How to choose a GlusterFS volume type
A GlusterFS volume type is hard to change after client data lands on the volume. The choice controls whether the cluster spreads files without redundancy, keeps full copies, uses erasure-coded fragments, or combines those layouts across several protection sets.
GlusterFS builds volumes from bricks in a trusted storage pool. Replicated and arbiter layouts group bricks into replica sets, while dispersed layouts group data and parity fragments into disperse sets. A distributed layer does not add protection by itself; it places files across multiple replica or disperse sets for more capacity.
Choose from the failure the volume must survive, the raw capacity available for protection, and the fault domains available for each set. Keep bricks from the same protection set on separate nodes or racks whenever node-level availability matters, and treat plain distributed volumes as capacity-only storage that still needs another protection layer or backups.
Steps to choose a GlusterFS volume type:
- Map each planned brick to a failure domain such as a host, rack, or site.
Replica and disperse sets should avoid placing multiple bricks from the same set in the same failure domain.
- Choose a replicated volume when the cluster needs full data copies and a small node count.
Replica 2 stores two full copies and uses about half of raw capacity. Replica 3 stores three full copies and uses about one third of raw capacity.
- Choose an arbiter volume when two full data copies need a third voting brick without storing a third full copy.
Replica 2 arbiter 1 creates two data bricks plus one metadata-only arbiter brick in each set.
- Choose a distributed replicated volume when one replica set is too small for the required capacity.
The total brick count must be a multiple of the replica count so each distributed set has complete replica protection.
- Choose a dispersed volume when erasure-coded redundancy fits the workload better than full copies.
Redundancy is the number of brick failures tolerated per disperse set. Current GlusterFS rules require redundancy above 0 and more total bricks than twice the redundancy value.
- Choose a distributed dispersed volume when one disperse set is too small for the required capacity.
The total brick count must be a multiple of the disperse count so each distributed set has a complete data-plus-parity layout.
- Use a plain distributed volume only when another layer already protects the data.
Pure distribution has no GlusterFS redundancy. Losing one brick makes the files stored on that brick unavailable until restored from backup or recovered by another layer.
- Record the selected layout before creating the volume.
Include the type, replica or disperse count, redundancy value when used, brick order, and the failure domain for every brick.
- Verify the created volume reports the selected type and brick grouping.
$ sudo gluster volume info gv_app Volume Name: gv_app Type: Distributed-Replicate Status: Started Number of Bricks: 2 x 3 = 6 Transport-type: tcp Bricks: Brick1: node1:/srv/gluster/brick1/gv_app Brick2: node2:/srv/gluster/brick1/gv_app Brick3: node3:/srv/gluster/brick1/gv_app Brick4: node4:/srv/gluster/brick1/gv_app Brick5: node5:/srv/gluster/brick1/gv_app Brick6: node6:/srv/gluster/brick1/gv_app
Related: How to create a GlusterFS volume
Related: How to check GlusterFS volume status