[Gluster-users] cluster/nufa disk full handling

[John Kilburg]

I am running glusterfs-2.0.0rc2 and I have noticed that cluster/nufa
volumes give out of space errors when the local volume is filled.
The other nodes can still write to their local volume until filled.
All nodes can read all of the files on the cluster/nufa volume.

This happened on pristine (all structure deleted) 2 node and 5
node clusters.

I assumed that cluster/nufa is a version of cluster/distribute
that prefers the local volume but uses other volumes in the cluster
when the local drive becomes full.

I see the same behavior with glusterfs-2.0.0rc1.

Volumes that are cluster/distribute fill all volumes in the cluster
on the same machines with "cluster/nufa" changed to "cluster/distribute"
and the structure cleaned.

How is cluster/nufa supposed to handle a full local volume?

							-john