Instead of relying on traditional backup tools like tar and rsync, PgBackRest implements all backup features internally and uses a custom protocol for communicating with remote systems. Removing reliance on tar and rsync allows for better solutions to database-specific backup issues. The custom remote protocol limits the types of connections that are required to perform a backup which increases security.
PgBackRest uses the gitflow model of development. This means that the master branch contains only the release history, i.e. each commit represents a single release and release tags are always from the master branch. The dev branch contains a single commit for each feature or fix and more accurately depicts the development history. Actual development is done on feature (dev_*) branches and squashed into dev after regression tests have passed. In this model dev is considered stable and can be released at any time. As such, the dev branch does not have any special version modifiers.
PgBackRest is written entirely in Perl and uses some non-standard modules that must be installed from CPAN. All examples below are for PostgreSQL 9.3 but should be easily adaptable to any recent version.
CAVEAT: Installing 8.4 with the 9.X series appears to break libpq.
### Regression Test Setup
* Create the backrest user
The backrest user must be created on the same system and in the same group as the user you will use for testing (which can be any user you prefer). For example:
```
adduser -g <test-user-group> backrest
```
* Setup password-less SSH login between the test user and the backrest user
The test user should be able to `ssh backrest@127.0.0.1` and the backrest user should be able to `ssh <testuser>@127.0.0.1` without requiring any passwords. This article (http://archive.oreilly.com/pub/h/66) has details on how to accomplish this. Do the logons both ways at the command line before running regression tests.
* Give group read and execute permissions to `~/backrest/test`:
Usually this can be accomplished by running the following as the test user:
```
chmod 750 ~
```
* Running regression:
Running the full regression suite is generally not necessary. Run the following first:
This will run full backup/restore regression with a variety of options on all installed versions of PostgreSQL. If you are only interested in one version then modify the `db-version` setting to X.X (e.g. 9.4). `--thread-max` can be omitted if you are running single-threaded.
If there are errors in this test then run full regression to help isolate problems:
Defines the stanza for the command. A stanza is the configuration for a database that defines where it is located, how it will be backed up, archiving options, etc. Most db servers will only have one Postgres cluster and therefore one stanza, whereas backup servers will have a stanza for every database that needs to be backed up.
This option prevents PgBackRest from running `pg_start_backup()` and `pg_stop_backup()` on the database. In order for this to work PostgreSQL should be shut down and PgBackRest will generate an error if it is not.
The purpose of this option is to allow cold backups. The `pg_xlog` directory is copied as-is and `archive-check` is automatically disabled for the backup.
When used with `--no-start-stop` a backup will be run even if PgBackRest thinks that PostgreSQL is running. **This option should be used with extreme care as it will likely result in a bad backup.**
There are some scenarios where a backup might still be desirable under these conditions. For example, if a server crashes and the database volume can only be mounted read-only, it would be a good idea to take a backup even if `postmaster.pid` is present. In this case it would be better to revert to the prior backup and replay WAL, but possibly there is a very important transaction in a WAL segment that did not get archived.
Run a `full` backup on the `db` stanza. `--type` can also be set to `incr` or `diff` for incremental or differential backups. However, if no `full` backup exists then a `full` backup will be forced even if `incr` or `diff` is requested.
Accepts a WAL segment from PostgreSQL and archives it in the repository. `%p` is how PostgreSQL specifies the location of the WAL segment to be archived.
Retrieves a WAL segment from the repository. This command is used in `recovery.conf` to restore a backup, perform PITR, or as an alternative to streaming for keeping a replica up to date. `%f` is how PostgreSQL specifies the WAL segment it needs and `%p` is the location where it should be copied.
PgBackRest does backup rotation, but is not concerned with when the backups were created. So if two full backups are configured for retention, PgBackRest will keep two full backups no matter whether they occur, two hours apart or two weeks apart.
Expire (rotate) any backups that exceed the defined retention. Expiration is run automatically after every successful backup, so there is no need to run this command separately unless you have reduced retention, usually to free up some space.
By default the PostgreSQL data and tablespace directories are expected to be present but empty. This option performs a delta restore using checksums.
```
required: n
default: n
```
##### `force` option
By itself this option forces the PostgreSQL data and tablespace paths to be completely overwritten. In combination with `--delta` a timestamp/size delta will be performed instead of using checksums.
```
required: n
default: n
```
##### `type` option
The following recovery types are supported:
-`default` - recover to the end of the archive stream.
-`name` - recover the restore point specified in `--target`.
-`xid` - recover to the transaction id specified in `--target`.
