Skip to content

Latest commit

 

History

History
584 lines (440 loc) · 22.1 KB

README.rst

File metadata and controls

584 lines (440 loc) · 22.1 KB

WAL-E Disaster Recovery

WAL-E is a program designed to perform continuous archiving of PostgreSQL WAL files and manage the use of pg_start_backup and pg_stop_backup.

To correspond on using WAL-E or to collaborate on its development, do not hesitate to send mail to the mailing list at [email protected]. Github issues are also currently being used to track known problems, so please feel free to submit those.

WAL-E has four critical operators:

  • backup-fetch
  • backup-push
  • wal-fetch
  • wal-push

Of these, the "push" operators send things to storage and "fetch" operators get things from storage. "wal" operators send/get write ahead log, and "backup" send/get a hot backup of the base database that WAL segments can be applied to.

All of these operators work in a context of three important environment-variable based settings:

  • AWS_ACCESS_KEY_ID or WABS_ACCOUNT_NAME
  • AWS_SECRET_ACCESS_KEY or WABS_ACCESS_KEY
  • WALE_S3_PREFIX or WALE_WABS_PREFIX

With the exception of AWS_SECRET_ACCESS_KEY and WABS_ACCESS_KEY, all of these can be specified as arguments as well. The AWS_* variables are the standard access-control keying system provided by Amazon, where the WABS_* are the standard access credentials defined by Windows Azure.

The WALE_S3_PREFIX and WALE_WABS_PREFIX (_PREFIX) variables can be thought of as a context whereby this program operates on a single database cluster at a time. Generally, for any one database the _PREFIX will be the same between all four operators. This context-driven approach attempts to help users avoid errors such as one database overwriting the WAL segments of another, as long as the _PREFIX is set uniquely for each database. Use whichever variable is appropriate for the store you are using.

Important

Ensure that all servers have different _PREFIXes set. Reuse of a value between two servers will likely cause unrecoverable backups.

  • lzop
  • psql (>= 8.4)
  • pv

This software is most frequently used with Python 2.6+. It will probably never support Python 2.5 or below because of the much more useful timeout semantics with sockets in Python 2.6+. Still, if you feel strongly about supporting Python 2.5, please send mail to the mailing list (see the Introduction).

This software also has Python dependencies; installing with setup.py will attempt to resolve them:

  • gevent>=0.13.1
  • boto>=2.6.0
  • azure==0.7.0
  • argparse, if not on Python 2.7

The storage backend is determined by the defined _PREFIX. Prefixes with the scheme s3 will be directed towards S3, those with the scheme wabs will be directed towards Windows Azure Blob Service, and those with the scheme swift will be directed towards an OpenStack Swift installation.

Example S3 Prefix:

s3://some-bucket/directory/or/whatever

Example WABS Prefix:

wabs://some-container/directory/or/whatever

Example OpenStack Swift Prefix:

swift://some-container/directory/or/whatever

Pushing a base backup to S3:

$ AWS_SECRET_ACCESS_KEY=... wal-e                     \
  -k AWS_ACCESS_KEY_ID                                \
  --s3-prefix=s3://some-bucket/directory/or/whatever  \
  backup-push /var/lib/my/database

Sending a WAL segment to WABS:

$ WABS_ACCESS_KEY=... wal-e                                   \
  -a WABS_ACCOUNT_NAME                                        \
  --wabs-prefix=wabs://some-bucket/directory/or/whatever      \
  wal-push /var/lib/my/database/pg_xlog/WAL_SEGMENT_LONG_HEX

It is generally recommended that one use some sort of environment variable management with WAL-E: working with it this way is less verbose, less prone to error, and less likely to expose secret information in logs.

At this time, AWS_SECRET_ACCESS_KEY and WABS_ACCESS_KEY are the only secret values, and recording it frequently in logs is not recommended. The tool has never and should never accept secret information in argv to avoid process table security problems. However, the user running PostgreSQL (typically 'postgres') must be able to run a program that can access this secret information, as part of its archive_command.

envdir, part of the daemontools package is one recommended approach to setting environment variables. One can prepare an envdir-compatible directory like so:

# Assumption: the group is trusted to read secret information
# S3 Setup
$ umask u=rwx,g=rx,o=
$ mkdir -p /etc/wal-e.d/env
$ echo "secret-key-content" > /etc/wal-e.d/env/AWS_SECRET_ACCESS_KEY
$ echo "access-key" > /etc/wal-e.d/env/AWS_ACCESS_KEY_ID
$ echo 's3://some-bucket/directory/or/whatever' > \
  /etc/wal-e.d/env/WALE_S3_PREFIX
$ chown -R root:postgres /etc/wal-e.d


# Assumption: the group is trusted to read secret information
# WABS Setup
$ umask u=rwx,g=rx,o=
$ mkdir -p /etc/wal-e.d/env
$ echo "secret-key-content" > /etc/wal-e.d/env/WABS_ACCESS_KEY
$ echo "access-key" > /etc/wal-e.d/env/WABS_ACCOUNT_NAME
$ echo 'wabs://some-container/directory/or/whatever' > \
  /etc/wal-e.d/env/WALE_WABS_PREFIX
$ chown -R root:postgres /etc/wal-e.d

