In order to resize GFS some requirements need to be meat.
The new or resized partitions need to be seen at any node. This can eigher be done via a SCSI/FC-Rescan script (HP: hp_rescan) or if an existing partition has been resized with sfdisk -R. Be sure that any node in the cluster sees the same partitions with the same sizes.
If LVM is used be sure the clusterlocking is enabled. Check /etc/lvm/lvm.conf for locking_type=3 or locking_type=2. Next check if clvmd is running (/etc/init.d/clvmd status) and start if appropriate labeling new pvs then resize the volume group and then the logical volume. Check that all nodes see the same volumegroup configuration (*vgdisplay -v *).
Now if you have resized all dependencies execute *gfs_grow * on ONE node and everything should be fine.
You will only see problems if I/Os go behind the old device - that means if your filesystem access the new devices. Then unmount the filesystem on all nodes, deactivate the volumegroup (*vgchange -an ), vgscan and activate the volumegroup again (vgchange -ay volumegroup). Now you can savly remount the filesystem on all nodes.