Default variable details

Some of debops.postgresql_server default variables have more extensive configuration than simple strings or lists, here you can find documentation and examples for them.

postgresql_server__preferred_version

By default the role installs the PostgreSQL version preferred by the APT package manager. This behaviour is influenced by how the PostgreSQL is packaged in Debian - each version has its own set of packages with the version as a suffix, and there's a set of metapackages which depend on the version available in the distribution (by default only 1 version is available).

Multiple PostgreSQL versions become available after enabling the upstream APT repository. To choose a different version than the default one, you need to set two variables in the inventory:

postgresql_server__preferred_version
The value of this variable should be set as the version of the PostgreSQL you wish the role to manage (it does not influence the APT packages the role installs, but what version is used in different file/directory paths managed by the role, what features are enabled/disabled in the configuration, etc.).
postgresql_server__base_packages
This is a list of APT packages which will be used by the role to install PostgreSQL. By default, it contains the metapackages which install the highest available version of PostgreSQL packages. To select a different version, you need to change the list of packages.

For example, to install PostgreSQL 9.3 instead of the default available version, in inventory you need to define:

postgresql_server__upstream: True
postgresql_server__preferred_version: '9.3'
postgresql_server__base_packages:
  - 'postgresql-9.3'
  - 'postgresql-client-9.3'
  - 'postgresql-contrib-9.3'

Remember that role does not support management of multiple PostgreSQL versions at the same time. The above variables should be defined in the inventory at all times, otherwise role might revert to the default PostgreSQL packages and version, and break your installation. This also is true for server upgrades. The preferred way to make an upgrade is to configure a new database server with desired PostgreSQL version and move the database to it.

You might also need to set similar set of variables for the debops.postgresql role to keep both of the roles in sync. Refer to its documentation for details.

postgresql_server__hba_*

Host-Based Authentication configuration in debops.postgresql_server Ansible role is specified in a set of lists:

Each PostgreSQL cluster by default uses all of the above lists in its pg_hba.conf configuration file. A cluster can disable any list by specifying its abbrevated name as a parameter with False. For example:

postgresql_server__cluster_main:
  name: 'main'
  port: '5432'
  hba_replication: False
  hba_public: False
  hba_trusted: False
  hba_local: False
  hba_remote: False

Above configuration will disable connections by trusted users (all users will be required to specify a password) and from remote clients.

Additionally, each cluster can specify its own HBA entries using item.hba parameter which will be added at the end of the pg_hba.conf file. By disabling selected global lists and adding custom entries you can redefine the HBA configuration file as needed. Example:

postgresql_server__cluster_main:
  name: 'main'
  port: '5432'
  hba_remote: False

  hba:
    - comment: 'Custom remote entries'
      type: 'hostssl'
      database: 'all'
      user: 'all'
      address: [ '192.0.2.0/24' ]
      method: 'md5'

Each entry in a HBA list is a YAML dictionary with parameters:

comment
Optional. Comment added to a given entry in pg_hba.conf file.
type

Required. Specifies connection type to use for a given entry. Known types are:

  • local: local connections by UNIX accounts
  • host: remote TCP connections, either with or without SSL
  • hostssl: remote TCP connections, SSL is required
  • hostnossl: remote TCP connections, plaintext is required
database

Required. String or a list of database names that are controlled by a given HBA entry. You can use special names:

  • all: all databases in a cluster
  • sameuser: database with the same name as the PostgreSQL role
  • samerole: all databases accessible by a given PostgreSQL role
  • @name: file with a list of database names, relative to a given cluster's configuration directory in /etc
user

Required. String or a list of PostgreSQL roles that are controlled by a given HBA entry. You can use special names:

  • all: all roles in on the PostgreSQL cluster
  • +role: a specified role and all roles that are included in it
  • @name: file with a list of roles, relative to a given cluster's configuration directory in /etc
  • *postgres*: a custom debops.postgresql_server name, it will be replaced by the UNIX system account name that manages a given cluster, usually postgres
address

Required by all types other than local. A string or list of IP addresses or CIDR networks (debops.postgresql_server does not support ip/netmask notation). You can use special names:

  • all: any network clients
  • samenet: any IP address from a subnet the host is directly connected to
method
Required. Authentication method used by this HBA entry. You most likely either want peer for local connections or md5 for remote connections. There are also other methods available, see the PostgreSQL documentation for information about how to use them.
options
Optional. List of additional options specific to a given authentication method.

