NAME¶

barman - Backup and Recovery Manager for PostgreSQL

DESCRIPTION¶

Barman is an administration tool for disaster recovery of PostgreSQL servers written in Python and maintained by EnterpriseDB. Barman can perform remote backups of multiple servers in business critical environments and helps DBAs during the recovery phase.

CONFIGURATION FILE LOCATIONS¶

The system-level Barman configuration file is located at

/etc/barman.conf

or

/etc/barman/barman.conf

and is overridden on a per-user level by

$HOME/.barman.conf

CONFIGURATION FILE SYNTAX¶

The Barman configuration file is a plain INI file. There is a general section called [barman] and a section [servername] for each server you want to backup. Rows starting with ; are comments.

CONFIGURATION FILE DIRECTORY¶

Barman supports the inclusion of multiple configuration files, through the configuration_files_directory option. Included files must contain only server specifications, not global configurations. If the value of configuration_files_directory is a directory, Barman reads all files with .conf extension that exist in that folder. For example, if you set it to /etc/barman.d, you can specify your PostgreSQL servers placing each section in a separate .conf file inside the /etc/barman.d folder.

OPTIONS¶

active

When set to true (default), the server is in full operational state. When set to false, the server can be used for diagnostics, but any operational command such as backup execution or WAL archiving is temporarily disabled. When adding a new server to Barman, we suggest setting active=false at first, making sure that barman check shows no problems, and only then activating the server. This will avoid spamming the Barman logs with errors during the initial setup.

archiver

This option allows you to activate log file shipping through PostgreSQL's archive_command for a server. If set to true, Barman expects that continuous archiving for a server is in place and will activate checks as well as management (including compression) of WAL files that Postgres deposits in the incoming directory. Setting it to false (default), will disable standard continuous archiving for a server. Note: If neither archiver nor streaming_archiver are set, Barman will automatically set this option to true. This is in order to maintain parity with deprecated behaviour where archiver would be enabled by default. This behaviour will be removed from the next major Barman version.

archiver_batch_size

This option allows you to activate batch processing of WAL files for the archiver process, by setting it to a value > 0. Otherwise, the traditional unlimited processing of the WAL queue is enabled. When batch processing is activated, the archive-wal process would limit itself to maximum archiver_batch_size WAL segments per single run. Integer.

autogenerate_manifest

This option enables the auto-generation of backup manifest files for rsync based backups and strategies. The manifest file is a JSON file containing the list of files contained in the backup. It is generated at the end of the backup process and stored in the backup directory. The manifest file generated follows the format described in the postgesql documentation, and is compatible with the pg_verifybackup tool. The option is ignored if the backup method is not rsync.

aws_profile

The name of the AWS profile to use when authenticating with AWS (e.g. INI section in AWS credentials file).

aws_region

The name of the AWS region containing the EC2 VM and storage volumes defined by snapshot_instance and snapshot_disks.

azure_credential

The credential type (either azure-cli or managed-identity) to use when authenticating with Azure. If this is omitted then the default Azure authentication flow will be used.

azure_resource_group

The name of the Azure resource group to which the compute instance and disks defined by snapshot_instance and snapshot_disks belong. Required when the snapshot value is specified for backup_method and snapshot_provider is set to azure.

azure_subscription_id

The ID of the Azure subscription which owns the instance and storage volumes defined by snapshot_instance and snapshot_disks. Required when the snapshot value is specified for backup_method and snapshot_provider is set to azure.

backup_compression

The compression to be used during the backup process. Only supported when backup_method = postgres. Can either be unset or gzip,lz4, zstd or none. If unset then no compression will be used during the backup. Use of this option requires that the CLI application for the specified compression algorithm is available on the Barman server (at backup time) and the PostgreSQL server (at recovery time). Note that the lz4 and zstd algorithms require PostgreSQL 15 (beta) or later. Note that none compression will create an archive not compressed.

backup_compression_format

The format pg_basebackup should use when writing compressed backups to disk. Can be set to either plain or tar. If unset then a default of tar is assumed. The value plain can only be used if the server is running PostgreSQL 15 or later and if backup_compression_location is server. Only supported when backup_method = postgres.

backup_compression_level

An integer value representing the compression level to use when compressing backups. Allowed values depend on the compression algorithm specified by backup_compression. Only supported when backup_method = postgres.

backup_compression_location

The location (either client or server) where compression should be performed during the backup. The value server is only allowed if the server is running PostgreSQL 15 or later.

backup_compression_workers

The number of compression threads to be used during the backup process. Only supported when backup_compression = zstd is set. Default value is 0. In this case default compression behavior will be used.

