barman-cloud-wal-restore(1)

BARMAN-CLOUD-WAL-RESTORE(1)

Barman

BARMAN-CLOUD-WAL-RESTORE(1)

NAME

barman-cloud-wal-restore - Barman-cloud Commands

Synopsis

barman-cloud-wal-restore


                [ { -V | --version } ]


                [ --help ]


                [ { { -v | --verbose } | { -q | --quiet } } ]


                [ { -t | --test } ]


                [ --cloud-provider { aws-s3 | azure-blob-storage | google-cloud-storage } ]


                [ --endpoint-url ENDPOINT_URL ]


                [ --addressing-style { auto | virtual | path } ]


                [ { -P | --aws-profile } AWS_PROFILE ]


                [ --profile AWS_PROFILE ]


                [ --read-timeout READ_TIMEOUT ]


                [ { --azure-credential | --credential } { azure-cli | managed-identity


                  | default } ]


                [ --no-partial ]


                [ { -p, --parallel } PARALLEL ]


                [ --spool-dir SPOOL_DIR ]


                SOURCE_URL SERVER_NAME WAL_NAME WAL_DEST

Description

The barman-cloud-wal-restore script functions as the restore_command for retrieving WAL files from cloud storage and placing them directly into a Postgres standby server, bypassing the Barman server.

This script is used to download WAL files that were previously archived with the barman-cloud-wal-archive command. Disable automatic download of .partial files by calling --no-partial option.

Note:

Partial WAL files (with a .partial suffix) can appear in the object store when a standby server is archiving WAL segments and is then promoted to primary. At the moment of promotion, the in-progress WAL segment has not been completed yet, so Postgres archives it as a .partial file before switching to a new timeline. These files represent an incomplete segment and are not suitable for crash recovery on their own.

In most cases, downloading .partial files during WAL restore is harmless, because Postgres knows how to handle them correctly. However, when running barman-cloud-wal-restore on a cluster that has experienced a promotion, the restore_command may retrieve a .partial (after renaming it without the suffix) file where a complete WAL is expected.

Use --no-partial to suppress the download of .partial files in this scenario, which will cause restore_command to signal to Postgres that the WAL is not available, prompting it to switch to the correct timeline via a timeline history file (.history) instead.

Important:

On the target Postgres node, when pg_wal and the spool directory are on the same filesystem, files are moved via renaming, which is faster than copying and deleting. This speeds up serving WAL files significantly. If the directories are on different filesystems, the process still involves copying and deleting, so there's no performance gain in that case.

Note:

For GCP, only authentication with GOOGLE_APPLICATION_CREDENTIALS env is supported. To use an alternative GCP universe (e.g. S3NS/T-Systems), set the GOOGLE_CLOUD_UNIVERSE_DOMAIN environment variable to the desired universe domain.

Parameters

SERVER_NAME: Name of the server that will have WALs restored.
SOURCE_URL: URL of the cloud source, such as a bucket in AWS S3. For example: s3://bucket/path/to/folder.
WAL_NAME: The value of the '%f' keyword (according to restore_command).
WAL_DEST: The value of the '%p' keyword (according to restore_command).
-V / --version: Show version and exit.
--help: show this help message and exit.
-v / --verbose: Increase output verbosity (e.g., -vv is more than -v).
-q / --quiet: Decrease output verbosity (e.g., -qq is less than -q).
-t / --test: Test cloud connectivity and exit.
--cloud-provider: The cloud provider to use as a storage backend.
Allowed options are:

aws-s3.
azure-blob-storage.
google-cloud-storage.

--no-partial: Do not download partial WAL files
-p / --parallel: Specifies the number of WAL files to fetch in parallel. Default is 0 (disabled). When set to a value greater than 1, fetches the requested WAL file along with the next N - 1 files simultaneously. The additional files are staged in a local spool directory (see --spool-dir) so that subsequent restore requests can be served immediately from local storage.
--spool-dir: Directory used for staging extra WALs fetched when using --parallel. Default is /var/tmp/walrestore.

Extra options for the AWS cloud provider

--endpoint-url: Override default S3 endpoint URL with the given one.
--addressing-style: The addressing style to use for S3 requests. This is useful when connecting to S3-compatible services that require a specific addressing style.
Allowed options are:

auto (default): Uses the addressing style determined by the underlying library.
virtual: Uses virtual-hosted style addressing (e.g., bucket-name.s3.amazonaws.com).
path: Uses path-style addressing (e.g., s3.amazonaws.com/bucket-name).

-P / --aws-profile: Profile name (e.g. INI section in AWS credentials file).
--profile (deprecated): Profile name (e.g. INI section in AWS credentials file) - replaced by --aws-profile.
--read-timeout: The time in seconds until a timeout is raised when waiting to read from a connection (defaults to 60 seconds).

Extra options for the Azure cloud provider

--azure-credential / --credential: Optionally specify the type of credential to use when authenticating with Azure. If omitted then Azure Blob Storage credentials will be obtained from the environment and the default Azure authentication flow will be used for authenticating with all other Azure services. If no credentials can be found in the environment then the default Azure authentication flow will also be used for Azure Blob Storage.
Allowed options are:

azure-cli.
managed-identity.
default.

Author

EnterpriseDB

Copyright

May 19, 2026

3.19