PostgreSQL Configuration

Users that are familiar with PostgreSQL are aware of the existence of the following two files to configure an instance:

  • postgresql.conf: main run-time configuration file of PostgreSQL
  • pg_hba.conf: clients authentication file

Due to the concepts of declarative configuration and immutability of the PostgreSQL containers, users are not allowed to directly touch those files. Configuration is possible through the postgresql section of the Cluster resource definition by defining custom postgresql.conf and pg_hba.conf settings via the parameters and the pg_hba keys.

These settings are the same across all instances.

Warning

Please don't use the ALTER SYSTEM query to change the configuration of the PostgreSQL instances in an imperative way. Changing some of the options that are normally controlled by the operator might indeed lead to an unpredictable/unrecoverable state of the cluster. Moreover, ALTER SYSTEM changes are not replicated across the cluster.

A reference for custom settings usage is included in the samples, see cluster-example-custom.yaml.

The postgresql section

The PostgreSQL instance in the pod starts with a default postgresql.conf file, to which these settings are automatically added:

listen_addresses = '*'
include custom.conf

The custom.conf file will contain the user-defined settings in the postgresql section, as in the following example:

  # ...
  postgresql:
    parameters:
      shared_buffers: "1GB"
  # ...

PostgreSQL GUCs: Grand Unified Configuration

Refer to the PostgreSQL documentation for more information on the available parameters, also known as GUC (Grand Unified Configuration).

The content of custom.conf is automatically generated and maintained by the operator by applying the following sections in this order:

  • Global default parameters
  • Default parameters that depend on the PostgreSQL major version
  • User-provided parameters
  • Fixed parameters

The global default parameters are:

dynamic_shared_memory_type = 'posix'
logging_collector = 'on'
log_destination = 'csvlog'
log_directory = '/controller/log'
log_filename = 'postgres'
log_rotation_age = '0'
log_rotation_size = '0'
log_truncate_on_rotation = 'false'
max_parallel_workers = '32'
max_replication_slots = '32'
max_worker_processes = '32'
shared_memory_type = 'mmap' # for PostgreSQL >= 12 only
wal_keep_size = '512MB' # for PostgreSQL >= 13 only
wal_keep_segments = '32' # for PostgreSQL <= 12 only
wal_sender_timeout = '5s'
wal_receiver_timeout = '5s'

Warning

It is your duty to plan for WAL segments retention in your PostgreSQL cluster and properly configure either wal_keep_size or wal_keep_segments, depending on the server version, based on the expected and observed workloads.

Alternatively, if the only streaming replication clients are the replica instances running in the High Availability cluster, you can take advantage of the replication slots feature, which adds support for replication slots at the cluster level. You can enable the feature with the replicationSlots.highAvailability option (for more information, please refer to the "Replication" section.)

Without replication slots nor continuous backups in place, configuring wal_keep_size or wal_keep_segments is the only way to protect standbys from falling out of sync. If a standby did fall out of sync it would produce error messages like: "could not receive data from WAL stream: ERROR: requested WAL segment ************************ has already been removed". This will require you to dedicate a part of your PGDATA, or the volume dedicated to storing WAL files, to keep older WAL segments for streaming replication purposes.

The following parameters are fixed and exclusively controlled by the operator:

archive_command = '/controller/manager wal-archive %p'
archive_mode = 'on'
full_page_writes = 'on'
hot_standby = 'true'
listen_addresses = '*'
port = '5432'
restart_after_crash = 'false'
ssl = 'on'
ssl_ca_file = '/controller/certificates/client-ca.crt'
ssl_cert_file = '/controller/certificates/server.crt'
ssl_key_file = '/controller/certificates/server.key'
unix_socket_directories = '/controller/run'
wal_level = 'logical'
wal_log_hints = 'on'

Since the fixed parameters are added at the end, they can't be overridden by the user via the YAML configuration. Those parameters are required for correct WAL archiving and replication.

Replication settings

The primary_conninfo, restore_command, and recovery_target_timeline parameters are managed automatically by the operator according to the state of the instance in the cluster.

primary_conninfo = 'host=cluster-example-rw user=postgres dbname=postgres'
recovery_target_timeline = 'latest'

Log control settings

The operator requires PostgreSQL to output its log in CSV format, and the instance manager automatically parses it and outputs it in JSON format. For this reason, all log settings in PostgreSQL are fixed and cannot be changed.

For further information, please refer to the "Logging" section.

Shared Preload Libraries

The shared_preload_libraries option in PostgreSQL exists to specify one or more shared libraries to be pre-loaded at server start, in the form of a comma-separated list. Typically, it is used in PostgreSQL to load those extensions that need to be available to most database sessions in the whole system (e.g. pg_stat_statements).