-`time` - recover to the time specified in `--target`.
-`preserve` - preserve the existing `recovery.conf` file.
-`none` - no `recovery.conf` file is written so PostgreSQL will attempt to achieve consistency using WAL segments present in `pg_xlog`. Provide the required WAL segments or use the `archive-copy` setting to include them with the backup.
Defines the recovery target when `--type` is `name`, `xid`, or `time`.
```
required: y
example: "--target=2015-01-30 14:15:11 EST"
```
##### `target-exclusive` option
Defines whether recovery to the target would be exclusive (the default is inclusive) and is only valid when `--type` is `time` or `xid`. For example, using `--target-exclusive` would exclude the contents of transaction `1007` when `--type=xid` and `--target=1007`. See `recovery_target_inclusive` option in the PostgreSQL docs for more information.
```
required: n
default: n
```
##### `target-resume` option
Specifies whether recovery should resume when the recovery target is reached. See `pause_at_recovery_target` in the PostgreSQL docs for more information.
```
required: n
default: n
```
##### `target-timeline` option
Recovers along the specified timeline. See `recovery_target_timeline` in the PostgreSQL docs for more information.
Recovery settings in recovery.conf options can be specified with this option. See http://www.postgresql.org/docs/X.X/static/recovery-config.html for details on recovery.conf options (replace X.X with your database version). This option can be used multiple times.
Note: `restore_command` will be automatically generated but can be overridden with this option. Be careful about specifying your own `restore_command` as PgBackRest is designed to handle this for you. Target Recovery options (recovery_target_name, recovery_target_time, etc.) are generated automatically by PgBackRest and should not be set with this option.
Recovery settings can also be set in the `restore:recovery-setting` section of pg_backrest.conf. For example:
```
[restore:recovery-setting]
primary_conn_info=db.mydomain.com
standby_mode=on
```
Since PgBackRest does not start PostgreSQL after writing the `recovery.conf` file, it is always possible to edit/check `recovery.conf` before manually restarting.
Moves a tablespace to a new location during the restore. This is useful when tablespace locations are not the same on a replica, or an upgraded system has different mount points.
Since PostgreSQL 9.2 tablespace locations are not stored in pg_tablespace so moving tablespaces can be done with impunity. However, moving a tablespace to the `data_directory` is not recommended and may cause problems. For more information on moving tablespaces http://www.databasesoup.com/2013/11/moving-tablespaces.html is a good resource.
Retrieve information about backups for a single stanza or for all stanzas. Text output is the default and gives a human-readable summary of backups for the stanza(s) requested. This format is subject to change with any release.
For machine-readable output use `--output=json`. The JSON output contains far more information than the text output, however **this feature is currently experimental so the format may change between versions**.
##### `output` option
The following output types are supported:
-`text` - Human-readable summary of backup information.
-`json` - Exhaustive machine-readable backup information in JSON format.
```
required: n
default: text
example: --output=json
```
##### Example: Backup information
```
/path/to/pg_backrest.pl --stanza=db --output=json info
PgBackRest can be used entirely with command-line parameters but a configuration file is more practical for installations that are complex or set a lot of options. The default location for the configuration file is `/etc/pg_backrest.conf`.
### Examples
#### Confguring Postgres for Archiving
Modify the following settings in `postgresql.conf`:
Replace the path with the actual location where PgBackRest was installed. The stanza parameter should be changed to the actual stanza name for your database.
#### Minimal Configuration
The absolute minimum required to run PgBackRest (if all defaults are accepted) is the database path.
`/etc/pg_backrest.conf`:
```
[main]
db-path=/data/db
```
The `db-path` option could also be provided on the command line, but it's best to use a configuration file as options tend to pile up quickly.
#### Simple Single Host Configuration
This configuration is appropriate for a small installation where backups are being made locally or to a remote file system that is mounted locally. A number of additional options are set:
-`cmd-psql` - Custom location and parameters for psql.
-`cmd-psql-option` - Options for psql can be set per stanza.
-`compress` - Disable compression (handy if the file system is already compressed).
-`repo-path` - Path to the PgBackRest repository where backups and WAL archive are stored.
-`log-level-file` - Set the file log level to debug (Lots of extra info if something is not working as expected).
-`hardlink` - Create hardlinks between backups (but never between full backups).
-`thread-max` - Use 2 threads for backup/restore operations.
This configuration is appropriate for a small installation where backups are being made remotely. Make sure that postgres@db-host has trusted ssh to backrest@backup-host and vice versa. This configuration assumes that you have pg_backrest_remote.pl and pg_backrest.pl in the same path on both servers.