You can find different examples of how to defined HBA lists in defaults/main.yml file of debops.postgresql_server role.

postgresql_server__ident_*

Ident maps stored in pg_ident.conf configuration file is used to map local UNIX accounts to PostgreSQL roles. This can be used to control what UNIX accounts can login to the PostgreSQL server as a given role.

Ident maps should only be used by the local UNIX accounts with the peer authentication method. Using them for ident method with remote clients is unreliable and discouraged - ident protocol is not meant to be used for authentication or authorization.

By default, PostgreSQL clusters managed by the debops.postgresql_server role use global lists of ident maps:

  • postgresql_server__ident_system: a user mapping which specifies which system users can login as the postgres superuser role.
  • postgresql_server__ident_trusted: this user mapping can be used with the "trusted" HBA list to specify which local UNIX accounts can login without specifying a password. It's not set by default.
  • postgresql_server__ident_local: this user mapping can be used to define local UNIX accounts globally for all clusters. It's not set by default.

Above ident maps can be disabled in a given cluster by specifying their abbvevated names in a parameter with False value. Example:

postgresql_server__cluster_main:
  name: 'main'
  port: '5432'
  ident_trusted: False
  ident_local: False

You can specify custom lists of ident maps in a PostgreSQL cluster configuration:

postgresql_server__cluster_main:
  name: 'main'
  port: '5432'
  ident_local: False

  ident:
    - map: 'main_local'
      user: [ 'user1', 'user2' ]
      role: 'role1'

Each ident map entry is a YAML dictionary with parameters:

map
Required. Name of the user map, can be repeated in different entries.
user

Required. String or list of UNIX user accounts to use in this map. You can use a regexp to specify accounts in various ways, see PostgreSQL documentation for more information.

Special string *postgres* will be replaced by Ansible to the owner of the PostgreSQL cluster, usually postgres.

role

Optional. String or list of PostgreSQL roles to map to the UNIX accounts.

If defined, specifies the PostgreSQL role to map to a given UNIX accounts.

If not defined, each entry role name will be the same as the UNIX account name. Don't use this option with regexp user entries.

Special string *postgres* will be replaced by Ansible to the owner of the PostgreSQL cluster, usually postgres.

Examples can be found in the defaults/main.yml file of the debops.postgresql_server Ansible role.

postgresql_server__clusters

On Debian and its derivatives, PostgreSQL installation is based around "clusters", each cluster being run on a particular PostgreSQL version and on a specific TCP port. debops.postgresql_server is designed to be used within that system, and allows you to create separate PostgreSQL clusters. A default <version>/main cluster will be created, based on default PostgreSQL version installed on a given host.

You can create and manage separate PostgreSQL clusters using postgresql_server__clusters list. Each cluster is defined as a YAML dict with at least two parameters - name and port. You should take care to always use separate port for each cluster you define. Role will create an entry for each cluster in /etc/services as well as maintain firewall configuration as needed.

Some of the global variables defined in debops.postgresql_server concerning clusters can be overriden on a cluster by cluster basis using their abbrevated names (without postgresql_server__ prefix) as cluster parameters. In addition, almost all of the PostgreSQL parameters found in the postgresql.conf configuration file can be specified as cluster parameters as well, to change the defaults.

List of some of the parameters that you can specify in a cluster configuration entry:

name
Required. Name of the cluster, used to separate different clusters based on the same PostgreSQL version.
port
Required. TCP port to use for a given cluster. Default PostgreSQL port is 5432, more clusters usually use the next port number available.
version
Optional. PostgreSQL version to use for a given cluster. If it's not specified, default detected version will be used, which is usually what you want.
environment

Optional. Dictionary which specifies environment variables and their values that should be set for a given PostgreSQL cluster. Example:

postgresql_server__cluster_main:
  name: 'main'
  port: '5432'

  environment:
    HOME: '/var/lib/postgresql'
    SHELL: '/bin/bash'
listen_addresses
List of network interfaces specified by their addresses a given cluster should bind to. If not set, global value of postgresql_server__listen_addresses will be used instead.
allow
List of IP addresses or CIDR subnets which should be allowed to connect to a given cluster.