mariadb-galera

Clusterable, auto-discoverable MariaDB galera cluster.

Built with the useful topic from Withblue.ink.

Project description

Goal of this project is to create an easily deployable MariaDB cluster, that can scale up and down without configuration/setup assle. It uses HOSTNAME (or DOCKERCLOUD_CONTAINER_HOSTNAME on Docker Cloud) env vars to discover other hosts (mariadb-1, mariadb-2, mariadb-3, etc.). Container named with {name}-1 ({name-0} when on Kubernetes) will be the galera cluster creator.

The container is ready to use with Docker Cloud, Docker Compose and simple Docker commands. Support for Kubernetes is in beta.

Build and run

• Docker Cloud
• Docker Compose
• Kubernetes

Manual (to try it locally - manual deployments with Docker)

docker network create --driver bridge galera
docker run -it -e MYSQL_ROOT_PASSWORD=root --name=mariadb-1 -e HOSTNAME=mariadb-1 --rm --network galera -p 3306:3306 pukoren/mariadb-galera:latest
docker run -it -e MYSQL_ALLOW_EMPTY_PASSWORD=true --name=mariadb-2 -e HOSTNAME=mariadb-2 --rm --network galera --link mariadb-1:mariadb-1 pukoren/mariadb-galera:latest

Env vars

Name	Example	Description
HOSTNAME	mariadb-1	The container hostname. Container named {name}-1 will be the Galera master.
MYSQL_ROOT_PASSWORD	pass	The cluster root password.
MYSQL_ALLOW_EMPTY_PASSWORD	anything	Allow empty root password
MYSQL_RANDOM_ROOT_PASSWORD	anything	Generates a random root password
MYSQL_DATABASE	anything	A database name to create on first launch
MYSQL_USER	username	An user name to create on first launch. Must provide MYSQL_PASSWORD env var. If MYSQL_DATABASE is provided, it will be granted access to it
MARIADB_DEFAULT_STORAGE_ENGINE	InnoDB	The default storage engine to use for the node
GALERA_SLAVE_THREADS	1	Number of slave threads to start, see doc
MARIADB_INNODB_BUFFER_POOL_SIZE	134217728	The pool size of your InnoDB buffers (in bytes), see doc
MARIADB_INNODB_BUFFER_POOL_INSTANCES	1	The number of instances for your InnoDB buffers, see doc. Each should be at least 1GB of size (POOL_SIZE)

Things to know

Cluster shut down and clustering recovery

Never shut down the whole cluster - if you do so the nodes may not be able to recover. If for some reason you need to shut down the whole cluster and restart it later, make sure to delete the file named clustered located in shared volume of the first container ({name}-1 or {name}-0 in Kubernetes).

Always set up a rolling deployment process to avoid any clustering shutdown when updating.

Setup examples

Behind a reverse proxy (HAProxy), secured by IP filtering (Docker Cloud)

Creating a MariaDB Galera cluster behind a reverse proxy, that will load balancer and filter incoming requests based on source IP address:

mariadb-production:
  environment:
    - 'EXCLUDE_PORTS=4567,4568,4444'
    - 'EXTRA_SETTINGS=acl network_allowed src 192.168.1.0,block if !network_allowed'
    - MYSQL_ROOT_PASSWORD=password
    - TCP_PORTS=3306
  image: 'pukoren/mariadb-galera:latest'
  sequential_deployment: true
  volumes:
    - '/data/mysql:/data'
mariadb-production-lb:
  environment:
    - 'STATS_AUTH=user:pass'
    - STATS_PORT=1936
  image: 'dockercloud/haproxy:latest'
  links:
    - mariadb-production
  ports:
    - '1936:1936'
    - '3306:3306'
  roles:
    - global