`/etc/pg_backrest.conf` on the db host:
```
[global:general]
repo-path=/path/to/db/repo
repo-remote-path=/path/to/backup/repo
[global:backup]
backup-host=backup.mydomain.com
backup-user=backrest
[global:archive]
archive-async=y
[main]
db-path=/data/db
```
`/etc/pg_backrest.conf` on the backup host:
```
[global:general]
repo-path=/path/to/backup/repo
[main]
db-host=db.mydomain.com
db-path=/data/db
db-user=postgres
```
### Options
#### `command` section
The `command` section defines the location of external commands that are used by PgBackRest.
##### `cmd-psql` key
Defines the full path to `psql`. `psql` is used to call `pg_start_backup()` and `pg_stop_backup()`.
If additional per stanza parameters need to be passed to `psql` (such as `--port` or `--cluster`) then add `%option%` to the command line and use `command-option::psql` to set options.
Allows per stanza command line parameters to be passed to `psql`.
```
required: n
example: cmd-psql-option --port=5433
```
##### `cmd-remote` key
Defines the location of `pg_backrest_remote.pl`.
Required only if the path to `pg_backrest_remote.pl` is different on the local and remote systems. If not defined, the remote path will be assumed to be the same as the local path.
The `general` section defines settings that are shared between multiple operations.
##### `buffer-size` key
Set the buffer size used for copy, compress, and uncompress functions. A maximum of 3 buffers will be in use at a time per thread. An additional maximum of 256K per thread may be used for zlib buffers.
Enable gzip compression. Backup files are compatible with command-line gzip tools.
```
required: n
default: y
example: compress=n
```
##### `compress-level` key
Sets the zlib level to be used for file compression when `compress=y`.
```
required: n
default: 6
allow: 0-9
example: compress-level=9
```
##### `compress-level-network` key
Sets the zlib level to be used for protocol compression when `compress=n` and the database is not on the same host as the backup. Protocol compression is used to reduce network traffic but can be disabled by setting `compress-level-network=0`. When `compress=y` the `compress-level-network` setting is ignored and `compress-level` is used instead so that the file is only compressed once. SSH compression is always disabled.
```
required: n
default: 3
allow: 0-9
example: compress-level-network=1
```
##### `repo-path` key
Path to the backrest repository where WAL segments, backups, logs, etc are stored.
```
required: n
default: /var/lib/backup
example: repo-path=/data/db/backrest
```
##### `repo-remote-path` key
Path to the remote backrest repository where WAL segments, backups, logs, etc are stored.
```
required: n
example: repo-remote-path=/backup/backrest
```
#### `backup` section
The `backup` section defines settings related to backup.
##### `backup-host` key
Sets the backup host when backup up remotely via SSH. Make sure that trusted SSH authentication is configured between the db host and the backup host.
When backing up to a locally mounted network filesystem this setting is not required.
Forces a checkpoint (by passing `true` to the `fast` parameter of `pg_start_backup()`) so the backup begins immediately. Otherwise the backup will start after the next regular checkpoint.
Enable hard-linking of files in differential and incremental backups to their full backups. This gives the appearance that each backup is a full backup. Be careful, though, because modifying files that are hard-linked can affect all the backups in the set.
Defines how often the manifest will be saved during a backup (in bytes). Saving the manifest is important because it stores the checksums and allows the resume function to work efficiently. The actual threshold used is 1% of the backup size or `manifest-save-threshold`, whichever is greater.
Defines whether the resume feature is enabled. Resume can greatly reduce the amount of time required to run a backup after a previous backup of the same type has failed. It adds complexity, however, so it may be desirable to disable in environments that do not require the feature.
Defines the number of threads to use for backup or restore. Each thread will perform compression and transfer to make the backup run faster, but don't set `thread-max` so high that it impacts database performance during backup.
```
required: n
default: 1
example: thread-max=4
```
##### `thread-timeout` key
Maximum amount of time (in seconds) that a backup thread should run. This limits the amount of time that a thread might be stuck due to unforeseen issues during the backup. Has no affect when `thread-max=1`.
```
required: n
example: thread-timeout=3600
```
##### `archive-check` key
Checks that all WAL segments required to make the backup consistent are present in the WAL archive. It's a good idea to leave this as the default unless you are using another method for archiving.
