Mirroring drives without RAID on Slackware 10.2

Jeiku · 06-20-2006, 12:29 AM

Hi all,

I have the following setup:

SCSI controller with 2x70GB hard drives attached.

I want these drives to act like they were on a RAID controller. Is there some software that will do this in Linux?
I want the drives to be constantly mirrored, so rsync is out of the question.

Thanks for any info!

Jake

Yalla-One · 06-20-2006, 01:10 AM

Have you looked into the Logical Volume Manager, LVM ?

Jeiku · 06-20-2006, 02:57 AM

Hi,

No I haven't. Have you managed to do this with LVM?

redcane · 06-21-2006, 05:19 AM

LVM can do mirroring. Otherwise you can use the Linux Software RAID system.

Jeiku · 06-22-2006, 03:08 AM

Thanks for the replies =]

Slim Backwater · 06-23-2006, 08:47 AM

Quote:

Originally Posted by redcane

LVM can do mirroring. Otherwise you can use the Linux Software RAID system.

Wait, LVM can do mirroring? I have LVM running on Software RAID, is that what you mean? I created a mirror (md0) with mdadm, then made a physical volume on the md0 device. (from that PV, I created a volume group, and then logical volumes). If LVM can do mirroring itself, is it 'better' than lvm on RAID?

As for the OP, check out the software Raid, mdadm and all. For your reference, here are my notes from my last install. Some stuff won't apply (like umounting /pub my own nfs mount) or make sense, and I only wrote this, I haven't followed it yet to fix errors, but maybe it will help.

I was using IDE drives so the references to hda and hdb, but as you are using SCSI drives, they will probably be sda and sdb. The most annoying things are the old version of LILO in Slackware, which doesn't support degraded RAID-1, and the default init-rd is missing md entries in /dev.

This will cover creating a sofware RAID-1 environment to run Slackware on. The two devices are hda and hdb. They do not need to be identical, only the portions that will be mirrors.

Install Slackware onto hda. Probably should apply patches, compile a new kernel, etc.
My custom compiled kernel was installed as /boot/vmlinuz-20060606-2.6.16.20 which included all the modules I needed to boot built-in (sata drivers, reiserfs, md-mod, dm-mod, raid1).
Install the LVM2 tools from Disc2
Probably need to modprobe dm-mod
Create a three partitions on hdb: Probably have to reboot after creating the partitions
- hdb1 128MB type FD (for /boot mirror)
- hdb2 2048 type FD (for swap mirror)
- hdb3 Rest type FD (for LVM PV mirror)
Create the degraded RAID-1 arrays
- mdadm --create --level=raid1 --raid-devices=2 /dev/md0 missing /dev/hdb1
- mdadm --create --level=raid1 --raid-devices=2 /dev/md1 missing /dev/hdb2
- mdadm --create --level=raid1 --raid-devices=2 /dev/md2 missing /dev/hdb3
Probably could stand to change the partition type to FD to have the raid autobuild. There's notes about having to run raidstart --all otherwise
- Make filesystems on the RAID devices
- mkfs -t ext2 /dev/md0
- mkswap /dev/md1
- pvcreate /dev/md2
- vgcreate main /dev/md2
- lvcreate main --size 256M --name root
- lvcreate main --size 1024M --name opt
- lvcreate main --size 256M --name tmp
- lvcreate main --size 4096M --name usr
- lvcreate main --size 256M --name var
- lvcreate main --size 4096M --name home
- mkreiserfs -q /dev/main/root
- mkreiserfs -q /dev/main/opt
- mkreiserfs -q /dev/main/tmp
- mkreiserfs -q /dev/main/usr
- mkreiserfs -q /dev/main/var
- mkreiserfs -q /dev/main/home
Mount all those partitions under /mnt/hd
- mount /dev/main/root /mnt/hd
- mkdir /mnt/hd/opt
- mkdir /mnt/hd/tmp
- mkdir /mnt/hd/usr
- mkdir /mnt/hd/var
- mkdir /mnt/hd/home
- mount /dev/main/opt /mnt/hd/opt
- mount /dev/main/tmp /mnt/hd/tmp
- mount /dev/main/usr /mnt/hd/usr
- mount /dev/main/var /mnt/hd/var
- mount /dev/main/home /mnt/hd/home
Create a mountpoint for, and mount /boot
- mkdir /mnt/hd/boot
- mount /dev/md0 /mnt/hd/boot

Make an initrd

If using a generic kernel
- mkinitrd -c -k 2.6.16.18 -m libata:reiserfs:sata_nv:sata_sil:md-mod:dm-mod:raid1
Or with a custom Kernel
- mkinitrd -c -k 2.6.16.20

