IBM’s GPFS filesystem expects that all members of the GPFS cluster, be they clients or servers, have a complete and consistent state of the cluster. This causes some difficulty in re-adding a node to the cluster after it’s lost it’s … Continue reading →