backup_directory

Directory where backup data for a server will be placed.

backup_method

Configure the method barman used for backup execution. If set to rsync (default), barman will execute backup using the rsync command over SSH (requires ssh_command). If set to postgres barman will use the pg_basebackup command to execute the backup. If set to local-rsync, barman will assume to be running on the same server as the PostgreSQL instance and with the same user, then execute rsync for the file system copy. If set to snapshot, barman will use the API for the cloud provider defined in the snapshot_provider option to create snapshots of disks specified in the snapshot_disks option and save only the backup label and metadata to its own storage.

backup_options

This option allows you to control the way Barman interacts with PostgreSQL for backups. It is a comma-separated list of values that accepts the following options:

bandwidth_limit

This option allows you to specify a maximum transfer rate in kilobytes per second. A value of zero specifies no limit (default).

barman_home

Main data directory for Barman.

barman_lock_directory

Directory for locks. Default: %(barman_home)s.

basebackup_retry_sleep

Number of seconds of wait after a failed copy, before retrying Used during both backup and recovery operations. Positive integer, default 30.

basebackup_retry_times

Number of retries of base backup copy, after an error. Used during both backup and recovery operations. Positive integer, default 0.

basebackups_directory

Directory where base backups will be placed.

check_timeout

Maximum execution time, in seconds per server, for a barman check command. Set to 0 to disable the timeout. Positive integer, default 30.

cluster

Name of the Barman cluster associated with a Barman server or model. Used by Barman to group servers and configuration models that can be applied to them. Can be omitted for servers, in which case it defaults to the server name. Must be set for configuration models, so Barman knows the set of servers which can apply a given model.

compression

Standard compression algorithm applied to WAL files. Possible values are: gzip (requires gzip to be installed on the system), bzip2 (requires bzip2), pigz (requires pigz), pygzip (Python's internal gzip compressor) and pybzip2 (Python's internal bzip2 compressor).

config_changes_queue

Barman uses a queue to apply configuration changes requested through barman config-update command. This allows it to serialize multiple requests of configuration changes, and also retry an operation which has been abruptly terminated. This configuration option allows you to specify where in the filesystem the queue should be written. By default Barman writes to a file named cfg_changes.queue under barman_home.

conninfo

Connection string used by Barman to connect to the Postgres server. This is a libpq connection string, consult the PostgreSQL manual (https://www.postgresql.org/docs/current/static/libpq-connect.html#LIBPQ-CONNSTRING) for more information. Commonly used keys are: host, hostaddr, port, dbname, user, password.

create_slot

When set to auto and slot_name is defined, Barman automatically attempts to create the replication slot if not present. When set to manual (default), the replication slot needs to be manually created.

custom_compression_filter

Customised compression algorithm applied to WAL files.

custom_compression_magic

Customised compression magic which is checked in the beginning of a WAL file to select the custom algorithm. If you are using a custom compression filter then setting this will prevent barman from applying the custom compression to WALs which have been pre-compressed with that compression. If you do not configure this then custom compression will still be applied but any pre-compressed WAL files will be compressed again during WAL archive.

custom_decompression_filter

Customised decompression algorithm applied to compressed WAL files; this must match the compression algorithm.

description

A human readable description of a server.

errors_directory

Directory that contains WAL files that contain an error; usually this is related to a conflict with an existing WAL file (e.g. a WAL file that has been archived after a streamed one).

forward_config_path

Parameter which determines whether a passive node should forward its configuration file path to its primary node during cron or sync-info commands. Set to true if you are invoking barman with the -c/--config option and your configuration is in the same place on both the passive and primary barman servers. Defaults to false.

gcp_project

The ID of the GCP project which owns the instance and storage volumes defined by snapshot_instance and snapshot_disks. Required when the snapshot value is specified for backup_method and snapshot_provider is set to gcp.

gcp_zone

The name of the availability zone where the compute instance and disks to be backed up in a snapshot backup are located. Required when the snapshot value is specified for backup_method and snapshot_provider is set to gcp.

immediate_checkpoint

This option allows you to control the way PostgreSQL handles checkpoint at the start of the backup. If set to false (default), the I/O workload for the checkpoint will be limited, according to the checkpoint_completion_target setting on the PostgreSQL server. If set to true, an immediate checkpoint will be requested, meaning that PostgreSQL will complete the checkpoint as soon as possible.

incoming_wals_directory

Directory where incoming WAL files are archived into. Requires archiver to be enabled.

last_backup_maximum_age

This option identifies a time frame that must contain the latest backup. If the latest backup is older than the time frame, barman check command will report an error to the user. If empty (default), latest backup is always considered valid. Syntax for this option is: "i (DAYS | WEEKS | MONTHS)" where i is an integer greater than zero, representing the number of days | weeks | months of the time frame.

