OpenStack-Ansible Galera server

OpenStack-Ansible Galera server

Ansible role to install and configure a Galera cluster powered by MariaDB

To clone or view the source code for this repository, visit the role repository for galera_server.

Default variables

# Set the package install state for distribution packages
# Options are 'present' and 'latest'
galera_server_package_state: "latest"

galera_cluster_members: "{{ groups['galera_all'] }}"
galera_server_bootstrap_node: "{{ galera_cluster_members[0] }}"
galera_ignore_cluster_state: false
galera_upgrade: false
galera_force_bootstrap: false

galera_wsrep_node_name: "{{ inventory_hostname }}"
galera_cluster_name: openstack_galera_cluster

# The galera server-id should be set on all cluster nodes to ensure
#  that replication is handled correctly and the error
#  "Warning: You should set server-id to a non-0 value if master_host is
#   set; we will force server id to 2, but this MySQL server will not act
#   as a slave." is no longer present.
# galera_server_id: 0

# These are here to stub out the internal ROLE API.
#  if these are used they should be set within the
#  distro specific variable files found in vars/
galera_debconf_items: []
galera_mariadb_server_package: "{{ _galera_mariadb_server_package }}"

# The major version used to select the repo URL path
galera_major_version: 10.1
galera_minor_version: 30

# Set the URL for the MariaDB repository
galera_repo_host: "downloads.mariadb.com"
galera_repo_url: "{{ _galera_repo_url }}"

# Set the repo information for the MariaDB repository
galera_repo: "{{ _galera_repo }}"

# Set the gpg keys needed to be imported
galera_gpg_keys: "{{ _galera_gpg_keys | default([]) }}"

# Set the rpo information for the Percona Xtrabackup repository
galera_percona_xtrabackup_repo: "{{ _galera_percona_xtrabackup_repo | default({}) }}"

# Enable the use of the upstream percona repo
use_percona_upstream: "{{ _use_percona_upstream }}"

galera_monitoring_user: monitoring
galera_monitoring_user_password: ""

# WARNING: Set this to open xinetd rules for galera monitoring.
# This is REQUIRED to run a working openstack-ansible deployment.
# If it's undefined the galera cluster state can't be reported,
# and haproxy would fail to do proper load balancing on the cluster.
# Because this opens connections to the cluster status, this
# should be restricted, which we do in the integrated build.
# Please override accordingly to your use case.
# This can be replaced with other hostnames, cidr, ips, and ips + wildcards.
#
#galera_monitoring_allowed_source: "0.0.0.0/0"

galera_root_user: root

# WARNING: This option is deprecated and will be removed in v12.0
galera_gcache_size: 1024M

galera_max_heap_table_size: 32M
galera_tmp_table_size: 32M

galera_file_limits: 65535
galera_wait_timeout: 3600
# Increase this value if large SST transfers cause mysql startup to fail due
# to timeout
galera_startup_timeout: 1800

## innodb options
galera_innodb_buffer_pool_size: 4096M
galera_innodb_log_file_size: 1024M
galera_innodb_log_buffer_size: 128M

## wsrep configuration
galera_wsrep_address: "{{ ansible_host }}"
galera_wsrep_address_port: "{{ galera_wsrep_address }}:3306"
galera_wsrep_cluster_address: >-
  {% set _var = [] -%}
  {% for cluster_host in galera_cluster_members -%}
  {% set _addr = hostvars[cluster_host]['galera_wsrep_address']
                 | default(hostvars[cluster_host]['ansible_host']) -%}
  {% if _var.append(_addr) %}{% endif -%}
  {% endfor -%}
  {# If only 1 cluster member is present output an empty string so the
     single-node member will re-bootstrap correctly upon restart #}
  {{ _var | join(',') if galera_cluster_members | length > 1 else '' }}
galera_wsrep_node_incoming_address: "{{ galera_wsrep_address }}"
## Cap the maximum number of threads / workers when a user value is unspecified.
galera_wsrep_slave_threads_max: 16
galera_wsrep_slave_threads: "{{ [[ansible_processor_vcpus|default(2), 2] | max, galera_wsrep_slave_threads_max] | min }}"
galera_wsrep_retry_autocommit: 3
galera_wsrep_debug: 0
galera_wsrep_sst_method: xtrabackup-v2
galera_wsrep_provider_options:
  - { option: "gcache.size", value: "{{ galera_gcache_size }}" }
galera_wsrep_sst_auth_user: "root"
galera_wsrep_sst_auth_password: "{{ galera_root_password  }}"

