Once again about 3TB

[Dima Veselov <i%kab00m.ru@localhost>]

Hello!

Several times I was trying to settle 3TB on recent stable version.
Now it works and I just wanted to drop something helpful for people
not to waste time.

Now I have production system working on two different 3Tb drives,
handled by NetBSD RAIDframe.

Few advices/notes:

1. Despite of some messages found - RAIDframe can't be built
over raw devices (at least on NetBSD 6.1.4/amd64) - it always
will be limited to 2Tb. Somewhere I saw "kernel know disk size,
but disklabel don't" - thats not true. Everything you try with 'd'
device will be limited to 2Tb.

2. Booting from GPT works well, but it is a real mess to settle.
There is no good documentation about 'gpt biosboot' command - some
says 'gpt biosboot wd0', some says 'gpt biosboot dk0'.
I don't understand much about offsets working there, and have a
sandwitch [wd0 [dk0 [raid0a]]] and somehow this worked out as a
magic. However few things should be fixed (or better documented):

gpt biosboot wd0 says no partition can be found
if disk wasn't zeroed - installboot complain 'not enough space
before FAT'. Worked after zeroing and dk re-creation.
documentation says to create FAT partition and boot it, this is
outdated and don't work at all.

3. dk-raid-dk sandwitch is working! Magic :)
Don't know if it is good or not - I created many wedges for both
disks, most of them work like wd0 -> dk -> raidXa and most big one
work as wd-dk-raid-dk

/dev/raid0a        241M        53M       175M  23% /
/dev/raid4a        7,8G       644M       6,7G   8% /var
/dev/raid5a         16G       2,3G        12G  15% /usr
/dev/dk10          2,7T       824G       1,7T  31% /export

4. dkctl can be "tired". If you play with it much (create/delete
wedges) it always do well, but suddenly all new wedges won't work.
Any try to touch it fail and only reboot will help. This is a bug,
but I don't know how to repeat it.

And few questions I have - where dkctl store wedge information?

-- 
Sincerelly yours