Using PostgreSQL on Debian and Ubuntu
- Installing the PostgreSQL Server
- Installing the PostgreSQL Management Tools
- Configuring the PostgreSQL Server
- Enabling Remote Access
- Creating Databases
- Database Maintenance
- Database Backups
- Using Multiple PostgreSQL Clusters
- Upgrading PostgreSQL
These instructions assume that you are using either Debian or Ubuntu, but should be adaptable to other operating systems. Ubuntu provides versions of the Debian PostgreSQL packages, and so has the same maintenance facilities as Debian.
The current releases of Ubuntu (8.04 and above) includes version 8.3 of PostgreSQL, rather than PostgreSQL 8.1, so amend the commands below accordingly.
Privileged Commands: Commands that require root (administrative) privileges are shown with the prefix #. On Ubuntu, simply replace the # with sudo.
The setup process installs the PostgreSQL server software and configures a cluster – a group of databases managed by a common service. The default cluster is called main, and the service for it runs on TCP port 5432. If you need additional clusters you may create them at any time.
The main cluster initially has three databases: postgres (for management purposes), template0 (a template database without any customizations) and template1 (a template database that you may customize). When you create your own databases they are generated as copies of template1.
Each PostgreSQL cluster uses a set of accounts, known as login roles, which are independent of the users and groups on the host system. For security and accountability, each user and application should have an individual login role to access the databases in the cluster. Login roles may belong to group roles, in the same way that system accounts may belong to groups.
The default configuration rejects all connections from other systems, and uses ident authentication to manage access from users on the same system. Ident security means that users are not prompted for a password when they login from the local system, instead they are either connected, with the login role that has the same name as their system account, or their connection is rejected if no login role exists to match their account.
You may easily enable secure remote access, if you require it, and instructions are below. Current versions of PostgreSQL also support Kerberos and LDAP, which allows clusters to use the central authentication and security facilities of more complex networks. Kerberos and LDAP integration are not covered here.
Like most SQL server products, PostgreSQL groups tables and other objects held in the databases into schemas. Only assign an object to the public schema if that object must be accessible to every user and application. Typically, you create schemas with the same names as login roles or group roles, and then assign database objects to the appropriate schemas.
To manage your PostgreSQL databases, the psql utility provides a command-line shell, and the pgAdmin III application offers a graphical interface. You must use separate utilities to create and upgrade clusters (see below).
PostgreSQL is highly standards-compliant, and also provides the additional features that you would expect from a enterprise SQL database. You may use popular programming languages within PostgreSQL functions, as well as standard SQL. Version 8.3 includes XML support and full-text search as part of the core system, and these features were available as extensions in previous releases.
Version 9.0 of PostgreSQL includes support for single-master replication. Use Bucardo, Slony-I, or a commercial replication product to support more advanced scenarios or older versions of PostgreSQL.
The development team prefer more specialized features to remain separate, rather than simply being bundled with the main software in the form of one preset solution. For example, the PostGIS extension adds support for spatial data. Similarly, database drivers and the extensions for third-party programming languages are maintained separately.
For community add-ons and extensions to PostgreSQL, refer to the PGFoundry Web site:
Installing the PostgreSQL Server
On a Debian server:
# apt-get install postgresql-8.1 postgresql-client-8.1 postgresql-contrib-8.1
On an Ubuntu server:
# apt-get install postgresql-8.3 postgresql-client-8.3 postgresql-contrib-8.3
Ubuntu also has packages for Slony-I replication and the PostGIS spacial data extensions:
# apt-get install postgresql-8.3-slony1
# apt-get install postgresql-8.3-postgis
Installing the PostgreSQL Management Tools
To manage a PostgreSQL service from an Ubuntu workstation, install these packages:
# apt-get install pgadmin3 postgresql-client-8.3 postgresql-doc-8.3
This installs the graphical management tool pgAdmin III, the command-line utilities, and the official documentation in HTML format.
Enter this URI in your Web browser to read the documentation:
If you use Debian stable, install these packages:
# apt-get install pgadmin3 postgresql-client-8.1 postgresql-doc-8.1
Configuring the PostgreSQL Server
First, set a strong password for the postgres role. This role automatically has unrestricted access to the cluster and everything held within those databases, so set a password for this role. To avoid potential risk, do this as soon as you have installed PostgreSQL, even if you do not currently intend to enable remote access, and are sure that no other users can login on the local system.
