Hello Heiko!
On Sat, 24 Nov 2001 00:33, Heiko Zuerker wrote:
> how should we handle the harddisk support?
> I think the first is to add the scripts from LFS which are related
> to mounting and checking of filesystems.
I would never use a non-journaling filesystem on devil-linux.
We do run from CD-Rom because we do not want a filesystem that's
not imune against interruption. If you read on you'll see that I
suggest we use our own strategy regarding mounting of filesystems.
> But how and where should we mount something?
> I think the best would be, that /var is mounted to the HDD,
> because most programs place there stuff there.
It looks like, but I wouldn't mount a single /var. I'd rather mount a
/var/squid, /var/spool/mqueue, .... Seperate partitions decouple
each service that needs the harddisk from the other. If eg. squid fills its
partition, ok, but mail works without interruption.
> What is about the initial file structure? Should this be created
> by a special script? How to check if the structure exists and was
> created from the correct version of the script?
I'd use LVM on top of a single harddiskpartition. So that should be
fairly easy to automate, because if we find an unpartitioned drive
we'd prompt the user if he want's to use this disk.
# fdisk -l
Disk /dev/hda: 16 heads, 63 sectors, 19885 cylinders
Units = cylinders of 1008 * 512 bytes
Device Boot Start End Blocks Id System
/dev/hda1 1 19885 9970632 8e Linux LVM
Now with LVM the logical partitioning should be fairly easy
too. See an exapmle output of the available volume groups (vg):
# vgdisplay
--- Volume group ---
VG Name devil-system
VG Access read/write
VG Status available/resizable
VG # 0
MAX LV 256
Cur LV 4
Open LV 4
MAX LV Size 255.99 GB
Max PV 256
Cur PV 1
Act PV 1
VG Size 9.51 GB
PE Size 4 MB
Total PE 2434
Alloc PE / Size 2434 / 9.51 GB
Free PE / Size 0 / 0
So grepping for "Total PE" gives us the total logical extends (=sectors)
which we then use to call the other lvm tools to make logical volumes
(=logical partitions).
So the startup script in initrd just needs to check if
vgscan finds the volume group (logical harddisk) named devil-system
then it could get on. If we are paranoid we'd check if we really
find all needed logical volumes.
eg. /dev/devil-system/mail-spool, /dev/devil-system/squid, ...
Now, why do I want to go this way? Here are some facts
* easy to automate fdisk in this way, just add on partition
that's a big as the disk
* we use logical partitions for the data, so we can resize them
whenever needed, even during runtime
* we could easily add software raid1 underneath lvm, there's just
one partition we need to mirror
* with ReiserFS/XFS/JFS on top we are up and running in "no" time
if the user ever feels the need to push the power button or if
there's power outage (UPS not really needed)
--
MfG / Regards
Friedrich Lobenstock
|