In CloudNativePG the shared_preload_libraries option is empty by default. Although you can override the content of shared_preload_libraries, we recommend that only expert Postgres users take advantage of this option.

Important

In case a specified library is not found, the server fails to start, preventing CloudNativePG from any self-healing attempt and requiring manual intervention. Please make sure you always test both the extensions and the settings of shared_preload_libraries if you plan to directly manage its content.

CloudNativePG is able to automatically manage the content of the shared_preload_libraries option for some of the most used PostgreSQL extensions (see the "Managed extensions" section below for details).

Specifically, as soon as the operator notices that a configuration parameter requires one of the managed libraries, it will automatically add the needed library. The operator will also remove the library as soon as no actual parameter requires it.

Important

Please always keep in mind that removing libraries from shared_preload_libraries requires a restart of all instances in the cluster in order to be effective.

You can provide additional shared_preload_libraries via .spec.postgresql.shared_preload_libraries as a list of strings: the operator will merge them with the ones that it automatically manages.

Managed extensions

As anticipated in the previous section, CloudNativePG automatically manages the content in shared_preload_libraries for some well-known and supported extensions. The current list includes:

  • auto_explain
  • pg_stat_statements
  • pgaudit

Some of these libraries also require additional objects in a database before using them, normally views and/or functions managed via the CREATE EXTENSION command to be run in a database (the DROP EXTENSION command typically removes those objects).

For such libraries, CloudNativePG automatically handles the creation and removal of the extension in all databases that accept a connection in the cluster, identified by the following query:

SELECT datname FROM pg_database WHERE datallowconn

Note

The above query also includes template databases like template1.

Enabling auto_explain

The auto_explain extension provides a means for logging execution plans of slow statements automatically, without having to manually run EXPLAIN (helpful for tracking down un-optimized queries).

You can enable auto_explain by adding to the configuration a parameter that starts with auto_explain. as in the following example excerpt (which automatically logs execution plans of queries that take longer than 10 seconds to complete):

  # ...
  postgresql:
    parameters:
      auto_explain.log_min_duration: "10s"
  # ...

Note

Enabling auto_explain can lead to performance issues. Please refer to the auto explain documentation

Enabling pg_stat_statements

The pg_stat_statements extension is one of the most important capabilities available in PostgreSQL for real-time monitoring of queries.

You can enable pg_stat_statements by adding to the configuration a parameter that starts with pg_stat_statements. as in the following example excerpt:

  # ...
  postgresql:
    parameters:
      pg_stat_statements.max: "10000"
      pg_stat_statements.track: all
  # ...

As explained previously, the operator will automatically add pg_stat_statements to shared_preload_libraries and run CREATE EXTENSION IF NOT EXISTS pg_stat_statements on each database, enabling you to run queries against the pg_stat_statements view.

Enabling pgaudit

The pgaudit extension provides detailed session and/or object audit logging via the standard PostgreSQL logging facility.

CloudNativePG has transparent and native support for PGAudit on PostgreSQL clusters. For further information, please refer to the "PGAudit" logs section.

You can enable pgaudit by adding to the configuration a parameter that starts with pgaudit. as in the following example excerpt:

#
postgresql:
  parameters:
    pgaudit.log: "all, -misc"
    pgaudit.log_catalog: "off"
    pgaudit.log_parameter: "on"
    pgaudit.log_relation: "on"
#

The pg_hba section

pg_hba is a list of PostgreSQL Host Based Authentication rules used to create the pg_hba.conf used by the pods.

Since the first matching rule is used for authentication, the pg_hba.conf file generated by the operator can be seen as composed of four sections:

  1. Fixed rules
  2. User-defined rules
  3. Optional LDAP section
  4. Default rules

Fixed rules:

local all all peer

hostssl postgres streaming_replica all cert
hostssl replication streaming_replica all cert

Default rules:

host all all all <default-authentication-method>

From PostgreSQL 14 the default value of the password_encryption database parameter is set to scram-sha-256. Because of that, the default authentication method is scram-sha-256 from this PostgreSQL version.

PostgreSQL 13 and older will use md5 as the default authentication method.

The resulting pg_hba.conf will look like this:

local all all peer

hostssl postgres streaming_replica all cert
hostssl replication streaming_replica all cert

<user defined rules>
<user defined LDAP>

host all all all scram-sha-256 # (or md5 for PostgreSQL version <= 13)

Refer to the PostgreSQL documentation for more information on pg_hba.conf.

Inside the cluster manifest, pg_hba lines are added as list items in spec.postgresql.pg_hba, as in the following excerpt:

  postgresql:
    pg_hba:
      - hostssl app app 10.244.0.0/16 md5

In the above example we are enabling access for the app user to the app database using MD5 password authentication (you can use scram-sha-256 if you prefer) via a secure channel (hostssl).

