On sam, oct 02, 1999 at 01:18:34 -0700, Jones, Clay wrote:
> I'm not sure if there is a political reason why the "alpha" raid is not in
> the kernel. It's been available since 2.0.30 as a patch (well over a year I
> think).
It was actually put in some 2.2.11-prex (or was it 12-prex), and was backed
out because some people who don't read docs complained that it broke their
0.50 array.
The problem is that if you install the newer kernel, and reboot without
doing anything else, things will break. You need a new suite of user level
tools, and you need to read the docs.
Don't flame me for saying this, I'm merely repeating what came from the
thread about this issue that was on linux-kernel a few weeks ago.
Ingo promised he would do his best so that this gets into 2.4 once and for
all.
The part that really sucks IHO is the useless error messages that mkraid
returns. I mean "failed" isn't very useful...
It should say: "your user tools protocol version 0.90.2 don't match the code
in the kernel which is version 0.90.1 (or version 0.50.3)"
Or it should say "you are missing option xyz in /etc/raidtab".
> The last I'd read was that the Kernel team did not want any changes that
> required "Tool" changes as well. So if the new raid would run with the old
> tools, they would probably add it.
Although the newer knfs requires new tools, it's apparently going to go in
2.2.13, but the idea is that if you reboot and knfs fails because you didn't
read the docs and didn't upgrade your tools, it's no big deal.
If you reboot your machine and a critical array (or even your root
filesystem) doesn't come up, that's much worse
I would personally put the newer patch in, but have the kernel Makefile for
raid pause the compilation process, display a message warning about the new
tools, and have the user type ok before the compilation resumes.
I didn't come up with the idea, the Debian folks have been doing this with
their packages.
I think it'd be nice if Alan did that with the knfs patch so that people get
a fair warning and can't complain :-)
(I can also envision a variable in the main Makefile that says "don't bother
me with warnings and don't ever pause my kenrel compile, I know what I'm
doing")
Alan?
> All in all, the newer raid has been much more stable. Other than a lack of
> detailed documentation, the newer raid tools generally work better.
I think we all agree here :-)
Marc
--
Microsoft is to software what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/ (friendly to non IE browsers)
Finger [EMAIL PROTECTED] for PGP key and other contact information
