Hyper-converged GlusterFS + heketi on Kubernetes

gluster-kubernetes is a project to provide Kubernetes administrators a mechanism to easily deploy a hyper-converged GlusterFS cluster along with heketi onto an existing Kubernetes cluster. This is a convenient way to unlock the power of dynamically provisioned, persistent GlusterFS volumes in Kubernetes.

Link: https://github.com/gluster/gluster-kubernetes

Component Projects

  • Kubernetes, the container management system.
  • GlusterFS, the scale-out storage system.
  • heketi, the RESTful volume management interface for GlusterFS.


You can find slides and videos of community presentations here.

>>> Video demo of the technology! <<<



You can start with your own Kubernetes installation ready to go, or you can use the vagrant setup in the vagrant/directory to spin up a Kubernetes VM cluster for you. To run the vagrant setup, you'll need to have the following installed:

  • ansible
  • vagrant
  • libvirt or VirtualBox

To spin up the cluster, simply run ./up.sh in the vagrant/ directory.

Next, copy the deploy/ directory to the master node of the cluster.

You will have to provide your own topology file. A sample topology file is included in the deploy/ directory (default location that gk-deploy expects) which can be used as the topology for the vagrant libvirt setup. When creating your own topology file:

  • Make sure the topology file only lists block devices intended for heketi’s use. heketi needs access to whole block devices (e.g. /dev/sdb, /dev/vdb) which it will partition and format.
  • The hostnames array is a bit misleading. manage should be a list of hostnames for the node, but storage should be a list of IP addresses on the node for backend storage communications.

If you used the provided vagrant libvirt setup, you can run:

$ vagrant ssh-config > ssh-config
$ scp -rF ssh-config ../deploy master:
$ vagrant ssh master
[vagrant@master]$ cd deploy
[vagrant@master]$ mv topology.json.sample topology.json

The following commands are meant to be run with administrative privileges (e.g. sudo su beforehand).

At this point, verify the Kubernetes installation by making sure all nodes are Ready:

$ kubectl get nodes
master Ready 22h
node0 Ready 22h
node1 Ready 22h
node2 Ready 22h

NOTE: To see the version of Kubernetes (which will change based on latest official releases) simply do kubectl version. This will help in troubleshooting.

Next, to deploy heketi and GlusterFS, run the following:

$ ./gk-deploy -g

If you already have a pre-existing GlusterFS cluster, you do not need the -g option.

After this completes, GlusterFS and heketi should now be installed and ready to go. You can set the HEKETI_CLI_SERVERenvironment variable as follows so that it can be read directly by heketi-cli or sent to something like curl:

$ export HEKETI_CLI_SERVER=$(kubectl get svc/heketi --template 'http://{{.spec.clusterIP}}:{{(index .spec.ports 0).port}}')
$ curl $HEKETI_CLI_SERVER/hello
Hello from Heketi

Your Kubernetes cluster should look something like this:

$ kubectl get nodes,pods
master Ready 22h
node0 Ready 22h
node1 Ready 22h
node2 Ready 22h
glusterfs-node0-2509304327-vpce1 1/1 Running 0 1d
glusterfs-node1-3290690057-hhq92 1/1 Running 0 1d
glusterfs-node2-4072075787-okzjv 1/1 Running 0 1d
heketi-3017632314-yyngh 1/1 Running 0 1d

You should now also be able to use heketi-cli or any other client of the heketi REST API (like the GlusterFS volume plugin) to create/manage volumes and then mount those volumes to verify they're working. To see an example of how to use this with a Kubernetes application, see the following:

Hello World application using GlusterFS Dynamic Provisioning

A single golf clap? Or a long standing ovation?

By clapping more or less, you can signal to us which stories really stand out.