2010


It's all a matter of taste.

You could put an extra buck to get an other disk on the secondary ide
controller.

hda 40g
hdb cdrom
hdc 40g (might be something else)
hdd cdrom

With only 192mb RAM and LFS in mind, you'll need lots of swap; the magnitude
being around 2.5gb. For this purpose, the partition layout would then
become:

hda1    100M    ext2
hda5    10G     raid (md0)
hda6    10G     raid (md1)
hda7    19G     raid (md2)
hda2    1.2G    swap

hdc5    10G     raid (md0)
hdc6    10G     raid (md1)
hdc7    19G     raid (md2)
hdc1    1.3G    swap

If you like the red color, you use a single boot partition. But if dark
green is more amongst your taste, you might simply clone the partition
scheme from hda to hdc, thus ending with 2 boot space, one at the begining
of each disk. Of course, we don't know when something is going to mess up
with the MBR of a drive and why it could happens. Still, when it does happen
we bang our heads on the keyboard telling ourselves "why did'nt I set my
second drive bootable?"

prtn    size    mount point
----    ----    -----------
hda1    100M    /boot
hdc1    100M    /boot   (backup)
md0     20G     /       (main)
md1     20G     /       (build)
md2     38G     /home
hda2    1.2G    swap
hdc2    1.2G    swap

Considering raid 0 in such a config, your hedgehog is going to hop around
like a mid-aged rabbit. You can afford the few cpu cycles for almost twice
the throughtput to your disk accesses. After all, if building LFS deals with
2G of swap at some points in spacetime, you'll gain a smile with (almost)
twice the speed. Same goes to compressing/decompressing archives like
everything else.

The above example uses raid0, striping for speed. If your needs are
different, you could use the same logic to build a raid1 mirror.

Building raid array is quite simple for its benifits. All you need is
"mdadm", availlable right from the knoppix prompt, with sufficient help.
mdadm is also easily built on LFS. When the array is first created with
"mdadm --build", the partition holds its configuration data and is magically
recognized by a kernel with proper modules built-in. There was some work in
kernel to allow md partitioning. So one might look into something more
elegant.

For the swap file vs partition debate, I'd stick with the partition and use
a file if needs be. The swap is accessed in a raid manner when spawned
across multiple disk partitions. For the /boot matter, it is required when /
is mounted in some esoteric fashion, raid being an example.

Finally, I'd first get a good secondary drive then would partition as
follow:

prtn    size    mp      type
----    ----    --      ----
#hda1    100M    /boot   raid 1
#hda2   19G     /home   raid 1
hda3     20G     /       ext3
hda4    1.2G    swap

#hdc1    100M    /boot   raid 1
#hdc2   19G     /home   raid 1
hdc3     20G     /       ext3
hdc4    1.2G    swap

md0     100M  /boot   ext2
md1     20G   /home   ext3

The day your disk die, you may simply move /home around, building a stripe
for it then set it back. After this, you'll have to make good use of your
optical drives for backup. The basic idea is to have a boot, minimum 2
partitions and a swap on each disks. Everything else is like choosing the
song on a jukebox.

Dominic.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfromscratch.org/pipermail/lfs-support/attachments/20100305/d1caf62a/attachment.html>


More information about the lfs-support mailing list