Edit the initrd

cd /boot/initrd-tree
rm dev/md[0-4]
mknod --mode=0660 dev/md0 b 9 0
mknod --mode=0660 dev/md1 b 9 1
mknod --mode=0660 dev/md2 b 9 2
mknod --mode=0660 dev/md3 b 9 3
mknod --mode=0660 dev/md4 b 9 4
mkdir sbin
cp /sbin/lvm.static sbin

Code:

( cd sbin ; rm -rf lvs )
( cd sbin ; ln -sf lvm lvs )
( cd sbin ; rm -rf pvs )
( cd sbin ; ln -sf lvm pvs )
( cd sbin ; rm -rf vgs )
( cd sbin ; ln -sf lvm vgs )
( cd sbin ; rm -rf vgck )
( cd sbin ; ln -sf lvm vgck )
( cd sbin ; rm -rf lvdisplay )
( cd sbin ; ln -sf lvm lvdisplay )
( cd sbin ; rm -rf pvchange )
( cd sbin ; ln -sf lvm pvchange )
( cd sbin ; rm -rf pvcreate )
( cd sbin ; ln -sf lvm pvcreate )
( cd sbin ; rm -rf vgmknodes )
( cd sbin ; ln -sf lvm vgmknodes )
( cd sbin ; rm -rf pvremove )
( cd sbin ; ln -sf lvm pvremove )
( cd sbin ; rm -rf pvresize )
( cd sbin ; ln -sf lvm pvresize )
( cd sbin ; rm -rf lvmchange )
( cd sbin ; ln -sf lvm lvmchange )
( cd sbin ; rm -rf lvmsadc )
( cd sbin ; ln -sf lvm lvmsadc )
( cd sbin ; rm -rf vgcfgrestore )
( cd sbin ; ln -sf lvm vgcfgrestore )
( cd sbin ; rm -rf lvmdiskscan )
( cd sbin ; ln -sf lvm lvmdiskscan )
( cd sbin ; rm -rf pvdisplay )
( cd sbin ; ln -sf lvm pvdisplay )
( cd sbin ; rm -rf lvmsar )
( cd sbin ; ln -sf lvm lvmsar )
( cd sbin ; rm -rf lvscan )
( cd sbin ; ln -sf lvm lvscan )
( cd sbin ; rm -rf vgchange )
( cd sbin ; ln -sf lvm vgchange )
( cd sbin ; rm -rf vgcreate )
( cd sbin ; ln -sf lvm vgcreate )
( cd sbin ; rm -rf pvmove )
( cd sbin ; ln -sf lvm pvmove )
( cd sbin ; rm -rf pvscan )
( cd sbin ; ln -sf lvm pvscan )
( cd sbin ; rm -rf vgexport )
( cd sbin ; ln -sf lvm vgexport )
( cd sbin ; rm -rf vgextend )
( cd sbin ; ln -sf lvm vgextend )
( cd sbin ; rm -rf vgimport )
( cd sbin ; ln -sf lvm vgimport )
( cd sbin ; rm -rf vgscan )
( cd sbin ; ln -sf lvm vgscan )
( cd sbin ; rm -rf vgmerge )
( cd sbin ; ln -sf lvm vgmerge )
( cd sbin ; rm -rf vgsplit )
( cd sbin ; ln -sf lvm vgsplit )
( cd sbin ; rm -rf lvchange )
( cd sbin ; ln -sf lvm lvchange )
( cd sbin ; rm -rf lvcreate )
( cd sbin ; ln -sf lvm lvcreate )
( cd sbin ; rm -rf lvextend )
( cd sbin ; ln -sf lvm lvextend )
( cd sbin ; rm -rf vgreduce )
( cd sbin ; ln -sf lvm vgreduce )
( cd sbin ; rm -rf vgrename )
( cd sbin ; ln -sf lvm vgrename )
( cd sbin ; rm -rf vgremove )
( cd sbin ; ln -sf lvm vgremove )
( cd sbin ; rm -rf vgconvert )
( cd sbin ; ln -sf lvm vgconvert )
( cd sbin ; rm -rf vgcfgbackup )
( cd sbin ; ln -sf lvm vgcfgbackup )
( cd sbin ; rm -rf lvreduce )
( cd sbin ; ln -sf lvm lvreduce )
( cd sbin ; rm -rf lvrename )
( cd sbin ; ln -sf lvm lvrename )
( cd sbin ; rm -rf lvremove )
( cd sbin ; ln -sf lvm lvremove )
( cd sbin ; rm -rf lvresize )
( cd sbin ; ln -sf lvm lvresize )
( cd sbin ; rm -rf vgdisplay )
( cd sbin ; ln -sf lvm vgdisplay )

