1 |
Hello! |
2 |
|
3 |
I have started to edit an /etc/raidtab file for a future raid and I |
4 |
just left off in the middle. Of course it was not in the fstab yet. |
5 |
Later I rebooted the machine and rc tried to start the half written |
6 |
array config by default (! #1), of course that failed and then it |
7 |
dropped me into a single user shell (! #2) although the system disk |
8 |
is a completely different non-md device, having all the filesystems |
9 |
needed to go multiuser. |
10 |
|
11 |
Seeing that I looked into the /etc/init.d/checkfs script and |
12 |
found the following behaviour: If there is an /etc/raidtab file |
13 |
and a /proc/mdstat, then it tries to start all the raid arrays, |
14 |
unless the raid device is _present_ in the fstab with a _noauto_ |
15 |
flag. IMHO thats like a cop shooting at everybody who doesn't |
16 |
hold a sign with the text "I'm not a gangster"... |
17 |
|
18 |
Here is what I think it should work like: |
19 |
|
20 |
- If there is a raidtab, _and_ in /etc/conf.d/raidstart (or call it as |
21 |
you like) it says (globally) RAID_AUTOSTART="yes" _and_ in the same |
22 |
file RAID_AUTOSTART_DEVICES="md0 md2 ..." lists the particular |
23 |
raid device as one to be automagically started _then_ it is started, |
24 |
regardless of what is in the fstab. Of course when someone is creating |
25 |
a new raid device, he has to explicitly put that device into the above |
26 |
variable when he thinks that the raid is ready to be autostarted, so |
27 |
he won't be surprised the way I've been. This might also suggest a |
28 |
separate /etc/init.d/raidstart script which has to be executed before |
29 |
checkfs, dedicated to the purpose of starting the raid arrays. |
30 |
|
31 |
Regards, |
32 |
Szabolcs Rumi |
33 |
|
34 |
-- |
35 |
gentoo-dev@g.o mailing list |