table of contents
PG_CREATECLUSTER(1) | Debian PostgreSQL infrastructure | PG_CREATECLUSTER(1) |
NAME¶
pg_createcluster - create a new PostgreSQL cluster
SYNOPSIS¶
pg_createcluster [options] version name [-- initdb options]
DESCRIPTION¶
pg_createcluster creates a new PostgreSQL server cluster (i. e. a collection of databases served by a postgres(1) instance) and integrates it into the multi-version/multi-cluster architecture of the postgresql-common package.
Every cluster is uniquely identified by its version and name. The name can be arbitrary. The default cluster that is created on installation of a server package is main. However, you might wish to create other clusters for testing, with other superusers, a cluster for each user on a shared server, etc. pg_createcluster will abort with an error if you try to create a cluster with a name that already exists for that version.
For compatibility with systemd service units, the cluster name should not contain any dashes (-). pg_ctlcluster will warn about the problem, but succeed with the operation.
Given a major PostgreSQL version (like "8.2" or "8.3") and a cluster name, it creates the necessary configuration files in /etc/postgresql/version/name/; in particular these are postgresql.conf, pg_ident.conf, pg_hba.conf, a postgresql-common specific configuration file start.conf (see STARTUP CONTROL below), pg_ctl.conf, and a symbolic link log which points to the log file (by default, /var/log/postgresql/postgresql-version-name.log).
postgresql.conf is automatically adapted to use the next available port, i. e. the first port (starting from 5432) which is not yet used by an already existing cluster.
If the data directory does not yet exist, PostgreSQL's initdb(1) command is used to generate a new cluster structure. If the data directory already exists, it is integrated into the postgresql-common structure by moving the configuration file and setting the data_directory option. Please note that this only works for data directories which were created directly with initdb, i. e. all the configuration files (postgresql.conf etc.) must be present in the data directory.
If a custom socket directory is given and it does not exist, it is created.
If the log file does not exist, it is created. In any case the permissions are adjusted to allow write access to the cluster owner. Please note that postgresql.conf can be customized to specify log_directory and/or log_filename; if at least one of these options is present, then the symbolic link log in the cluster configuration directory is ignored.
If the default snakeoil SSL certificate exists (/etc/ssl/certs/ssl-cert-snakeoil.pem and /etc/ssl/private/ssl-cert-snakeoil.key), and the postgres user is in the ssl-cert Unix group, pg_createcluster configures the cluster to use this certificate, and enables SSL. Therefore all clusters will use the same SSL certificate by default. For versions up to 9.1, symlinks in the data directory will be created (server.crt and server.key); for 9.2 and later, the appropriate postgresql.conf options will be set (ssl_cert_file and ssl_key_file). Of course you can replace this with a cluster specific certificate. Similarly for /etc/postgresql-common/root.crt and /etc/postgresql-common/root.crl, these files will be configured as client certificate CA and revocation list, when present. (root.crt is initially a placeholder that will only be used if real certificates are added to the file.)
OPTIONS¶
- -u user, --user=user
- Set the user who owns the cluster and becomes the database superuser to the given name or uid. By default, this is the user postgres. A cluster must not be owned by root.
- -g group, --group=group
- Change the group of the cluster related data files. By default this will be the primary group of the database owner.
- -d dir, --datadir=dir
- Explicitly set the data directory path, which is used to store all the actual databases and tables. This will become quite big (easily in the order of five times the amount of actual data stored in the cluster). Defaults to /var/lib/postgresql/version/cluster.
- -s dir, --socketdir=dir
- Explicitly set the directory where the postgres(1) server stores the Unix socket for local connections. Defaults to /var/run/postgresql/ for clusters owned by the user postgres, and /tmp for clusters owned by other users. Please be aware that /tmp is an unsafe directory since everybody can create a socket there and impersonate the database server. If the given directory does not exist, it is created with appropriate permissions.
- -l path, --logfile=path
- Explicitly set the path for the postgres(1) server log file. Defaults to /var/log/postgresql/postgresql-version-cluster.log.
- --locale=locale
- Set the default locale for the database cluster. If this option is not specified, the locale is inherited from the environment that pg_createcluster runs in.
- --lc-collate=locale
- --lc-ctype=locale
- --lc-messages=locale
- --lc-monetary=locale
- --lc-numeric=locale
- --lc-time=locale
- Like --locale, but only sets the locale in the specified category.
- -e encoding, --encoding=encoding
- Select the encoding of the template database. This will also be the
default encoding of any database you create later, unless you override it
there. The default is derived from the locale, or SQL_ASCII if that does
not work. The character sets supported by the PostgreSQL server are
described in the documentation.
Note: It is not recommended to set this option directly! Set the locale instead.