last_backup_minimum_size

This option identifies lower limit to the acceptable size of the latest successful backup. If the latest backup is smaller than the specified size, barman check command will report an error to the user. If empty (default), latest backup is always considered valid. Syntax for this option is: "i (k|Ki|M|Mi|G|Gi|T|Ti)" where i is an integer greater than zero, with an optional SI or IEC suffix. k=kilo=1000, Ki=Kibi=1024 and so forth. Note that the suffix is case-sensitive.

last_wal_maximum_age

This option identifies a time frame that must contain the latest WAL file archived. If the latest WAL file is older than the time frame, barman check command will report an error to the user. If empty (default), the age of the WAL files is not checked. Syntax is the same as last_backup_maximum_age (above).

lock_directory_cleanup

enables automatic cleaning up of the barman_lock_directory from unused lock files.

log_file

Location of Barman's log file.

log_level

Level of logging (DEBUG, INFO, WARNING, ERROR, CRITICAL).

max_incoming_wals_queue

Maximum number of WAL files in the incoming queue (in both streaming and archiving pools) that are allowed before barman check returns an error (that does not block backups). Default: None (disabled).

minimum_redundancy

Minimum number of backups to be retained. Default 0.

model

By default any section configured in the Barman configuration files define the configuration for a Barman server. If you set model = true in a section, that turns that section into a configuration model for a given cluster. Cannot be set as false.

network_compression

This option allows you to enable data compression for network transfers. If set to false (default), no compression is used. If set to true, compression is enabled, reducing network usage.

parallel_jobs

This option controls how many parallel workers will copy files during a backup or recovery command. Default 1. For backup purposes, it works only when backup_method is rsync.

parallel_jobs_start_batch_period

The time period in seconds over which a single batch of jobs will be started. Default: 1 second.

parallel_jobs_start_batch_size

Maximum number of parallel jobs to start in a single batch. Default: 10 jobs.

path_prefix

One or more absolute paths, separated by colon, where Barman looks for executable files. The paths specified in path_prefix are tried before the ones specified in PATH environment variable.

post_archive_retry_script

Hook script launched after a WAL file is archived by maintenance. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. In a post archive scenario, ABORT_STOP has currently the same effects as ABORT_CONTINUE.

post_archive_script

Hook script launched after a WAL file is archived by maintenance, after 'post_archive_retry_script'.

post_backup_retry_script

Hook script launched after a base backup. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. In a post backup scenario, ABORT_STOP has currently the same effects as ABORT_CONTINUE.

post_backup_script

Hook script launched after a base backup, after 'post_backup_retry_script'.

post_delete_retry_script

Hook script launched after the deletion of a backup. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. In a post delete scenario, ABORT_STOP has currently the same effects as ABORT_CONTINUE.

post_delete_script

Hook script launched after the deletion of a backup, after 'post_delete_retry_script'.

post_recovery_retry_script

Hook script launched after a recovery. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. In a post recovery scenario, ABORT_STOP has currently the same effects as ABORT_CONTINUE.

post_recovery_script

Hook script launched after a recovery, after 'post_recovery_retry_script'.

post_wal_delete_retry_script

Hook script launched after the deletion of a WAL file. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. In a post delete scenario, ABORT_STOP has currently the same effects as ABORT_CONTINUE.

post_wal_delete_script

Hook script launched after the deletion of a WAL file, after 'post_wal_delete_retry_script'.

pre_archive_retry_script

Hook script launched before a WAL file is archived by maintenance, after 'pre_archive_script'. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. Returning ABORT_STOP will propagate the failure at a higher level and interrupt the WAL archiving operation.

pre_archive_script

Hook script launched before a WAL file is archived by maintenance.

pre_backup_retry_script

Hook script launched before a base backup, after 'pre_backup_script'. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. Returning ABORT_STOP will propagate the failure at a higher level and interrupt the backup operation.

pre_backup_script

Hook script launched before a base backup.

pre_delete_retry_script

Hook script launched before the deletion of a backup, after 'pre_delete_script'. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. Returning ABORT_STOP will propagate the failure at a higher level and interrupt the backup deletion.

pre_delete_script

Hook script launched before the deletion of a backup.

pre_recovery_retry_script

Hook script launched before a recovery, after 'pre_recovery_script'. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. Returning ABORT_STOP will propagate the failure at a higher level and interrupt the recover operation.

pre_recovery_script

Hook script launched before a recovery.

