Deploy the BeeGFS HA cluster

08/13/2024 Contributors

PDFs

Specify what tasks should run to deploy the BeeGFS HA cluster using a playbook.

Overview

This section covers how to assemble the standard playbook used to deploy/manage BeeGFS on NetApp.

Steps

Create the Ansible Playbook

Create the file playbook.yml and populate it as follows:

First define a set of tasks (commonly referred to as a play) that should only run on NetApp E-Series block nodes. We use a pause task to prompt before running the installation (to avoid accidental playbook runs), then import the nar_santricity_management role. This role handles applying any general system configuration defined in group_vars/eseries_storage_systems.yml or individual host_vars/<BLOCK NODE>.yml files.

- hosts: eseries_storage_systems
  gather_facts: false
  collections:
    - netapp_eseries.santricity
  tasks:
    - name: Verify before proceeding.
      pause:
        prompt: "Are you ready to proceed with running the BeeGFS HA role? Depending on the size of the deployment and network performance between the Ansible control node and BeeGFS file and block nodes this can take awhile (10+ minutes) to complete."
    - name: Configure NetApp E-Series block nodes.
      import_role:
        name: nar_santricity_management

Define the play that will run against all file and block nodes:

- hosts: all
  any_errors_fatal: true
  gather_facts: false
  collections:
    - netapp_eseries.beegfs

Within this play we can optionally define a set of "pre-tasks" that should run before deploying the HA cluster. This can be useful to verify/install any prerequisites like Python. We can also inject any pre-flight checks, for example verifying the provided Ansible tags are supported:

  pre_tasks:
    - name: Ensure a supported version of Python is available on all file nodes.
      block:
        - name: Check if python is installed.
          failed_when: false
          changed_when: false
          raw: python --version
          register: python_version

        - name: Check if python3 is installed.
          raw: python3 --version
          failed_when: false
          changed_when: false
          register: python3_version
          when: 'python_version["rc"] != 0 or (python_version["stdout"] | regex_replace("Python ", "")) is not version("3.0", ">=")'

        - name: Install python3 if needed.
          raw: |
            id=$(grep "^ID=" /etc/*release* | cut -d= -f 2 | tr -d '"')
            case $id in
              ubuntu) sudo apt install python3 ;;
              rhel|centos) sudo yum -y install python3 ;;
              sles) sudo zypper install python3 ;;
            esac
          args:
            executable: /bin/bash
          register: python3_install
          when: python_version['rc'] != 0 and python3_version['rc'] != 0
          become: true

        - name: Create a symbolic link to python from python3.
          raw: ln -s /usr/bin/python3 /usr/bin/python
          become: true
          when: python_version['rc'] != 0
      when: inventory_hostname not in groups[beegfs_ha_ansible_storage_group]

    - name: Verify any provided tags are supported.
      fail:
        msg: "{{ item }} tag is not a supported BeeGFS HA tag. Rerun your playbook command with --list-tags to see all valid playbook tags."
      when: 'item not in ["all", "storage", "beegfs_ha", "beegfs_ha_package", "beegfs_ha_configure", "beegfs_ha_configure_resource", "beegfs_ha_performance_tuning", "beegfs_ha_backup", "beegfs_ha_client"]'
      loop: "{{ ansible_run_tags }}"

Lastly, this play imports the BeeGFS HA role for the version of BeeGFS you want to deploy:

  tasks:
    - name: Verify the BeeGFS HA cluster is properly deployed.
      import_role:
        name: beegfs_ha_7_4 # Alternatively specify: beegfs_ha_7_3.

A BeeGFS HA role is maintained for each supported major.minor version of BeeGFS. This allows users to choose when they want to upgrade major/minor versions. Currently either BeeGFS 7.3.x (beegfs_7_3) or BeeGFS 7.2.x (beegfs_7_2) are supported. By default both roles will deploy the latest BeeGFS patch version at the time of release, though users can opt to override this and deploy the latest patch if desired. Refer to the latest upgrade guide for more details.

Optional: If you wish to define additional tasks, keep in mind if the tasks should be directed to all hosts (including the E-Series storage systems) or only the file nodes. If needed define a new play specifically targeting file nodes using - hosts: ha_cluster.

Click here for an example of a complete playbook file.

Install the NetApp Ansible Collections

The BeeGFS collection for Ansible and all dependencies are maintained on Ansible Galaxy. On your Ansible control node run the following command to install the latest version:

ansible-galaxy collection install netapp_eseries.beegfs

Though not typically recommended, it is also possible to install a specific version of the collection:

ansible-galaxy collection install netapp_eseries.beegfs:==<MAJOR>.<MINOR>.<PATCH>

Run the Playbook

From the directory on your Ansible control node containing the inventory.yml and playbook.yml files, run the playbook as follows:

ansible-playbook -i inventory.yml playbook.yml

Based on the size of the cluster the initial deployment can take 20+ minutes. If the deployment fails for any reason, simply correct any issues (e.g., miscabling, node wasn't started, etc.), and restart the Ansible playbook.

When specifying common file node configuration, if you choose the default option to have Ansible automatically manage connection based authentication, a connAuthFile used as a shared secret can now be found at <playbook_dir>/files/beegfs/<sysMgmtdHost>_connAuthFile (by default). Any clients needing to access the file system will need to use this shared secret. This is handled automatically if clients are configured using the BeeGFS client role.