vgscan --force-all-update-reread-and-take-that-kernel

Hello lvm folks,
Before I give you the whole shpiel... the issue I am having here
is that vgscan is unable to find an existing volume group after a repartitioning of the disk using 'parted' (hence the subject line that implies that vgscan needs an option to slap the kernel a little harder in some cases).

I am using lvm to automaticly extend a logical partition
to span the "rest of the harddisk", in the case that there are some unused gigs on the disk.

I've written a little script (which I'll attach here) that basicly:
- Resizes the extended partition
- Creates a new logical partition in the newly allocated space inside the extended partition
- Creates a new "physical volume" on the new partition
- Adds the new "physical volume" to the already existing "volume group"
- Extends the "logical volume" to span the entire "volume group"

The script works fine it you type it all out by hand in the shell and wait a few secs and repeatedly check vgscan to see if its back up, when I run the script... I get:

From when I first issue a vgscan:
=================================
Reading all physical volumes. This may take a while... No volume groups found

From my next call to pvcreate:
==============================
Couldn't find device with uuid 'HAkK8E-D3CS-87kk-GtL0-6ZGC-1NKI-x65WQg'. Couldn't find all physical volumes for volume group massgaol. Couldn't find device with uuid 'HAkK8E-D3CS-87kk-GtL0-6ZGC-1NKI-x65WQg'. Couldn't find all physical volumes for volume group massgaol.
get_pv_from_vg_by_id: vg_read failed to read VG massgaol Physical volume "/dev/hdb7" successfully created

Anyway, I guess at this point y'all know what I'm rambling on about...

please let me know what I'm doing wrong or what I could be doing better :)

Cheers,
-Tristan

#!/bin/bash

# This scrip makes some evil assumptions:
#

# o That the disk was partitioned as 120 Gigs
# o That the real disk size will be reported as numGB (in Gigs)
#   (i.e. this wont work if parted decides to display it in terabytes
#   or megabytes on some disks)
#

OUTPATTERN="Disk /dev/hdb: "
DISKOUT=parted /dev/hdb print | grep "$OUTPATTERN"

DISKSIZE=${DISKOUT#$OUTPATTERN}

echo "this disk size is $DISKSIZE"

echo "that would be ${DISKSIZE%GB} Gigs I think"

DISKGIGS=${DISKSIZE%GB}

if [ $DISKGIGS -gt 120 ]; then
# Found a disk that is physically bigger than 120GB

# Resize the extended partition to span till the end of the disk parted /dev/hdb resize 4 3208MB $DISKSIZE

# Create /dev/hdb7 with the remaining disk space parted /dev/hdb mkpart logical 120GB $DISKSIZE

# XXX massgaol dissapears here for a few secs.... sleep 2
vgscan

# Give the system some time to create the fs nodes needed # to proceed
sleep 5

# Create a physical volume for
pvcreate -y -ff /dev/hdb7

# Adding more ...
sleep 2

# Scan volume groups... needed to make 'massgaol' recognized # in the next step.
vgscan

# Add new physical volume to the volume group vgextend massgaol /dev/hdb7

# Get the current size of the virtual group now... # (get the 12'th field, separated by ':'). VGSIZE=vgdisplay -c massgaol | cut -d : -f12

# Activate the group -- this creates /dev/massgaol/massgaolLV vgchange -ay

# Extend the logical volume to span the entire volume group lvextend /dev/massgaol/massgaolLV -L${VGSIZE}KB

# Resize sometimes wants an fsck first.
e2fsck -f /dev/massgaol/massgaolLV

# Resize the overlying filesystem
resize2fs /dev/massgaol/massgaolLV

# deactivate the group for now
vgchange -an

fi

Hi again...
seems I have found a way to make the script work, by making the
"volume group" active throughout the whole procedure.

So, before I commit that and close this book... is what I am doing insane ? is it crazy to go and add a new physical volume to a volume group that is living and breathing ? (i.e. "active" via `vgchange -ay')

Cheers,
-Tristan

Yes. That's part of the whole point of LVM, it's operations can all be done online.

--
"Genius might be described as a supreme capacity for getting its possessors into trouble of all kinds."
-- Samuel Butler