pre_wal_delete_retry_script

Hook script launched before the deletion of a WAL file, after 'pre_wal_delete_script'. Being this a retry hook script, Barman will retry the execution of the script until this either returns a SUCCESS (0), an ABORT_CONTINUE (62) or an ABORT_STOP (63) code. Returning ABORT_STOP will propagate the failure at a higher level and interrupt the WAL file deletion.

pre_wal_delete_script

Hook script launched before the deletion of a WAL file.

primary_checkpoint_timeout

This defines the amount of seconds that Barman will wait at the end of a backup if no new WAL files are produced, before forcing a checkpoint on the primary server.

primary_conninfo

The connection string used by Barman to connect to the primary Postgres server during backup of a standby Postgres server. Barman will use this connection to carry out any required WAL switches on the primary during the backup of the standby. This allows backups to complete even when archive_mode = always is set on the standby and write traffic to the primary is not sufficient to trigger a natural WAL switch.

primary_ssh_command

Parameter that identifies a Barman server as passive. In a passive node, the source of a backup server is a Barman installation rather than a PostgreSQL server. If primary_ssh_command is specified, Barman uses it to establish a connection with the primary server. Empty by default, it can also be set globally.

recovery_options

Options for recovery operations. Currently only supports get-wal. get-wal activates generation of a basic restore_command in the resulting recovery configuration that uses the barman get-wal command to fetch WAL files directly from Barman's archive of WALs. Comma separated list of values, default empty.

recovery_staging_path

A path to a location on the recovery host (either the barman server or a remote host if --remote-ssh-command is also used) where files for a compressed backup will be staged before being uncompressed to the destination directory. Backups will be staged in their own directory within the staging path according to the following naming convention: "barman-staging-SERVER_NAME-BACKUP_ID". The staging directory within the staging path will be removed at the end of the recovery process. This option is required when recovering from compressed backups and has no effect otherwise.

retention_policy

Policy for retention of periodic backups and archive logs. If left empty, retention policies are not enforced. For redundancy based retention policy use "REDUNDANCY i" (where i is an integer > 0 and defines the number of backups to retain). For recovery window retention policy use "RECOVERY WINDOW OF i DAYS" or "RECOVERY WINDOW OF i WEEKS" or "RECOVERY WINDOW OF i MONTHS" where i is a positive integer representing, specifically, the number of days, weeks or months to retain your backups. For more detailed information, refer to the official documentation. Default value is empty.

retention_policy_mode

Currently only "auto" is implemented.

reuse_backup

This option controls incremental backup support. Possible values are:

slot_name

Physical replication slot to be used by the receive-wal command when streaming_archiver is set to on. Default: None (disabled).

snapshot_disks

A comma-separated list of disks which should be included in a backup taken using cloud snapshots. Required when the snapshot value is specified for backup_method.

snapshot_instance

The name of the VM or compute instance where the storage volumes are attached. Required when the snapshot value is specified for backup_method.

snapshot_provider

The name of the cloud provider which should be used to create snapshots. Required when the snapshot value is specified for backup_method. Supported values: gcp.

ssh_command

Command used by Barman to login to the Postgres server via ssh.

streaming_archiver

This option allows you to use the PostgreSQL's streaming protocol to receive transaction logs from a server. If set to on, Barman expects to find pg_receivewal (known as pg_receivexlog prior to PostgreSQL 10) in the PATH (see path_prefix option) and that streaming connection for the server is working. This activates connection checks as well as management (including compression) of WAL files. If set to off (default) barman will rely only on continuous archiving for a server WAL archive operations, eventually terminating any running pg_receivexlog for the server. Note: If neither streaming_archiver nor archiver are set, Barman will automatically set archiver to true. This is in order to maintain parity with deprecated behaviour where archiver would be enabled by default. This behaviour will be removed from the next major Barman version.

streaming_archiver_batch_size

This option allows you to activate batch processing of WAL files for the streaming_archiver process, by setting it to a value > 0. Otherwise, the traditional unlimited processing of the WAL queue is enabled. When batch processing is activated, the archive-wal process would limit itself to maximum streaming_archiver_batch_size WAL segments per single run. Integer.

streaming_archiver_name

Identifier to be used as application_name by the receive-wal command. Only available with pg_receivewal (or pg_receivexlog >= 9.3). By default it is set to barman_receive_wal.

streaming_backup_name

Identifier to be used as application_name by the pg_basebackup command. By default it is set to barman_streaming_backup.

streaming_conninfo

Connection string used by Barman to connect to the Postgres server via streaming replication protocol. By default it is set to conninfo.

streaming_wals_directory

Directory where WAL files are streamed from the PostgreSQL server to Barman. Requires streaming_archiver to be enabled.

