Ansible Role: Percona XtraDB Cluster

Installs and configures Percona XtraDB Cluster (only 5.7 at this stage) on EL 8/9. This role takes in consideration a standard Galera architecture: three nodes minimum of which two act as server and one acts as arbitrator.

This role, at this stage, does not support:

• node scaling: you cannot re-run this role on a pre-installed architecture to add one or more nodes.
• pmm: You can add PMM Client manually to your architecture.
• ssl: You can encrypt your PXC traffic manually

Some approaches used in this role are definitely inspired by the excellent work of csuka/ansible_role_percona_xtradb_cluster, available here. But the implementation needs were different, so that we decided to open a new project: other that, the csuka project supports only EL8 and Percona 8.0. Hopefully, in a near future, these project could be merged.

Requirements

No special requirements; note that this role requires root access, so either run it in a playbook with a global become: yes, or invoke the role in your playbook like:

- hosts: database
  roles:
    - role: bmeme.percona_xtradb_cluster
      become: yes

Copy the variables from default/main.yml and adjust them for each node in it's variable folder. Take a look at the molecule/default/group_vars and molecule/default/host_vars folder to have an idea of how to configure every single node of your cluster architecture.

Installation

This is an Ansible role distributed using Ansible Galaxy. In order to install this role you can use the following command.

$ ansible-galaxy install bmeme.percona_xtradb_cluster

Update

If you want to update the role, you need to pass --force parameter when installing. Please, check the following command:

$ ansible-galaxy install --force bmeme.percona_xtradb_cluster

## Clustering When there are at least 2 nodes in the play, a multi-master cluster is possible. A 2 node galera cluster is a fundamentally broken design, as it cannot maintain uptime without a quorum and the ability of a node to go down to aid recovery.

Role Variables

percona_cluster:
  role: master
  bootstrapper: false
  server_id: '1'
  ip_address: "{{ ansible_default_ipv4.address }}"

The percona_cluster object is what you really need to configure for each node:

• role: string. Defines what is the role of the specific node. Could be master or arbiter. If you have three or more nodes dedicated to your cluster, all of them could be master. If you want to have only two nodes in the cluster and use an arbiter (installable on any other server in your infrastructure), you will have two nodes as master and one as arbiter
• bootstrapper: boolean. Defines the node responsible for bootstrapping the cluster. Pay attention: there must only be one bootstrapper and cannot be the arbiter!!! There is no assertion in the role to check any of that (at this stage).
• server_id: int. Defines the unique ID of a node. Pay attention: this is a unique id! Do not use the same ID for more than a node. There isn't any assertion in the role to check this (always at this stage).
• ip_address: string. Defines the IP Address of the node. To automatically calculate IPv4 address of a node, use {{ ansible_default_ipv4.address }} ansible variable as used in default/main.yml file.

percona_cluster_name: "galera-cluster"

Defines the name you want to have your cluster. Change it at your need.

percona_root_user: "root"
percona_root_password: "S3cr3tS/$"
percona_sst_user: "sstuser"
percona_sst_password: "S3cr3tS/$"

Defines root user/password and sst user/password. sst is the user responsible for replication transactions between nodes.

percona_databases:
  - mynewbranddb

percona_users:
  - name: mynewbranddb_user
    password: "yie?XuiNaedu"
    privs: 'mynewbranddb.*:ALL,GRANT'
    state: present
    encrypted: false
    host: '%'

Create databases and/or users during the role execution. Leave blank to skip. Databases and users are created only on bootstrapper node.

# Datafile directory.
mysql_datadir: "/var/lib/mysql"

# Address on which mysql bind to
mysql_bind_address: "{{ percona_cluster.ip_address }}"

# Mysql symbolic links configuration
mysql_symbolic_links: 0

# pxc_strict_mode allowed values: DISABLED,PERMISSIVE,ENFORCING,MASTER
mysql_cluster_pxc_strict_mode: ENFORCING

Basic Mysqld configuration.

# All specific configuration items
# Example:
#
# mysql_cluster_configuration: |-
#   max_allowed_packet=64M
#
mysql_cluster_configuration: ""

This allows to add custom configuration to mysqld.conf file, stored in /etc/percona-xtradb-cluster.conf.d directory. Add mysqld configuration items exactly as if you were putting them in mysqld.conf file.

Here an example of advanced configurations:

mysql_cluster_configuration: |-
  max_allowed_packet=64M
  some_other_conf=some_other_value

Supported Linux distro and versions

• EL 8
• EL 9

EOL 7 is no longer officially supported. As far as we know, the role can currently continue to work properly on EL 7.9; however, we're uncertain whether it works on lower versions.

Dependencies

N/A

Example Playbook

- hosts: db-servers
  become: yes
  roles:
    - { role: bmeme.percona_xtradb_cluster }

Inside group_vars/all.yml:

percona_databases:
  - mynewbranddb

percona_users:
  - name: mynewbranddb_user
    password: "yie?XuiNaedu"
    privs: 'mynewbranddb.*:ALL,GRANT'
    state: present
    encrypted: false
    host: '%'

Inside host_vars/master01.yml:

percona_cluster:
  role: master
  bootstrapper: true
  server_id: '1'
  ip_address: "{{ ansible_default_ipv4.address }}"

Inside host_vars/master02.yml:

percona_cluster:
  role: master
  bootstrapper: false
  server_id: '2'
  ip_address: "{{ ansible_default_ipv4.address }}"

Inside host_vars/arbiter.yml:

percona_cluster:
  role: arbiter
  bootstrapper: false
  server_id: '3'
  ip_address: "{{ ansible_default_ipv4.address }}"

GitHub Actions and molecule test

We couldn't run the role's test steps via Molecule in the GitHub Action workflow. This because (for unknown reasons) executing a molecule test fails due to the inability to start MySQL with the following error:

fatal: [test-cluster-03]: FAILED! => {"changed": false, "msg": "Unable to start service mysql: Job for mysql.service failed because the control process exited with error code.\nSee \"systemctl status mysql.service\" and \"journalctl -xe\" for details.\n"}

All molecule tests run outside the GitHub Action environment succeeded. As soon as this issue is resolved, we'll add the tests to the CI. Meanwhile, if you can assist us in debugging the problem, we would be infinitely and eternally grateful!

License

MIT / BSD

Author Information

This role was created in 2022 by Bmeme. It is actually maintained by Daniele Piaggesi and Roberto Mariani.

Credits

This role was really inspired by csuka Ansible Role XtraDB Cluster.