Update the initrd.gz

cd /boot
mkinitrd

Edit lilo.conf, add a new entry to bootstrap the RAID devices

Don't change the boot= at this time, wait until booted off the mirror

Code:

vi /etc/lilo.conf
image = /boot/vmlinuz-20060606-2.6.16.20
  root = /dev/main/root
  initrd = /boot/initrd.gz
  label = nRAID-2.6.16.18
  read-only

Copy everything over to the new home

- umount /pub
- rsync -av --exclude /mnt/hd --exclude /proc --exclude /sys / /mnt/hd/
  - or cp -avx / /mnt/hd
  - cp -avx /boot /mnt/hd/

Edit new fstab

vi /mnt/hd/etc/fstab

Code:

/dev/md1         swap             swap        defaults         0   0
/dev/main/root   /                reiserfs    defaults         0   0
/dev/md0         /boot            ext2        defaults         1   1
/dev/cdrom       /mnt/cdrom       auto        noauto,owner,ro  0   0
/dev/fd0         /mnt/floppy      auto        noauto,owner     0   0
devpts           /dev/pts         devpts      gid=5,mode=620   0   0
proc             /proc            proc        defaults         0   0
/dev/main/var    /var       reiserfs       defaults    0    0
/dev/main/opt    /opt       reiserfs       defaults    0    0
/dev/main/tmp    /tmp       reiserfs       defaults    0    0
/dev/main/usr    /usr       reiserfs       defaults    0    0
/dev/main/var    /var       reiserfs       defaults    0    0
/dev/main/home   /home      reiserfs       defaults    0    0
scrambled:/home/brian /pub        nfs         defaults 0    0

Reboot and see if it boots off the raid!
If it boots, update the boot= line in lilo.conf

vi /etc/lilo.conf
- boot=/dev/md0
- raid-extra-boot=mbr
delete stuff from /boot that is no longer relevant (old kernels, old initrds, etc.)

Hard work, LILO 22.5.9 (with Slackware 10.2 and -current as of 2006-06-06) can't handle the degraded RAID-1. Download, compile and install Lilo 22.7.1 (atleast).
Run lilo

lilo
reboot and try to boot off the second disk.

Create mirror partitions

Wipe out the partition table (and some) on the old drive
- dd if=/dev/zero of=/dev/sda bs=512 count=2000 # BE VERY CAREFUL!
Reboot
Duplicate the partition table from the second drive to the first drive
- sfdisk -d /dev/sdb | sfdisk /dev/sda # BE VERY CAREFUL!
Reboot
Add the new partitions to the correct RAID array
- mdadm --detail /dev/md0
- mdadm /dev/md0 --add /dev/sda1
- mdadm --detail /dev/md1
- mdadm /dev/md1 --add /dev/sda2
- mdadm --detail /dev/md2
- mdadm /dev/md2 --add /dev/sda3
run lilo and see that both mbr's get updated (SDA and SDB)

HTH.

BananaRepublic · 01-23-2008, 04:24 PM

I'm bit lost trying to figure what I really need to accomplish in my situation.

I already created a RAID1 for my boot partition, and it seems to boot from the /dev/md0 as expected at the root= line.

However, lilo.conf has this line:

Code:

boot= /dev/sda

Which isn't really what I want as if this particular hard drive fails, the system will not boot at all.

I tried to add this:

Code:

raid-extra-boot="/dev/sdb,/dev/sdc,/dev/sdd"

but this didn't make any difference.

Also, this won't work:

Code:

boot= /dev/md0

I do not have a initrd, and am not sure if this is required in my case.

What should I look for to get LILO to boot from RAID1 array completely?

mRgOBLIN · 01-23-2008, 04:58 PM

Did you run lilo after making the changes to lilo.conf?

You might find some useful info here http://www.userlocal.com/articles/ra...ackware-12.php

BananaRepublic · 01-23-2008, 05:42 PM

Quote:

Originally Posted by mRgOBLIN

Did you run lilo after making the changes to lilo.conf?

You might find some useful info here http://www.userlocal.com/articles/ra...ackware-12.php

Excellent- that got everything working correctly. I didn't realize I had to run lilo whenever I edit the lilo.conf.

Now, the mdadm --detail indicates that array is still degraded even though all HDs are now plugged in now. Am I supposed to do something to get it reconstructed?

EDIT: I was under the mistaken impression that mdadm would automatically detect the HDs that was re-plugged back after the test and reconstruct it. This is not the case, apparently. To reconstruct the array:

Code:

mdadm --manage --add /dev/md[X] /dev/[X]