tablespace_bandwidth_limit

This option allows you to specify a maximum transfer rate in kilobytes per second, by specifying a comma separated list of tablespaces (pairs TBNAME:BWLIMIT). A value of zero specifies no limit (default).

wal_conninfo

A connection string which, if set, will be used by Barman to connect to the Postgres server when checking the status of the replication slot used for receiving WALs. If left unset then Barman will use the connection string defined by wal_streaming_conninfo. If wal_conninfo is set but wal_streaming_conninfo is unset then wal_conninfo will be ignored.

wal_retention_policy

Policy for retention of archive logs (WAL files). Currently only "MAIN" is available.

wal_streaming_conninfo

A connection string which, if set, will be used by Barman to connect to the Postgres server when receiving WAL segments via the streaming replication protocol. If left unset then Barman will use the connection string defined by streaming_conninfo for receiving WAL segments.

wals_directory

Directory which contains WAL files.

HOOK SCRIPTS¶

The script definition is passed to a shell and can return any exit code.

The shell environment will contain the following variables:

BARMAN_CONFIGURATION

configuration file used by barman

BARMAN_ERROR

error message, if any (only for the 'post' phase)

BARMAN_PHASE

'pre' or 'post'

BARMAN_RETRY

1 if it is a retry script (from 1.5.0), 0 if not

BARMAN_SERVER

name of the server

Backup scripts specific variables:

BARMAN_BACKUP_DIR

backup destination directory

BARMAN_BACKUP_ID

ID of the backup

BARMAN_PREVIOUS_ID

ID of the previous backup (if present)

BARMAN_NEXT_ID

ID of the next backup (if present)

BARMAN_STATUS

status of the backup

BARMAN_VERSION

version of Barman

Archive scripts specific variables:

BARMAN_SEGMENT

name of the WAL file

BARMAN_FILE

full path of the WAL file

BARMAN_SIZE

size of the WAL file

BARMAN_TIMESTAMP

WAL file timestamp

BARMAN_COMPRESSION

type of compression used for the WAL file

Recovery scripts specific variables:

BARMAN_DESTINATION_DIRECTORY

the directory where the new instance is recovered

BARMAN_TABLESPACES

tablespace relocation map (JSON, if present)

BARMAN_REMOTE_COMMAND

secure shell command used by the recovery (if present)

BARMAN_RECOVER_OPTIONS

recovery additional options (JSON, if present)

Only in case of retry hook scripts, the exit code of the script is checked by Barman. Output of hook scripts is simply written in the log file.

EXAMPLE¶

Here is an example of configuration file:

[barman]
; Main directory
barman_home = /var/lib/barman
; System user
barman_user = barman
; Log location
log_file = /var/log/barman/barman.log
; Default compression level
;compression = gzip
; Incremental backup
reuse_backup = link
; 'main' PostgreSQL Server configuration
[main]
; Human readable description
description =  "Main PostgreSQL Database"
; SSH options
ssh_command = ssh postgres@pg
; PostgreSQL connection string
conninfo = host=pg user=postgres
; PostgreSQL streaming connection string
streaming_conninfo = host=pg user=postgres
; Minimum number of required backups (redundancy)
minimum_redundancy = 1
; Retention policy (based on redundancy)
retention_policy = REDUNDANCY 2

SEE ALSO¶

barman (1).

AUTHORS¶

Barman maintainers (in alphabetical order):

•

Abhijit Menon-Sen

•

Jane Threefoot

•

Michael Wallace

Past contributors (in alphabetical order):

•

Anna Bellandi (QA/testing)

•

Britt Cole (documentation reviewer)

•

Carlo Ascani (developer)

•

Francesco Canovai (QA/testing)

•

Gabriele Bartolini (architect)

•

Gianni Ciolli (QA/testing)

•

Giulio Calacoci (developer)

•

Giuseppe Broccolo (developer)

•

Jonathan Battiato (QA/testing)

•

Leonardo Cecchi (developer)

•

Marco Nenciarini (project leader)

•

Niccolò Fei (QA/testing)

•

Rubens Souza (QA/testing)

•

Stefano Bianucci (developer)

RESOURCES¶

•

Homepage: <https://www.pgbarman.org/>

•

Documentation: <https://docs.pgbarman.org/>

•

Professional support: <https://www.enterprisedb.com/>

COPYING¶

Barman is the property of EnterpriseDB UK Limited and its code is distributed under GNU General Public License v3.

© Copyright EnterpriseDB UK Limited 2011-2023

AUTHORS¶

EnterpriseDB <https://www.enterprisedb.com>.