How-to Deploy GlusterFS (DRAFT)



This guide is a "draft" (i.e. it could be out-of-sync with the installation procedure anytime and it might not be updated here)

Infrastructure Setup



Rancher

This guide assumes a Rancher deployment, but any deployment with multiple (3+) Kubernetes nodes, with Docker and Helm will work. 

This guide is based on a deployment using Windriver ONAP lab.

You can follow this guide to set up your infrastructure - but skip the NFS setup section ("Setting up an NFS share for Multinode Kubernetes Clusters") if you do.



Storage

Once you have your Rancher and Kubernetes environment running on your Openstack environment, you will need to add raw volumes to each Kubernetes node from Openstack console. In this document, we assume that the volumes were mounted on /dev/vdb.

(Openstack will tell you which device path your volume is mounted when you attach your volume to your instance).

Here's a quick video on how to add volumes to Openstack:

Adding a volume to Openstack - Demo

GlusterFS setup

Once you have your infrastructure running and your RAW volumes mounted to your kubernetes nodes, deploy your Heketi / GlusterFS infrastructure. You can use the scripts included in OOM to automate this in your lab. (Not recommended for production install).

Demo Video

You can watch the execution of the script as per below instructions here:
GlusterFS Script Demo Video

Downloading the scripts 

There are 2 scripts contained within OOM resources: deploy_glusterfs.bash, which will set up your initial GlusterFS infrastructure. This will also deploy a Heketi pod, which is the RestAPI interface to manage your GlusterFS cluster.

There is also a cleanup script that you can use to cleanup your GlusterFS infrastructure when you are done or would like to re-deploy a clean GlusterFS infrastructure.

Grab the OOM artifacts from Gerrit (we did this on the Rancher master node in our deployment).

Currently the scripts are available by cherry-picking the following changeset: https://gerrit.onap.org/r/#/c/67049/

e.g. Run this after downloading Cassandra OOM: 
git fetch https://gerrit.onap.org/r/oom refs/changes/49/67049/1 && git cherry-pick FETCH_HEAD

git clone http://gerrit.onap.org/r/oom #git fetch https://gerrit.onap.org/r/oom refs/changes/49/67049/1 && git cherry-pick FETCH_HEAD cd oom/kubernetes cd contrib/resources/scripts ls -l
total 16 -rw-r--r-- 1 root root 1612 Sep 11 16:18 cleanup_gluster.bash -rw-r--r-- 1 root root 9956 Sep 11 18:51 deploy_glusterfs.bash
bash deploy_glusterfs.bash

The script will prompt you to hit "Enter" every once in a while to let you confirm there are no errors. There is minimal error checking in this script, so pay attention, especially when you are re-running the script after previously deploying GlusterFS.

The script will start off to direct you to run some commands manually on the other Kubernetes worker nodes (Openstack VMs):

 

Validation

Once the script is finished, check to make sure you have a valid StorageClass defined, and GlusterFS/Heketi Pods running on each Kubernetes node:

(Pod names and IP addresses will vary)

e.g

Deploy ONAP with OOM

You can choose any of the documented methods on this site or on onap.readthedocs.io, but here is a brief example of how you can deploy ONAP with GlusterFS.

Note: any persistent storage technology can be used in the example going forward, just make sure you have a storageClass already defined.



Edit / validate your values.yaml file

There is a custom values file in ~oom/kubernetes/onap/resources/environments called values_global_gluster.yaml that can be used, or you can edit the master values.yaml file at ~oom/kubernetes/onap/values.yaml.

We will assume that you are doing the former.

Ensure that you have your storageClass defined in the global section of your values file:

Ensure you have your storage class defined within global:persistence section:



Enable any components you want to deploy. In the values_global_gluster.yaml file, they are disabled by default. E.g. to enable APPC:

Once you have properly edited your values file, deploy ONAP:

Wait until components are up, and validate that your persistent volumes are using your persistent storageClass:





You should see "glusterfs-sc" (or whatever storageClass name you chose) under the STORAGECLASS column:



Ensure your pods are running and bound to your persistent volume:

e.g.



Make sure the pod is "Status: Running" and it is using the persistent volume that is associated with your storageClass: