[Linux-cluster] Basic question on CLVMD / GFS

Michael Weitzel michael.weitzel at uni-siegen.de
Wed Dec 21 15:52:32 UTC 2005


I am confused about the usage of clvmd. On a blank 80 GB disk I created
a logical volume using the following commands (after the cluster was
started with "ccsd; cman_tool join; fence_tool join" on all nodes):

# overwrite a former partition table:
dd if=/dev/zero of=/dev/hdb bs=512 count=1
# initialize the physical volume:
pvcreate /dev/hdb
# create a volume group "vg301":
vgcreate vg301 /dev/hdb
# create a 40 GB logical volume "lvol0" inside vg301:
lvcreate -L 40G vg301
# start clvmd
# mark the volume group as "clustered" (there are no other vg's):
vgchange -cy
# activate the volume group with/without cluster locking
vgchange -aey ...or... vgchange -aly

vgchange prints a notice if on some other node of the cluster clvmd is
not running and activates the logical volume.

Is this enough to use "vg301" or "vg301/lvol0" on the other nodes? If
yes, how do I access vg301 from another (than the local) node? vgscan,
executed on the other nodes, does not report anything.

Distributing the local vg301/lvol0 (with GFS on it) with GNBD 'nearly'
worked: At some point, the I/O was blocked on nodes accessing a specific

gfs_mkfs -p lock_dlm -t Isabella:lvol0_301 -j 4 /dev/vg301/lvol0
gnbd_export -v -e exp301_0 -d /dev/vg301/lvol0

Is there any information / documentation about clvmd? The lvm-HOWTO does
not cover this topic :-(

Many thanks!
Michael Weitzel

More information about the Linux-cluster mailing list