metal3d/glusterfs repository overview

⁠GlusterFS in Docker container

A Centos based docker container running GlusterFS daemon.

⁠Usage

There you'll find several methods to launch GlusterFS with docker. The "docker-compose" method is nice to make some tests. Host method is nice to be used on real servers.

⁠Real server

On each server, you may launch:

docker run --net host --privileged --name glusterfs -v /hostdata:/data  metal3d/glusterfs

Take a look at "-v" options that mount your "/hostdata" directory inside the container as "/data". When you will create volumes and bricks, you'll need to use the "container" path.

Note the "--net host" usage that will open glusterfs ports to be accessible by other nodes.

When you're ready (having at least 2 containers running on 2 hosts), you can access glusterfs from one node and add peers. Here, we're using "node1.tld" and "node2.tld", and we're using node1 to launch commands:

# on node1.tld
$ docker exec -it glusterfs bash
/# gluster peer proble node2.tld
/# gluster create volume gv0 replica 2 transport tcp node1.tld:/data/brick1 node2.tld:/data/brick1 force
/# gluster volume start gv0
/# exit

Right now, we've declared a "gv0" volume that is able to be mounted on other hosts.

# NOT IN DOCKER
$ mount -t glusterfs node1.tld:/gv0 /mnt
$ date > /mnt/example.output

You should now have replication of example.output file on node1.tld and node2.tld in "/hostdata/brick1"

⁠Using docker-compose

This is a simple test that may be adapted to your needs. This example is really interessing to check how works GlusterFS and what's make options, replications, stripping, and so on.

I strongly recommand to use docker-compose "version 2" syntax that easilly configure network to let the hosts to be able to ping each others.

This is a docker-compose.yml file that create 2 servers:

version: '2'
services:
    gs1:
        image: metal3d/glusterfs
        privileged: true
        volumes:
        - ./volumes/server1:/data

    gs2:
        image: metal3d/glusterfs
        privileged: true
        volumes:
        - ./volumes/server2:/data

Then:

$ docker-compose up -d
$ docker-compose exec gs1 bash
/# gluster peer probe gs1
/# gluster volume create gv0 replica 2 gs1:/data/brick1 gs2:/data/brick1 force
/# gluster volume start gv0
/# mount gs1:/gv0 /mnt
/# date > /mnt/example.file

On the host, you will see "example.file" to be replicated in "./volumes/server1/brick1" and "./volumes/server2/brick1" directories

You may check logs by using:

docker-compose logs