A logical volume/filesystem question. I'm rather new to Linux, but know a particular UNIX rather well.


There is a volume on a RHEL 5 system consisting of a number of disk drives arranged in RAID 0 configuration.

One of the disks becomes unusable.

What is the best way of restoring filesystem functionality on the remaining drives (with reduced capacity)?

Note: Outage is (obviously!) expected. There are no data considerations to worry about.

Which one????
You back up your data, period. RAID 0 has no redundancy, so if you're getting errors, you will lose the entire array.
http://en.wikipedia.org/wiki/RAID#Standard_levels

You don't say if this is a hardware or software RAID setup, so it's hard to provide specific advice. Also, RHEL 5 is old...the last supported version is 5.9, I believe. If you're having hardware problems, now would be an EXCELLENT time to upgrade the system to the latest 7, since you've got hardware to fix, and are obviously going to have an outage anyway.

a) Solaris.

b) i) I don't care about the data. ii) Hardware.

It's just a question about how to resume operations using a (somewhat slower) 3-disk stripe when a failure occurs. A new disk would be ordered, of course, but it would be nice to get something working before it arrives (probably about 5 hours from time of putative failure).

Then you know a 'real' unix, and not something like AIX. It should be pretty easy for you to get around in Linux.
You don't. As said, RAID0 provides ZERO redundancy. If you lose a disk, you lose the array, period. There is nothing that holds parity or provides a hot failover. You would resume normal operations after restoring your backup to whatever new array you build.

Taking your question at face value, to get back up and running with the drives that are working, all you need to do is partition and format the new RAID entity. You can use parted or fdisk or cfdisk or Gnome Disks (GUI) for that, and it should just work:

1. Find the device node for the drive using dmesg.
2. Assuming the drive is at /dev/sdb, partition it:
3. Get total size of disk:
4. Assume the drive is 500G. Create a partition spanning entire disk:
5. Create filesystem:
Done.

But given that you specifically state that this is an LVM question, maybe you are asking how to get LVM running, using this hardware-RAIDed array as part of the storage pool, because you intend to expand it later. In that case, do something similar, but with LVM commands so that your RAID drive becomes part of your logical volume space:

1. Mark your drive as available for inclusion into your storage pool:
2. Create a volume group to include your drive:
3. Create a logical volume within this storage pool:
4. Create a filesystem on it:

Hope that helps.

Hi, thanks for that, it got me on the right track.

I removed the volume, to re-RAID the drives to use only the outer cylinders of the disks, as a test (performance etc.):

480 umount -f /fs2
482 vgdisplay
483 vgremove VolGroup02
485 pvscan
486 pvremove /dev/sdc1

The H/W RAID person did (whatever), which involved a reboot for some reason.

Unfortunately, I had forgotten to remove /fs2 from the fstab, so the system complained after booting, and I had to edit the fstab:


489 grep fs2 /etc/fstab
490 mount -o remount,rw /

vi /etc/fstab....)

The filesystem was then re-created:


491 pvscan
492 fdisk /dev/sdc
493 ls /dev/sdc1
494 pvcreate /dev/sdc1
495 vgcreate VolGroup02 /dev/sdc1
497 lvcreate -n LogVol00 -l 100%FREE VolGroup02
498 lvdisplay
499 mkfs -t ext4 /dev/VolGroup02/LogVol00


And that was that! Perhaps I'll be able to refine this and script it up, and also find out about the H/W RAIDing.

Thank you - but I don't care about the data at all.

It looks as though I have the start of a workable procedure.

Now, if "swapoff" wasn't so very, very ss-l-o-w! But that's a topic for another time...