ObieBent / postgresql

PostgreSQL

Ansible role to managed PostgreSQL. This role used official PostgreSQL repository http://download.postgresql.org/pub/repos/ instead of version available in distribution.

This role have been tested in CentOS 7 with PostgreSQL version 9.2, 9.3, 9.4, 9.5, 9.6, 10 and 11.

Requirements

The only external dependency is the python bindings for PostgreSQL, which this role will install.

Role Variables

Variable	Default value	Descriptions
postgresql_version	9.6	PostgreSQL version to deploy. Supported version: 9.2 to 9.6, 10 and 11
postgresql_service_name	postgresql-{{ postgresql_version }}	Name of postgresql service for the init
postgresql_admin_user	postgres	Name of the administrative user for
postgresql_admin_group	postgresql_admin_user	Name of the administrative user for
postgresql_listen_addresses	0.0.0.0	Listen address postgresql service bind to
postgresql_locale	en_US.UTF-8	Default locale
postgresql_conf_dir	/etc/postgresql	Default configuration directory
postgresql_data_dir	/var/lib/pgsql/data	Path to the PostgreSQL data directory
postgresql_enable_cron_backup	no	Whether to deploy a cron to database backups
postgresql_backup_dir	/var/backup/postgresql	Path to where database backups should be
postgresql_backup_age	0	Maximum age of backups to keep in days
postgresql_backup_compression	gzip	Compression to use for backups

Note: Custom user configuration must be defined in postgresql_conf_dir/conf.d/02-custom.conf file which override default PostgreSQL settings.

Monitoring user options :

Variable	Default value	Descriptions
postgresql_monitoring_enable	False	Boolean to enable monitoring user
postgresql_monitoring_user	centreon	PostgreSQL monitoring username
postgresql_monitoring_pass	autogenerated	PostgreSQL monitoring assword

Kernel tuning options for postgresql :

Variable	Default value	Descriptions
postgresql_sysctl_vm_dirty_background_ratio	1	Avoid letting too much data waiting in memory (vm.dirty_background_ratio)
postgresql_sysctl_vm_dirty_ratio	10	Avoid saturation of the disk during write time (vm.dirty_ratio)
postgresql_sysctl_vm_swappiness	10	Push down the swap usage, which is not good for postgresql (vm.swappiness)
postgresql_sysctl_vm_zone_reclaim_mode	0	This allocation mechanism is useless for postgresql (vm.zone_reclaim_mode)
postgresql_sysctl_kernel_shmmax	100% of RAM	Maximum size of shared memory segment in bytes (kernel.shmmax)
postgresql_sysctl_kernel_shmall	4194304	Total amount of shared memory available in pages of 4K (kernel.shmall)

PostgreSQL tuning options:

Variable	Default value	Descriptions
postgresql_shared_buffers_prct	0.25	Shared buffers is the area where PostgreSQL keeps a cache of data blocks, contrary to MySQL, you must not put all your memory there, because PostgreSQL is intended to be used with the OS-level filesystem cache. Using too much memory here will lead to inefficient cache duplication.
postgresql_shared_buffers	25% of RAM
postgresql_work_mem	1MB	This is the amount of memory that PostgreSQL can use for sort operation, hash tables (GROUP BY...) etc. before switching to a temporary table, the default value is usually fine unless you have really big tables. In which case you might want to increase this parameter. Be careful however, because each client can allocate work_mem more than once, so putting too much here might make your server run out of memory.
postgresql_maintenance_work_mem_prct	0.05	This is the amount of memory that PostgreSQL can use for heavy operations like ALTER, VACCUUM, etc... Not many of them will run simultaneously, so you can put a bigger value than work_mem here.
postgresql_maintenance_work_mem	5% of RAM
postgresql_checkpoint_segments	16	The number of WALs that can be filled before a checkpoint is triggered.
postgresql_effective_cache_size_prct	0.5	This parameter is a bit different than others, it will not cause any memory allocation, it is only used to help the query planner by telling how much memory it can count on. You can fine-tune it by observing how much OS-level cache is available during normal server operation, and adding it to postgresql_shared_buffers.
postgresql_effective_cache_size	50% of RAM
postgresql_log_min_duration_statement	500	Log query ran for at least the specified number of milliseconds.

You can define client authentication with what is traditionally named pg_hba.conf. There are two variables to control this: postgresql_pg_hba_conf_default and postgresql_pg_hba_conf

Variable	Default value	Descriptions
postgresql_pg_hba_conf_default	[]	A list of allowed clients.keys type, database, user,
		address, and method
postgresql_pg_hba_conf	empty list	This is defined the same as postgresql_pg_hba_conf_default
postgresql_auth_method_default	md5	postgresql_auth_method_default

Defaults value:

postgresql_pg_hba_conf_default:
 - { type: local, database: all, user: "{{postgresql_admin_user}}", address: '',             method: peer  }
 - { type: host,  database: all, user: all,                         address: "127.0.0.1/32", method: md5 }
 - { type: host,  database: all, user: all,                         address: "::1/128",      method: md5 }

You can define what databases and users to create using postgresql_databases and postgresql_users.

Variable	Default value	Descriptions
postgresql_databases	[]	A list of databases to manage
name		name of the database. It mandatory
owner		owner of the database
encoding	UTF-8	language encoding
lc_collate	en_US.UTF-8	Collation order
lc_ctype	lc_collate	Character classification, defaults to same as lc_collate
template	template0	template used to create the database
state	present

Example

postgresql_databases:
  - name: example

Variable	Default value	Descriptions
postgresql_users	[]	A list of users to manage
name		name of the database user.It mandatory
password		the user's password. It mandatory
encrypted		a boolean for whether the password is stored hashed
priv		privileges to grant user on a database
db		name of database privileges are associated with
role_attr_flags		role attributes for the user
state	present

Variable	Default value	Description
use_satellite_repo	false	Use official internet repository to install package of this role. In case host have no internet access and package must be install from a local satellite, set this option to true. Repository must be previously declare in satellite and VM must be register to it before deploy this role.
manage_by_cluster	False	Variable set to true if service state must be managed by a cluster solution (like PCS)

Example

postgresql_users:
  - name: icarus
    password: poor-password
    db: example
    priv: ALL
  - name: another_user
    password: poor-password
    db: anotherdb
    priv: ALL

Dependencies

No dependency on other roles.

Example Playbook

- hosts: servers
  roles:
    - role: postgresql
      postgresql_data_dir: /srv/postgresql
      postgresql_databases:
        - name: "odoo"
      postgresql_users:
        - name: odoo
          password: "y6uC65m91CMO2"
          db: "odoo"
          priv: ALL

License

BSD

Author Information

• ASSOGBA Boris [e-mail]