```
required: n
default: y
example: archive-check=n
```
##### `archive-copy` key
Store WAL segments required to make the backup consistent in the backup's pg_xlog path. This slightly paranoid option protects against corruption or premature expiration in the WAL segment archive. PITR won't be possible without the WAL segment archive and this option also consumes more space.
Even though WAL segments will be restored with the backup, PostgreSQL will ignore them if a `recovery.conf` file exists and instead use `archive_command` to fetch WAL segments. Specifying `type=none` when restoring will not create `recovery.conf` and force PostgreSQL to use the WAL segments in pg_xlog. This will get the database to a consistent state.
The `archive` section defines parameters when doing async archiving. This means that the archive files will be stored locally, then a background process will pick them and move them to the backup.
##### `archive-async` key
Archive WAL segments asynchronously. WAL segments will be copied to the local repo, then a process will be forked to compress the segment and transfer it to the remote repo if configured. Control will be returned to PostgreSQL as soon as the WAL segment is copied locally.
- An error will be logged to the console and also to the Postgres log.
- A stop file will be written in the lock directory and no more archive files will be backed up until it is removed.
If this occurs then the archive log stream will be interrupted and PITR will not be possible past that point. A new backup will be required to regain full restore capability.
The purpose of this feature is to prevent the log volume from filling up at which point Postgres will stop completely. Better to lose the backup than have the database go down.
To start normal archiving again you'll need to remove the stop file which will be located at `${repo-path}/lock/${stanza}-archive.stop` where `${repo-path}` is the path set in the `general` section, and `${stanza}` is the backup stanza.
The `restore` section defines settings used for restoring backups.
##### `tablespace` key
Defines whether tablespaces will be be restored into their original (or remapped) locations or stored directly under the `pg_tblspc` path. Disabling this setting produces compact restores that are convenient for development, staging, etc. Currently these restores cannot be backed up as PgBackRest expects only links in the `pg_tblspc` path. If no tablespaces are present this this setting has no effect.
The `expire` section defines how long backups will be retained. Expiration only occurs when the number of complete backups exceeds the allowed retention. In other words, if full-retention is set to 2, then there must be 3 complete backups before the oldest will be expired. Make sure you always have enough space for retention + 1 backups.
Number of full backups to keep. When a full backup expires, all differential and incremental backups associated with the full backup will also expire. When not defined then all full backups will be kept.
```
required: n
example: retention-full=2
```
##### `retention-diff` key
Number of differential backups to keep. When a differential backup expires, all incremental backups associated with the differential backup will also expire. When not defined all differential backups will be kept.
Type of backup to use for archive retention (full or differential). If set to full, then PgBackRest will keep archive logs for the number of full backups defined by `retention-archive`. If set to differential, then PgBackRest will keep archive logs for the number of differential backups defined by `retention-archive`.
If not defined then archive logs will be kept indefinitely. In general it is not useful to keep archive logs that are older than the oldest backup, but there may be reasons for doing so.
Number of backups worth of archive log to keep. If this is set less than your backup retention then be sure you set `archive-copy=y` or you won't be able to restore some older backups.
For example, if `retention-archive=2` and `retention-full=4`, then any backups older than the most recent two full backups will not have WAL segments in the archive to make them consistent. To solve this, set `archive-copy=y` and use `type=none` when restoring. This issue will be addressed in a future release but for now be careful with this setting.
A stanza defines a backup for a specific database. The stanza section must define the base database path and host/user if the database is remote. Also, any global configuration sections can be overridden to define stanza-specific settings.
##### `db-host` key
Define the database host. Used for backups where the database host is different from the backup host.
```
required: n
example: db-host=db.domain.com
```
##### `db-user` key
Defines user account on the db host when `db-host` is defined.
```
required: n
example: db-user=postgres
```
##### `db-path` key
Path to the db data directory (data_directory setting in postgresql.conf).
### v0.75: New repository format, info command and experimental 9.5 support
* IMPORTANT NOTE: This flag day release breaks compatibility with older versions of PgBackRest. The manifest format, on-disk structure, and the binary names have all changed. You must create a new repository to hold backups for this version of PgBackRest and keep your older repository for a time in case you need to do a restore. The `pg_backrest.conf` file has not changed but you'll need to change any references to `pg_backrest.pl` in cron (or elsewhere) to `pg_backrest` (without the `.pl` extension).
* More efficient file ordering for backup. Files are copied in descending size order so a single thread does not end up copying a large file at the end. This had already been implemented for restore.