After having done this preparation, it is possible to run WAL-E commands much more simply, with less risk of accidentally using incorrect values:

$ envdir /etc/wal-e.d/env wal-e backup-push ...
$ envdir /etc/wal-e.d/env wal-e wal-push ...

envdir is conveniently combined with the archive_command functionality used by PostgreSQL to enable continuous archiving. To enable continuous archiving, one needs to edit postgresql.conf and restart the server. The important settings to enable continuous archiving are related here:

wal_level = archive # hot_standby in 9.0 is also acceptable
archive_mode = on
archive_command = 'envdir /etc/wal-e.d/env wal-e wal-push %p'
archive_timeout = 60

Every segment archived will be noted in the PostgreSQL log.

Warning

PostgreSQL users can check the pg_settings table and see the archive_command employed. Do not put secret information into postgresql.conf for that reason, and use envdir instead.

A base backup (via backup-push) can be uploaded at any time, but this must be done at least once in order to perform a restoration. It must be done again if any WAL segment was not correctly uploaded: point in time recovery will not be able to continue if there are any gaps in the WAL segments.

Pulling a base backup from S3:

$ sudo -u postgres bash -c                          \
"envdir /etc/wal-e.d/pull-env wal-e                 \
--s3-prefix=s3://some-bucket/directory/or/whatever  \
backup-fetch /var/lib/my/database LATEST"

This command makes use of the "LATEST" pseudo-name for a backup, which queries S3 to find the latest complete backup. Otherwise, a real name can be used:

$ sudo -u postgres bash -c                          \
"envdir /etc/wal-e.d/pull-env wal-e                 \
--s3-prefix=s3://some-bucket/directory/or/whatever  \
backup-fetch                                        \
/var/lib/my/database base_LONGWALNUMBER_POSITION_NUMBER"

One can find the name of available backups via the experimental backup-list operator, or using one's remote data store browsing program of choice, by looking at the PREFIX/basebackups_NNN/... directory.

It is also likely one will need to provide a recovery.conf file, as documented in the PostgreSQL manual, to recover the base backup, as WAL files will need to be downloaded to make the hot-backup taken with backup-push. The WAL-E's wal-fetch subcommand is designed to be useful for this very purpose, as it may be used in a recovery.conf file like this:

restore_command = 'envdir /etc/wal-e.d/env wal-e wal-fetch "%f" "%p"'

Warning

If the archived database contains user defined tablespaces please review the backup-fetch section below before utilizing that command.

backup-push, backup-fetch, wal-push, wal-fetch represent the primary functionality of WAL-E and must reside on the database machine. Unlike wal-push and wal-fetch commands, which function as described above, the``backup-push`` and backup-fetch require a little additional explination.

By default backup-push will include all user defined tablespaces in the database backup. please see the backup-fetch section below for WAL-E's tablespace restoration behavior.

There are two possible scenarios in which backup-fetch is run:

If the archived database did not contain any user defined tablespaces at the time of backup it is safe to execute backup-fetch with no additional work by following previous examples.

If the archived database did contain user defined tablespaces at the time of backup you there are specific behaviors of WAL-E you must be aware of:

WAL-E expects that tablespace symlinks will be in place prior to a backup-fetch run. This means prepare your target path by insuring ${PG_CLUSTER_DIRECTORY}/pg_tblspc contains all required symlinks before restoration time. If any expected symlink does not exist backup-fetch will fail.

If you are unable to reproduce tablespace storage structures prior to running backup-fetch you can set the option flag --blind-restore. This will direct WAL-E to skip the symlink verification process and place all data directly in the ${PG_CLUSTER_DIRECTORY}/pg_tblspc path.

Optionally, you can provide a restoration specification file to WAL-E using the backup-fetch --restore-spec RESTORE_SPEC option. This spec must be valid JSON and contain all contained tablespaces as well as the target storage path they require, and the symlink postgres expects for the tablespace. Here is an example for a cluster with a single tablespace:

{
    "12345": {
        "loc": "/data/postgres/tablespaces/tblspc001/",
        "link": "pg_tblspc/12345"
    },
    "tablespaces": [
        "12345"
    ],
}

Given this information WAL-E will create the data storage directory and symlink it appropriately in ${PG_CLUSTER_DIRECTORY}/pg_tblspc.

Warning

"link" properties of tablespaces in the restore specification must contain the pg_tblspc prefix, it will not be added for you.

These are commands that are not used expressly for backup or WAL pushing and fetching, but are important to the monitoring or maintenance of WAL-E archived databases. Unlike the critical four operators for taking and restoring backups (backup-push, backup-fetch, wal-push, wal-fetch) that must reside on the database machine, these commands can be productively run from any computer with the appropriate _PREFIX set and the necessary credentials to manipulate or read data there.

backup-list is useful for listing base backups that are complete for a given WAL-E context. Its output is subject to change, but currently it's a CSV with a one-line prepended header. Some fields are only filled in when the --detail option is passed to backup-list [1].