As the configuration defaults to ident authentication for local access, we must use the system account postgres to login with the postgres role. Enter this command at the server:
sudo -u postgres psql
Once logged in to the SQL interface, set a password for the postgres role:
ALTER ROLE postgres WITH ENCRYPTED PASSWORD 'mypassword';
You need this password to connect to the PostgreSQL server remotely with the postgres role, as described below.
To exit the PostgreSQL shell, type:
To enable some of the functions of the pgAdmin utility, you must run a script against the postgres database:
sudo -u postgres psql -d postgres < /usr/share/postgresql/8.1/contrib/admin81.sql
Use /usr/share/postgresql/8.3/contrib/adminpack.sql on current versions of Ubuntu.
Enabling Remote Access
If you wish to enable remote access, modify the cluster settings. The configuration files for each cluster are in the directory /etc/postgresql/version/clustername/, e.g. /etc/postgresql/8.1/main/.
Edit the file postgresql.conf, and remove the comment marker on the line for the listen_addresses setting, so that it reads:
listen_addresses = '*'
Open the file pg_hba.conf:
# nano /etc/postgresql/8.1/main/pg_hba.conf
To permit users to connect from remote systems on your network with any role, add this line:
host all all 192.168.1.0/24 md5
Replace 192.168.1.0/24 with the appropriate subnet definition for your network.
This enables md5 authentication, which means that login roles are secured with passwords that PostgreSQL itself stores in an encrypted form, and that PostgreSQL will require a valid password for any remote connection to use a role. After you make this change, ident authentication remains enabled for local logins.
You may, of course, change the local line in pg_hba.conf to disable ident authentication. Make sure that you can actually login to your PostgreSQL cluster with the postgres role and a password first!
To make your changes take effect, restart the service:
# /etc/init.d/postgresql-8.1 restart
Once the service restarts, you may access your PostgreSQL service from remote systems, either using tools such as psql or pgAdmin III, or with applications that support SQL. The Debian version of PostgreSQL is configured to automatically protect network connections with SSL, so that all communication between the server and remote clients is encrypted.
By default, the postgres system account on Debian is locked, and you should not unlock it. Cracking tools now try to use postgres, root, and other well-known system account names when they attempt to gain access to UNIX-like operating systems.
Under no circumstance should you enable PostgreSQL ident authentication for any remote access. The ident system cannot safely verify or guarantee the identity of any user on a remote system.
You may either use the createdb command-line utility to create new databases, or enter the appropriate SQL statement.
Remember that PostgreSQL does not create databases with UTF-8 by default. If you require Unicode support you must explicitly specify the UTF-8 encoding when you create the database:
# postgres createdb my_database -u -E UTF-8 --password
The corresponding SQL statement is:
CREATE DATABASE 'my_database' WITH ENCODING 'UTF-8';
You must manually set up an appropriate backup routine (see below).
The standard configuration enables the PostgreSQL auto-vacuum service by default, to ensure that the health of your databases is maintained. This typically just works without any intervention or extra configuration being required. If very large volumes of data are frequently added to or removed from your databases you may adjust the auto-vacuum settings to be more aggressive, in order to keep your clusters well optimized.
The Debian packages also install a scheduled task to automatically update the database statistics and run a standard vacuum each day if the auto-vacuum service is not running on the system (for whatever reason). In this case, edit /etc/cron.d/postgresql-common/ to enable a weekly full VACUUM. These tasks use the utility /usr/sbin/pg_maintenance, which you may also run manually. Refer to the man page for pg_maintenance for more information.
The correct backup strategy for your databases depends on their size and rate of change. For smaller databases, the best solution is to simply use the supplied pg_dumpall utility. Refer to the documentation for a detailed explanation of backup options:
Using Multiple PostgreSQL Clusters
Debian includes additional facilities to support multiple PostgreSQL clusters on the same system. Refer to the file /usr/share/doc/postgresql-common/architecture.html for a brief explanation, and the relevant man pages for more details:
- pg_createcluster – utility to create new clusters
- pg_lsclusters – utility to display the clusters on the system
- pg_ctlcluster – utility to manage clusters
- pg_upgradecluster – utility to upgrade clusters
- pg_dropcluster – utility to remove clusters
- user_clusters – configuration file to specify the default cluster for users and groups
Clusters may use different versions of PostgreSQL. Install the packages for the appropriate PostgreSQL versions, and specify the version that you require when you make a new cluster with pg_createcluster.
To safely upgrade a PostgreSQL cluster, use the pg_upgradecluster utility. This command actually creates a new cluster with the data and configuration from the specified cluster. The utility also automatically reconfigures the existing database cluster to use a different network port. The original copy of the specified cluster is not altered in any other way.