# Enable compression of backups.
galera_xtrabackup_compression: "{{ (ansible_architecture == 'x86_64') or (ansible_architecture == 'ppc64le' and ansible_distribution == 'Ubuntu') }}"
# xtrabackup parallel/compression/sync threads
galera_xtrabackup_threads: 4

galera_pip_packages:
  - MySQL-python
  - pycrypto

galera_pip_package_state: "latest"

# Galera slow/unindexed query logging
galera_slow_query_logging: 0
galera_unindexed_query_logging: 0

## Tunable overrides
galera_my_cnf_overrides: {}
galera_cluster_cnf_overrides: {}
galera_debian_cnf_overrides: {}

# Set the max connections value for galera. Set this value to override the
# computed value which is (100 x vCPUs) with a cap of 1600. If computed, the
# lowest value throughout the cluster will be used which is something to note
# if deploying galera on different hardware.
# galera_max_connections: 500

# This is only applied if the ansible_pkg_mgr is 'apt'
galera_server_distro_package_pins:
  - package: '*'
    release: MariaDB
    priority: 1001
  # Blacklist 10.1.27 release due to https://jira.mariadb.org/browse/MDEV-13908
  # which causes MariaDB to segfault with galera enabled.
  - package: mariadb-server
    version: 10.1.27+maria-1~xenial
    priority: -1
  # TODO(evrardjp): Remove pin when clustering issue is figured out
  # Clustering issue
  - package: mariadb-server
    version: 10.1.31+maria-1~xenial
    priority: -1

# Galera Server SSL functionality.

galera_use_ssl: false
galera_ssl_cert: /etc/ssl/certs/galera.pem
galera_ssl_key: /etc/mysql/ssl/galera.key
galera_ssl_ca_cert: /etc/ssl/certs/galera-ca.pem
## These options should be specified in user_variables if necessary, otherwise self-signed certs are used.
# galera_user_ssl_cert: /etc/openstack_deploy/self_signed_certs/galera.pem
# galera_user_ssl_key: /etc/openstack_deploy/self_signed_certs/galera.key
# galera_user_ssl_ca_cert: /etc/openstack_deploy/self_signed_certs/galera-ca.pem

## Set galera_ssl_self_signed_regen to true if you want to generate a new
# SSL certificate for Galera when this playbook runs.  You can also change
# the subject of the self-signed certificate here if you prefer.
galera_ssl_self_signed_regen: false
galera_ssl_self_signed_subject: "/C=US/ST=Texas/L=San Antonio/O=IT/CN={{ galera_address }}"
galera_ssl_ca_self_signed_subject: "/C=US/ST=Texas/L=San Antonio/O=IT"
# This option is used for creating the CA and overriding the Galera address on the clients side.
# Should be set to either public VIP of VIP FQDN, depending on what is currently used in the env.
galera_address: "{{ ansible_host }}"

# MariaDB 10.1+ ships with 'PrivateDevices=True' in the systemd unit file. This
# provides some additional security, but it causes problems with systemd 219.
# While the security enhancements are helpful on bare metal hosts with multiple
# services running, they are not as helpful when MariaDB is running in a
# container with its own isolated namespaces.
#
# Related bugs:
#   https://bugs.launchpad.net/openstack-ansible/+bug/1697531
#   https://github.com/lxc/lxc/issues/1623
#   https://github.com/systemd/systemd/issues/6121
#
# Setting the following variable to 'yes' will disable the PrivateDevices
galera_disable_privatedevices: "{{ _galera_disable_privatedevices }}"

## Set default mirror for openSUSE repositories
# NOTE(hwoarang): Ensure that the full path to the 'opensuse' directory is used.
#galera_server_opensuse_mirror_obs_url: "http://widehat.opensuse.org"

# Where the extra package download is executed from.
# Options are ['deployment-host', 'target-host']
galera_server_extra_package_downloader: "deployment-host"

# The location where the extra packages are downloaded to
galera_server_extra_package_path: "/opt/cache/files"

# Toggle whether certificate validation should be enabled/disabled
galera_server_extra_package_validate_certs: yes

Required variables

To use this role, define the following variables:

galera_root_password: secrete

Example playbook

- name: Install Galera server
  hosts: galera_all
  user: root
  roles:
    - { role: "galera_server" }
  vars:
    galera_root_password: secrete

External Restart Hooks

When the role performs a restart of the mariadb service, it will notify an Ansible handler named Manage LB, which is a noop within this role. In the playbook, other roles may be loaded before and after this role which will implement Ansible handler listeners for Manage LB, that way external roles can manage the load balancer endpoints responsible for sending traffic to the MariaDB servers being restarted by marking them in maintenance or active mode, draining sessions, etc. For an example implementation, please reference the ansible-haproxy-endpoints role used by the openstack-ansible project.

Creative Commons Attribution 3.0 License

Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.