GlusterFS Cluster deployment scripts

GlusterFS Cluster deployment scripts
- Requirements
  - System
  - Dependencies
- Usage
  - Local usage
    - Using docker
    - Customization
  - Remote usage

This folder hosts code for provisioning and testing a GlusterFS cluster ready with DMS deployed to the main node.

Requirements

System

16GB ram (2GB for each of the four VMs and 8GB for the host)
8 VCPU

Dependencies

incus
ansible
jq
openssh-client
unzip

Usage

These scripts are intended to be used locally or for provisioning a remote cluster.

Local usage

First, make sure incus is installed and configured with incus admin init.

Use ./clear-and-launch.sh to clear previous glusterfs VMs and reprovision them.

If config.yml isn't set, a new one is created from config.dist.yml. At the time of creation, the attribute extra_allowed_ips is modified to add the ipv4 address of the incusbr0 network interface, so that the glusterfs cluster trusts all the IPs from the incus server, including the host IP.

If ACC_TEST_DMS_DEB_FILE is set, the script will write it to config.yml under the local_dms_deb attribute.

The script will also create an ssh key in the root folder and add it to cloud-init.yml that is used for launching the VMs. It'll wait for cloud-init to finish and will populate a hosts_glusterfs file ready for ansible.

After this initial provisioning, run the ansible scripts with ssh-agent to deploy the glusterfs cluster:

ssh-agent bash run.sh

After the deployment is done, you can test it with:

ssh-agent bash run.sh test

Machines can be accessed either via incus shell or via ssh.sh:

incus shell glusterfs-test-node1
# or
./ssh.sh [client|server]

Using docker

A prepared Dockerfile can be found in the project root. First build the image:

docker build -t glusterfs-cluster-ansible .

Then launch the container:

docker run -it --rm \
  --network host \
  -v $PWD:/app \
  -v /var/lib/incus/unix.socket:/var/lib/incus/unix.socket \
  --workdir /app \
  glusterfs-cluster-ansible bash

Then you can proceed with the instructions in Local usage.

Customization

You can customize the deployment of the GlusterFS cluster by copying config.dist.yml to config.yml and overriding the variables in it.

Supported configuration:

Variable	Description	Default value
`cluster_size`	Specifies the size of the cluster when deploying locally	1
`clients_size`	Specifies the number of clients when deploying locally	1
`dms_passphrase`	The passphrase used to deploy and configure DMS in the cluster's main node and clients	`very-secure-passphrase`
`extra_allowed_ips`	Extra IPs to be added to the glusterfs cluster firewall. Format `IP/mask`. Ex: `170.11.22.33/32`, `10.0.0.1/16`, `192.168.0.1/24`	`[]`
`local_dms_deb`	If set this binary is used instead of latest available via permalink (https://d.nunet.io/nunet-dms-amd64-latest.zip)	null

Environment variables:

Variable	Description	Default value
`GLUSTERFS_ROLE_CONFIG`	Config file to be used by glusterfs-ubuntu ansible role	`config.yml`
`GLUSTERFS_SSH_KEY`	SSH Key used to authenticate to the nodes	`glusterfs-key`
`GLUSTERFS_ANSIBLE_HOSTS`	Hosts file used by glusterfs-ubuntu ansible role	`hosts_glusterfs`

Remote usage

The remote cluster can be pre-provisioned using cloud-init.tpl.yml, you only have to change ##SSH_KEY with an ssh public key. The private key will be used by ansible to authenticate with the servers and clients for provisioning.

The ssh private and public keys should be named with the prefix glusterfs-key and be deployed in this project's root folder at test-suite/infrastrucure/glusterfs-cluster.

To create the key and populate the cloud-init-remote.yml run the following command:

ssh-keygen -q -t ed25519 -f glusterfs-key-remote -C glusterfs-key-remote -N ""
sed "s~##SSH_KEY~$(cat glusterfs-key-remote.pub)~g" cloud-init.tpl.yaml >cloud-init-remote.yaml

After provisioning the remote servers with cloud-init-remote.yml, populate the file hosts_glusterfs_remote according to the following structure:

[glusterfs_servers]
glusterfs-main-node ansible_host={server_ip} ansible_connection=ssh ansible_user=ubuntu ansible_python_interpreter=auto_silent ansible_become=true ansible_become_method=sudo ansible_ssh_private_key_file=/path/to/glusterfs-key-remote
# ...

[glusterfs_clients]
glusterfs-client ansible_host={client_ip} ansible_connection=ssh ansible_user=ubuntu ansible_python_interpreter=auto_silent ansible_become=true ansible_become_method=sudo ansible_ssh_private_key_file=/path/to/glusterfs-key-remote
# ...

You can also customize the role config specifically for production deployment. This is useful if adding for instance extra IPs to the glusterfs cluster firewall without affecting a local deployment. To do so, cp config.dist.yml config.prod.yml and export GLUSTERFS_ROLE_CONFIG=config.prod.yml and change the file accordingly.

To execute the ansible role glusterfs-ubuntu with the correct key and hosts file, populate GLUSTERFS_SSH_KEY and GLUSTERFS_ANSIBLE_HOSTS with the path to their respective files:

export GLUSTERFS_ROLE_CONFIG=config.prod.yml
export GLUSTERFS_SSH_KEY=$PWD/glusterfs-key-remote
export GLUSTERFS_ANSIBLE_HOSTS=$PWD/hosts_glusterfs_remote
ssh-agent bash run.sh

These assume you are running from test-suite/infrastructure/glusterfs-cluster.

Requirements​

System​

Dependencies​

Usage​

Local usage​

Using docker​

Customization​

Remote usage​