I have searched, and many folks have asked - but nobody has gotten an answer.

I am building my companies third Linux system (RH 7.2) - and this one is using Software RAID-1 and RAID-0 mirrors mixed.

Here is my setup:
2 - IDE 40GB (no other drives)
/dev/hda1 & /dev/hdc1 - 101MB - /boot - (mirror)
/dev/hda2 & /dev/hdc2 - 1024MB - <swap> - (striped)
/dev/hda3 & /dev/hdc3 - 38GB - / (mirror)

The point is - I want to be able to keep my data if one of these two drives goes down. So I did an experiment. I turned off the system, unplugged /dev/hdc - and rebooted. So far so good - the MD driver sees that the drive is gone, and boots anyway - great. Swap is gone completely (obviously) - but it still boots and runs.

Shut it down again - plug in drive /dev/hdc - and the MD will not pick up hdc1 and hdc3. Says that the disk is "dirty" (boot text flies to fast - and this message didn't get captured by the log). Other items were captured in the log, but nothing about hdc1 or hdc3 - only the details on what it did do (not what failed).

I have no problem at this point if I have to destroy the system. It is only about two hours worth of setup - but I would like to make sure that I have my recovery procedure down pat. Is there some sort of rebuild the mirrors command I run? Do I need to "zero out" these devices? Do I need to shut down the disk swap and zero out the whole drive?

Thank you very much for any assistance you can give.

Gary
(who happily answers questions in the Networking Forum as time permits)

No need to stripe your swap space as the swap is suposed to use whatever device it deems most 'ready'.

What does 'cat /proc/mdstat' have to say about the issue?

You can rebuild the array with the raidhotadd command.

Funny - actually - when I tried to build the system without making every device part of a meta-device - the system could not initialize the swap. So, I assumed that a disk must be entirely part of a swap scheme or completely outside of a swap scheme.

Not much:

Personalities : [RAID0] [RAID1]
read_ahead 1024 sectors
md2 : active raid1 hda3[0]
37399232 blocks [2/1] [U_]

md0 : active raid0 hda2[0] hdc2[1]
3148544 blocks 54k chunks

md1 : active raid1 hda1[0]
104320 blocks [2/1] [U_]

unused devices: <none>

Could you direct me to a Man Page for this - seemingly - RH7.2 didn't have one, and --help isn't very usefull.

Thank you for your help!

raidhotadd /dev/hd__ /dev/md_

once things start whirring away, you can use 'watch cat /proc/mdstat' to see updates on how the rebuild is going.

try the boot partition first as it should be a quick rebuild.


As for the swap partitions - read up on the swap stuff in the most excellent RAID How-To's over at tldp.org. I think my favorite was Boot/Root RAID howto. Good reads fer sure.

The command line was backwards, but once I found out that two parameters were involved - all is going quite well...

raidhotadd /dev/md_ /dev/hd__

The command I used to start was:
# raidhotadd /dev/md2 /dev/hdc3

For those whom are curious - the /proc/mdstat (during rebuild) looks like the following:

Personalities : [RAID0] [RAID1]
read_ahead 1024 sectors
md2 : active raid1 hdc3[2] hda3[0]
37399232 blocks [2/1] [U_]
[====>................] recovery = 29% (16429956/74798464) finish=80.3m in speed = 12125K/sec
md0 : active raid0 hda2[0] hdc2[1]
3148544 blocks 54k chunks

md1 : active raid1 hda1[0]
104320 blocks [2/1] [U_]

unused devices: <none>