Hey,

I am kicking myself for buying into the Windows Home Server hype. It looks like MS was vaguely aware of this problem (http://support.microsoft.com/kb/946676/en-us) and decided to ship it anyhow. What I have now is a file server I'm afraid to put files on.

What interested me most was the drive pooling feature. It combines multiple drive and displays them as one similar to JBOD. But what I was most concerned about was being able to access the data when a drive crashed or even if the OS got fubar.

With WHS, all the data is in NTFS format. I can pull a drive out of the pool and plug it into a windows machine and read the data. It's not very organized but it's all there. Until recently I never knew linux was capable of something similar and has been for a long time using LVM, I had always used ext2 or 3.

A diagram of the layout: (6 port onboard fakeraid)

.RAID1_250.__.RAID1_500.__.RAID1_750
_____/\___________/\___________/\
____/**\_________/**\_________/**\
___/****\_______/****\_______/****\
250GB.250GB__500GB.500GB__750GB.750GB
__A_____B______C_____D______E_____F

WHS sees 250GB+500GB+750GB and displays it as one 1.5TB drive. So for instance I can put out drive "B" and plug it into a windows PC and read the data (standard NTFS).

So to test linux LVM I install fedora 8 on a couple drive without using RAID to simplify it and set up the partitions like so:

sda: 80GB
-sda1: /boot 100MB
-sda2: swap 2048GB
-sda3: / 20GB
-sda4: LVM (VolGroup00-LogVol00) 40GB

Then I added another drive.

sdb: 200GB
-sdb1: LVM (VolGroup00-LogVol00) 200GB

So I had a logical volume of 240GB mounted at /shares. I copied 70GB of data to /shares using samba (40GB on sda4 and 30GB on sdb1). Then I removed sda to simulate a total OS drive failure and try to recover the 30GB of data on sdb using the fedora 8 live cd. I have scoured the net, read countless forums, read the man pages - it just eludes me.

This tutorial (http://linuxwave.blogspot.com/2007/1...tu-livecd.html) along with a radhat one I lost the link to got me pretty close but it required I have the /etc/lvm/* data handy, which I copied to a thumbdrive before I pulled the plug on the OS drive.

Using this method:

pvscan
vgscan
vgchange -ay
lvscan
mount /dev/VolGroup00/LogVol00 /mnt

without the /etc/lvm/* data got me nowhere. Then I tried adding a temp drive (sdc) and doing this first (http://kbase.redhat.com/faq/FAQ_85_5843.shtm):

pvcreate --restorefile /etc/lvm/backup/volGroup00 --uuid uuid-of-missing-os-device /dev/sdc1

vgcfgrestore --file /etc/lvm/backup/VolGroup00 VolGroup00

Which appears to work without error. But when I try to mount it, I get this error:

mount /dev/VolGroup00/LogVol00 /mnt
mount: you must specify the filesystem type

I tried ext2 and 3 to no avail. I hoping someone can shed some light on this for me.

Thanks in advance.

What is happening in Windows is more like unionfs. Three distinctly different filesystems being presented as a logical whole. The directory you are in determines where files go and when that isn't clear, then there is the notion of a preferred order.

But broken apart, each area should be able to stand alone.

LVM is different. It allows you to create pools of storage area (NOT filesystems) and lets you creates logical volumes (similar to a partition) on which you can create filesystems. However, in order to pull a drive, you have to use a pvmove command (which can take some time) to physically move all data off of a physical volume onto other available physical volumes prior to being able to remove the physical volume from the volume group and ultimately removing the drive.

Of course, that's pretty proactive... in other words, not really designed to tolerate a total drive failure. But the above MIGHT work in the case of a failing drive that hasn't completley failed. A better solution for handling failure would be to use RAID.

I don't have a document on unionfs, but you can look at: http://www.linuxjournal.com/article/7714 and the unionfs web site.

I do have a good document on LVM: http://ntlug.org/Articles/LVM