[CentOS-virt] Resize guest filesystem question

[Ed Heron]

On Fri, 2012-02-24 at 12:05 -0800, Jeff Boyce wrote:
> Greetings -
> 
> I am going through some testing steps to expand a logical volume and the 
> corresponding filesystem on a KVM guest and have run across a deficiency in 
> my knowledge.  I spent the afternoon yesterday googling for answers, but had 
> have come up blank still.  What I am trying to do is resize the file system 
> to use the additional disk space that I added to the logical volume that the 
> guest uses.  Here is what I have done and the details of my system.
> 
> 0.  Both my host and guest are running Centos 6.2.
> 
> 1.  My KVM host system has the LVM volume group that is divided into logical 
> volumes which are then presented to the KVM guests as raw space.
> 
> 2.  A guest may use 2 or 3 logical volumes from the host system for its 
> filesystem (/, /var, /data) and I have logical volumes named within the host 
> system by guest and mount point so that I know what each logical volume is 
> assigned to by it's name.
> 
> 3.  I expanded a specific logical volume on the host (/dev/vg/lv_guest1root) 
> that is used by Guest1, and I can see in vgdisplay and lvdisplay that the 
> logical volume was properly expanded.
> 
> 4.  I then issued a  resize2fs /dev/vg/lv_guest1root  command (on the host) 
> to resize the filesystem to the expanded logical volume.  This resulted in a 
> message that it essentially couldn't find a valid filesystem superblock. 
> Well of course then I realized that there is no filesystem on the logical 
> volume from the perspective of the host.  The filesystem wasn't set on the 
> logical volume until the guest installation occurred.
> 
> 5.  So then I switched over to the guest system and ran  df -h  to see the 
> existing filesystem
> 
> [root at guest1 jeffb]# df -h
> Filesystem            Size  Used Avail Use% Mounted on
> /dev/vda2             4.5G  2.3G  2.0G  53% /
> tmpfs                1004M   88K 1004M   1% /dev/shm
> /dev/vda1             485M   30M  430M   7% /boot
> /dev/vdb1             2.0G  219M  1.7G  12% /var
> 
> 6.  Then I ran  resize2fs /dev/vda2  and got the result that the filesystem 
> is already xx blocks long.  Nothing to do!
> 
> So here is where I am stuck.  Guest1 is my test system so it only has the / 
> and /var logical volumes, whereas the production guest (guest2) that I will 
> be expanding also has /data, which will be the logical volume that I will 
> expand.  So two things I did not do where, I did not shut down the guest VM, 
> and I did not unmount the filesystem before asking it to resize.  However my 
> research before doing this did not seem to indicate that I had to do either, 
> and the message about nothing to do also seems to indicate that they were 
> not necessary.
> 
> So I am missing a hole in my knowledge and additional googling has not 
> helped to fill it.  I must be missing something simple.  Is this result due 
> to the fact that I am testing on expanding the / filesystem, and it would 
> work properly on a guest system that had  /data?  Do I need to unmount the 
> filesystem, or shut down the guest VM, or mount the guest from a LiveCD?  Or 
> do I need to give it  resize2fs /dev/vda  rather than specifically 
> /dev/vda2 ?  Any clues, or pointers to good documentation is greatly 
> appreciated.  Thanks.

  The guest put a partition table on the LV.  To resize the filesystem,
the partition must be resized, as well.  There are several ways to do
it.  I think some people use gparted.  To recap the steps, resize the
LV, resize the partition, resize the filesystem.

  I setup a separate guest with the required tools to resize my guest
filesystems and temporarily take down the guest to give the disk space
to the utility guest.