Redis

The role is to deploy a redis cluster across multiple machines.

Requirements

Prior to running the roles, the following must be in place :

• Access to an RPM with the EPEL redis RPM (In testing we add the EPEL repo)
• A version of Ruby installed (In testing we use the Centos packages)

Optionaly:

• An Ansible group where the nodes are to be installed.
• An Ansible group where the management node can be installed

Role Usage

The role is currently broken into 3 parts:

• redis/core : to install the required components for redis
• redis/node : which will configure everything for a node, and get it running
• redis/cluster : Use to create the cluster from the individual nodes

There is also the redis role which only contains the common default, of which all other roles are dependant (see meta/main.yml)

Example 1: A single host with 3 master-nodes, no replication.

 - hosts: all
   become: true
   roles:
     - redis/core
     - { role: redis/node, redis_port: 7000 }
     - { role: redis/node, redis_port: 7001 }
     - { role: redis/node, redis_port: 7002 }
     - { role: redis/cluster, redis_cluster_replicas: 0, redis_node_list: "{{ groups['all'] | map('extract', hostvars, ['ansible_eth0', 'ipv4', 'address']) | arraypermute( [':'] ) | arraypermute( [7000,7001,7002] ) }}" }

Example 2: 3 host cluster (defined in redis-nodes), with 3 master nodes with 1 replicas (6 nodes total). Notice that we define a node as management to install the management application.

- hosts: redis-nodes
  become: true
  roles:
    - redis/core
    - { role: redis/node, redis_port: 7000 }
    - { role: redis/node, redis_port: 7001 }

- hosts: redis-mgt
  become: true
  roles:
    - { role: redis/cluster, redis_cluster_replicas: 1, redis_node_list: "{{ groups['redis-nodes'] | map('extract', hostvars, ['ansible_eth1', 'ipv4', 'address']) | arraypermute( [':'] ) | arraypermute( [7000,7001] ) | list }}" }

Role Variables

This role has been developed with the goal to simplify the deployment of a redis cluster across multiple nodes.

Redis/Core

The following user variable can be redefined for the redis/core.

Variable	Description	Default
redis_data_dir	The root of the data directory	/var/lib/redis
redis_log_dir	The root direcotry for redis logs	/var/log/redis
redis_run_dir	The root direcotry for redis runtime information	/var/run/redis
redis_conf_dir	The root directry for redis node configuraitons	/etc/

NOTE: The role currently only supports Centos / rpm based installs. The role will install the redis package - which creates some default configs and directories - but will be ignored

The following variables are also defined, but not usualy redefined in this role.:

Variable	Description	Default
redis_packages	A list of packages required to install redis	[ redis ]
redis_port	The default port that redis will listen to.	6379

NOTE: redis_port can be overridden in the redis/node deployment.

Redis/Node

The redis node role will configure and start a redis node on a specific port. Tasks performed by the redis/node role include:

• Creating the data sub-direcotries ( under redis_data_dir as redis_data_dir/)
• Configure a custom redis.conf file ( under redis_conf_dir as redis_conf_dir/redis_.conf )
• Configure a custom systemd service definition ( In system standard location as redis_.service )
• Add the listening ports to the SElinux security contexts ( for RPM based systems )
• Start the configured redis node.

The following variables can be redefined by users of this role:

Variable	Description	Default
redis_port	The port that redis will listen to.	6379

This port number is used to configure all above items. This is a sane default as the host cannot have more than one redis instance listening to a single port. The default redis shutdown script uses this information to derive the proper config.

Redis/cluster

Redis cluster will take a series of cluster nodes, and configure them in a single cluster based on configurable parameters. This has been done to simplefy the deployment of the cluster.

The follwing are commonly overridden when creating the cluster:

Variables	Description	Default
redis_cluster_replicas	The number of replicas each master requires	0
redis_node_list	a list of 'IP:PORT' endpoint combinations	see defaults

Other variables used that should not be overridden:

Variables	Description	Default
redis_conf_dir	to create a "cluster defined" flag for idempotency	see redis/core

Other notes:

• In the redis.io redis-trib ruby application, we added a --yes flag to accept the proposed configuration without waiting for user input.
• redis-trib does an inteligent distribution of the nodes based on number of replicas, number of nodes and hosts. One assumption is that different IP's are different hosts. It will rotate through the servers to increase availability in case of a host failure.

Dependencies

No external role or Ansible module dependancies.

License

BSD

Author Information

An optional section for the role authors to include contact information, or a website (HTML is not allowed).