* Fixed an issue where archive-copy would fail on an incr/diff backup when hardlink=n. In this case the pg_xlog path does not already exist and must be created. Reported by Michael Renner
* Fixed an issue in async archiving where archive-push was not properly returning 0 when archive-max-mb was reached and moved the async check after transfer to avoid having to remove the stop file twice. Also added unit tests for this case and improved error messages to make it clearer to the user what went wrong. Reported by Michael Renner.
* Fixed a locking issue that could allow multiple operations of the same type against a single stanza. This appeared to be benign in terms of data integrity but caused spurious errors while archiving and could lead to errors in backup/restore. Reported by Michael Renner.
* Better resume support. Resumed files are checked to be sure they have not been modified and the manifest is saved more often to preserve checksums as the backup progresses. More unit tests to verify each resume case.
* Resume is now optional. Use the `resume` setting or `--no-resume` from the command line to disable.
* Added `tablespace` setting to allow tablespaces to be restored into the `pg_tblspc` path. This produces compact restores that are convenient for development, staging, etc. Currently these restores cannot be backed up as PgBackRest expects only links in the `pg_tblspc` path.
* Fixed a buffering error that could occur on large, highly-compressible files when copying to an uncompressed remote destination. The error was detected in the decompression code and resulted in a failed backup rather than corruption so it should not affect successful backups made with previous versions.
* Pushing duplicate WAL now generates an error. This worked before only if checksums were disabled.
* Database System IDs are used to make sure that all WAL in an archive matches up. This should help prevent misconfigurations that send WAL from multiple clusters to the same archive.
* Regression tests working back to PostgreSQL 8.3.
* Improved threading model by starting threads early and terminating them late.
* Added option `--no-start-stop` to allow backups when Postgres is shut down. If `postmaster.pid` is present then `--force` is required to make the backup run (though if Postgres is running an inconsistent backup will likely be created). This option was added primarily for the purpose of unit testing, but there may be applications in the real world as well.
* Fixed broken checksums and now they work with normal and resumed backups. Finally realized that checksums and checksum deltas should be functionally separated and this simplified a number of things. Issue #28 has been created for checksum deltas.
* Complete rewrite of BackRest::File module to use a custom protocol for remote operations and Perl native GZIP and SHA operations. Compression is performed in threads rather than forked processes.
* Fairly comprehensive unit tests for all the basic operations. More work to be done here for sure, but then there is always more work to be done on unit tests.
* Working on improving error handling in the file object. This is not complete, but works well enough to find a few errors that have been causing us problems (notably, find is occasionally failing building the archive async manifest when system is under load).
* Found and squashed a nasty bug where `file_copy()` was defaulted to ignore errors. There was also an issue in file_exists that was causing the test to fail when the file actually did exist. Together they could have resulted in a corrupt backup with no errors, though it is very unlikely.
* The `archive-get` operation returns a 1 when the archive file is missing to differentiate from hard errors (ssh connection failure, file copy error, etc.) This lets Postgres know that that the archive stream has terminated normally. However, this does not take into account possible holes in the archive stream.
* If an archive directory which should be empty could not be deleted backrest was throwing an error. There's a good fix for that coming, but for the time being it has been changed to a warning so processing can continue. This was impacting backups as sometimes the final archive file would not get pushed if the first archive file had been in a different directory (plus some bad luck).
* Removed `master_stderr_discard` option on database SSH connections. There have been occasional lockups and they could be related to issues originally seen in the file code.
* No restore functionality, but the backup directories are consistent Postgres data directories. You'll need to either uncompress the files or turn off compression in the backup. Uncompressed backups on a ZFS (or similar) filesystem are a good option because backups can be restored locally via a snapshot to create logical backups or do spot data recovery.
* Archiving is single-threaded. This has not posed an issue on our multi-terabyte databases with heavy write volume. Recommend a large WAL volume or to use the async option with a large volume nearby.
* Backups are multi-threaded, but the Net::OpenSSH library does not appear to be 100% thread-safe so it will very occasionally lock up on a thread. There is an overall process timeout that resolves this issue by killing the process. Yes, very ugly.
* Checksums are lost on any resumed backup. Only the final backup will record checksum on multiple resumes. Checksums from previous backups are correctly recorded and a full backup will reset everything.
* The backup.manifest is being written as Storable because Config::IniFile does not seem to handle large files well. Would definitely like to save these as human-readable text.
Crunchy Data Solutions (http://www.crunchydata.com) has contributed time and resources to PgBackRest and continues to support development. Resonate (http://www.resonate.com/) also contributed to the development of PgBackRest and allowed me to install early (but well tested) versions as their primary PostgreSQL backup solution.