- -p port, --port=port
- Select the port the new cluster listens on (for the Unix socket and the TCP port); this must be a number between 1024 and 65535, since PostgreSQL does not run as root and thus needs an unprivileged port number. By default the next free port starting from 5432 is assigned.
- -q --quiet --no-status
- Suppress output from initdb and (or only) the cluster status message at the end of the output.
- --start
- Immediately start a server for the cluster after creating it (i. e. call pg_ctlcluster version cluster start on it). By default, the cluster is not started.
- --start-conf=auto|manual|disabled
- Set the initial value in the start.conf configuration file. See STARTUP CONTROL below. By default, auto is used, which means that the cluster is handled by /etc/init.d/postgresql, i. e. starts and stops automatically on system boot.
- -o guc=value, --pgoption guc=value
- Configuration option to set in the new postgresql.conf file.
- --createclusterconf=file
- Alternative createcluster.conf file to use. Default is /etc/postgresql-common/createcluster.conf (or $PGSYSCONFDIR/createcluster.conf).
- --environment=file
- Alternative default environment file to use. Default is /etc/postgresql-common/environment (or $PGSYSCONFDIR/environment). If the file is missing, a placeholder string is used. %v and %c are replaced; see DEFAULT VALUES below.
- -- initdb options
- Options passed directly to initdb(1).
Per default, pg_createcluster will update the pg_hba.conf file generated by initdb to use peer authentication on local (unix) connections, and md5 on TCP (host) connections. If explicit authentication config is included here (-A, --auth, --auth-host, --auth-local), the pg_hba.conf file will be left untouched.
Note: If only one of --auth-host and --auth-local is provided, the other setting will default to trust as per initdb's defaults, opening a potential security risk.
STARTUP CONTROL¶
The start.conf file in the cluster configuration directory controls the start/stop behavior of that cluster's postgres process. The file can contain comment lines (started with '#'), empty lines, and must have exactly one line with one of the following keywords:
- auto
- The postgres process is started/stopped automatically in the init script.
When running from systemd, the cluster is started/stopped when postgresql.service is started/stopped. This is also the default if the file is missing.
- manual
- The postgres process is not handled by the init script, but manually
controlling the cluster with pg_ctlcluster(1) is permitted.
When running from systemd, the cluster is not started automatically when postgresql.service is started. However, stopping/restarting postgresql.service will stop/restart the cluster. The cluster can be started using systemctl start postgresql@version-cluster.
- disabled
- Neither the init script, pg_ctlcluster(1), nor postgresql@.service are permitted to start/stop the cluster. Please be aware that this will not stop the cluster owner from calling lower level tools to control the postgres process; this option is only meant to prevent accidents during maintenance, not more.
When running from systemd, invoke systemctl daemon-reload after editing start.conf.
The pg_ctl.conf file in the cluster configuration directory can contain additional options passed to pg_ctl of that cluster.
DEFAULT VALUES¶
Some default values used by pg_createcluster can be modified in /etc/postgresql-common/createcluster.conf. Occurrences of %v are replaced by the major version number, and %c by the cluster name. Use %% for a literal %.
- create_main_cluster (Default: true)
- Create a main cluster when a new postgresql-NN server package is installed.
- start_conf (Default: auto)
- Default start.conf value to use.
- data_directory (Default: /var/lib/postgresql/%v/%c)
- Default data directory.
- waldir|xlogdir (Default: unset)
- Default directory for transaction logs. When used, initdb will create a symlink from pg_wal (PostgreSQL 9.6 and earlier: pg_xlog) in the data directory to this location. Unset by default, i.e. transaction logs remain in the data directory. Both spellings of this option are accepted, and translated to the correct initdb invocation depending on the cluster version.
- initdb_options (Default: unset)
- Other options to pass to initdb.
- Other options
- All other options listed are copied into the new cluster's
postgresql.conf, e.g.:
listen_addresses = '*' log_line_prefix = '%%t '
Some postgresql.conf options are treated specially:
- ssl
- Only added to postgresql.conf if the default snakeoil certificates exist and are readable for the cluster owner as detailed above.
- stats_temp_directory
- Only added to postgresql.conf if existing, and writable for the cluster owner, or else if the parent directory is writable. Not used on PostgreSQL 15 or later.
- include
- include_if_exists
- include_dir
- createcluster.conf supports the same include directives as postgresql.conf.
- add_include
- add_include_if_exists
- add_include_dir
- To add include directives to the new postgresql.conf file, use the add_* directives. The add_ prefix is removed.
SEE ALSO¶
initdb(1), pg_ctlcluster(8), pg_lsclusters(1), pg_wrapper(1)
AUTHORS¶
Martin Pitt <mpitt@debian.org>, Christoph Berg <myon@debian.org>
2024-09-26 | Debian |