Gluster concepts intro - the other distributed filesystem

GlusterFS is another open-source distributed filesystem. It takes a different approach than Ceph: assemble the filesystem from bricks across servers. This lesson covers the concepts.

How GlusterFS works

GlusterFS is a userspace filesystem (FUSE) that aggregates bricks (directories) across multiple servers:

• Each server exports a brick (a directory).
• Gluster assembles the bricks into a volume.
• The volume is mounted on clients via FUSE.

Server 1: /data/brick1
Server 2: /data/brick1
Server 3: /data/brick1

GlusterFS volume (distributed-replicated):
  /mnt/gluster

A client mounts /mnt/gluster and sees the combined filesystem. Gluster handles the distribution and replication in userspace.

Volume types

For most production, distributed-replicated is the right choice: files spread across bricks, with 2-3 replicas for HA.

Configure a basic volume

# On each server, create a brick directory
sudo mkdir -p /data/brick1

# Start glusterd
sudo systemctl enable --now glusterd

# On one server, create a trusted storage pool
sudo gluster peer probe server2
sudo gluster peer probe server3

# Create a distributed-replicated volume
sudo gluster volume create gv0 replica 3 \
    server1:/data/brick1 \
    server2:/data/brick1 \
    server3:/data/brick1

# Start the volume
sudo gluster volume start gv0

# Mount on a client
sudo mount -t glusterfs server1:/gv0 /mnt/gluster

When to use GlusterFS

• File-based workloads (web content, archives, logs).
• POSIX semantics required.
• Scale-out to many servers.

When not to use GlusterFS:

• Block storage for VMs (use Ceph RBD or SAN).
• Object storage (use Ceph RGW or S3).
• High-performance database (use a database).

Knowledge check