LDAP Configuration

Under the postgres section of the cluster spec there is an optional ldap section available to define an LDAP configuration to be converted into a rule added into the pg_hba.conf file.

This will support two modes: simple bind mode which requires specifying a server, prefix and suffix in the LDAP section and the search+bind mode which requires specifying server, baseDN, binDN, and a bindPassword which is a secret containing the ldap password. Additionally, in search+bind mode you have the option to specify a searchFilter or searchAttribute. If no searchAttribute is specified the default one of uid will be used.

Additionally, both modes allow the specification of a scheme for ldapscheme and a port. Neither scheme nor port are required, however.

This section filled out for search+bind could look as follows:

postgresql:
  parameters:
    ldap:
      server: 'openldap.default.svc.cluster.local'
      bindSearchAuth:
        baseDN: 'ou=org,dc=example,dc=com'
        bindDN: 'cn=admin,dc=example,dc=com'
        bindPassword:
          name: 'ldapBindPassword'
          key: 'data'
        searchAttribute: 'uid'

Changing configuration

You can apply configuration changes by editing the postgresql section of the Cluster resource.

After the change, the cluster instances will immediately reload the configuration to apply the changes. If the change involves a parameter requiring a restart, the operator will perform a rolling upgrade.

Dynamic Shared Memory settings

PostgreSQL supports a few implementations for dynamic shared memory management through the dynamic_shared_memory_type configuration option. In CloudNativePG we recommend to limit ourselves to any of the following two values:

  • posix: which relies on POSIX shared memory allocated using shm_open (default setting)
  • sysv: which is based on System V shared memory allocated via shmget

In PostgreSQL, this setting is particularly important for memory allocation in parallel queries. For details, please refer to this thread from the pgsql-general mailing list.

POSIX shared memory

The default setting of posix should be enough in most cases, considering that the operator automatically mounts a memory-bound EmptyDir volume called shm under /dev/shm. You can verify the size of such volume inside the running Postgres container with:

mount | grep shm

You should get something similar to the following output:

shm on /dev/shm type tmpfs (rw,nosuid,nodev,noexec,relatime,size=******)

System V shared memory

In case your Kubernetes cluster has a high enough value for the SHMMAX and SHMALL parameters, you can also set:

dynamic_shared_memory_type: "sysv"

You can check the SHMMAX/SHMALL from inside a PostgreSQL container, by running:

ipcs -lm

For example:

------ Shared Memory Limits --------
max number of segments = 4096
max seg size (kbytes) = 18014398509465599
max total shared memory (kbytes) = 18014398509481980
min seg size (bytes) = 1

As you can see, the very high number of max total shared memory recommends setting dynamic_shared_memory_type to sysv.

An alternate method is to run:

cat /proc/sys/kernel/shmall
cat /proc/sys/kernel/shmmax

Fixed parameters

Some PostgreSQL configuration parameters should be managed exclusively by the operator. The operator prevents the user from setting them using a webhook.

Users are not allowed to set the following configuration parameters in the postgresql section:

  • allow_system_table_mods
  • archive_cleanup_command
  • archive_command
  • archive_mode
  • bonjour
  • bonjour_name
  • cluster_name
  • config_file
  • data_directory
  • data_sync_retry
  • event_source
  • external_pid_file
  • full_page_writes
  • hba_file
  • hot_standby
  • ident_file
  • jit_provider
  • listen_addresses
  • log_destination
  • log_directory
  • log_file_mode
  • log_filename
  • log_rotation_age
  • log_rotation_size
  • log_truncate_on_rotation
  • logging_collector
  • port
  • primary_conninfo
  • primary_slot_name
  • promote_trigger_file
  • recovery_end_command
  • recovery_min_apply_delay
  • recovery_target
  • recovery_target_action
  • recovery_target_inclusive
  • recovery_target_lsn
  • recovery_target_name
  • recovery_target_time
  • recovery_target_timeline
  • recovery_target_xid
  • restart_after_crash
  • restore_command
  • shared_preload_libraries
  • ssl
  • ssl_ca_file
  • ssl_cert_file
  • ssl_ciphers
  • ssl_crl_file
  • ssl_dh_params_file
  • ssl_ecdh_curve
  • ssl_key_file
  • ssl_max_protocol_version
  • ssl_min_protocol_version
  • ssl_passphrase_command
  • ssl_passphrase_command_supports_reload
  • ssl_prefer_server_ciphers
  • stats_temp_directory
  • synchronous_standby_names
  • syslog_facility
  • syslog_ident
  • syslog_sequence_numbers
  • syslog_split_messages
  • unix_socket_directories
  • unix_socket_group
  • unix_socket_permissions
  • wal_level
  • wal_log_hints