Note

Some --detail only fields are not strictly to the right of fields that do not require --detail be passed. This is not a problem if one uses any CSV parsing library (as two tab-delimiters will be emitted) to signify the empty column, but if one is hoping to use string mangling to extract fields, exhibit care.

Firstly, the fields that are filled in regardless of if --detail is passed or not:

Header in CSV Meaning
name The name of the backup, which can be passed to the delete and backup-fetch commands.
last_modified The date and time the backup was completed and uploaded, rendered in an ISO-compatible format with timezone information.
wal_segment_backup_start The wal segment number. It is a 24-character hexadecimal number. This information identifies the timeline and relative ordering of various backups.
wal_segment_offset_backup_start The offset in the WAL segment that this backup starts at. This is mostly to avoid ambiguity in event of backups that may start in the same WAL segment.

Secondly, the fields that are filled in only when --detail is passed:

Header in CSV Meaning
expanded_size_bytes The decompressed size of the backup in bytes.
wal_segment_backup_stop The last WAL segment file required to bring this backup into a consistent state, and thus available for hot-standby.
wal_segment_offset_backup_stop The offset in the last WAL segment file required to bring this backup into a consistent state.
[1]backup-list --detail is slower (one web request per backup, rather than one web request per thousand backups or so) than backup-list, and often (but not always) the information in the regular backup-list is all one needs.

delete contains additional subcommands that are used for deleting data from storage for various reasons. These commands are organized separately because the delete subcommand itself takes options that apply to any subcommand that does deletion, such as --confirm.

All deletions are designed to be reentrant and idempotent: there are no negative consequences if one runs several deletions at once or if one resubmits the same deletion command several times, with or without canceling other deletions that may be concurrent.

These commands have a dry-run mode that is the default. The command is basically optimized for not deleting data except in a very specific circumstance to avoid operator error. Should a dry-run be performed, wal-e will instead simply report every key it would otherwise delete if it was not running in dry-run mode, along with prominent HINT-lines for every key noting that nothing was actually deleted from the blob store.

To actually delete any data, one must pass --confirm to wal-e delete. If one passes both --dry-run and --confirm, a dry run will be performed, regardless of the order of options passed.

Currently, these kinds of deletions are supported. Examples omit environment variable configuration for clarity:

  • before: Delete all backups and wal segment files before the given base-backup name. This does not include the base backup passed: it will remain a viable backup.

    Example:

    $ wal-e delete [--confirm] before base_00000004000002DF000000A6_03626144
    
  • retain: Leave the given number of backups in place, and delete all base backups and wal segment files older than them.

    Example:

    $ wal-e delete [--confirm] retain 5
    
  • old-versions: Delete all backups and wal file segments with an older format. This is only intended to be run after a major WAL-E version upgrade and the subsequent base-backup. If no base backup is successfully performed first, one is more exposed to data loss until one does perform a base backup.

    Example:

    $ wal-e delete [--confirm] old-versions
    
  • everything: Delete all backups and wal file segments in the context. This is appropriate if one is decommissioning a database and has no need for its archives.

    Example:

    $ wal-e delete [--confirm] everything
    

All assets pushed to storage are run through the program "lzop" which compresses the object using the very fast lzo compression algorithm. It takes roughly 2 CPU seconds to compress a gigabyte, which when sending things to storage at about 25MB/s occupies about 5% CPU time. Compression ratios are expected to make file sizes 50% or less of the original file size in most cases, making backups and restorations considerably faster.

Because storage services generally require the Content-Length header of a stored object to be set up-front, it is necessary to completely finish compressing an entire input file and storing the compressed output in a temporary file. Thus, the temporary file directory needs to be big enough and fast enough to support this, although this tool is designed to avoid calling fsync(), so some memory can be leveraged.

Base backups first have their files consolidated into disjoint tar files of limited length to avoid the relatively large per-file transfer overhead. This has the effect of making base backups and restores much faster when many small relations and ancillary files are involved.

To encrypt backups as well as compress them, first generate a key pair using gpg --gen-key. You don't need the private key on the machine to back up, but you will need it to restore. It'll need to have no passphrase.

Once this is done, just set the WALE_GPG_KEY_ID environment variable or the --gpg-key-id command line option to the ID of the secret key for backup and restore commands.

To reduce the read load on base backups, they are sent through the tool pv first. To use this rate-limited-read mode, use the option --cluster-read-rate-limit as seen in wal-e backup-push.

To restrict log statements to warnings and errors, use the --terse option. This is supported on all WAL-E operations.

In certain situations, the wal-push process can take long enough that it can't keep up with WAL segments being produced by Postgres, which can lead to unbounded disk usage and an eventual crash of the database.

One can instruct WAL-E to pool WAL segments together and send them in groups by passing the --pool-size parameter to wal-push. This can increase throughput significantly.

As of version 0.7.x, --pool-size defaults to 8.

Development is heavily reliant on the tool tox being existent within the development environment. All additional dependencies of WAL-E are managed by tox. In addition, the coding conventions are checked by the tox configuration included with WAL-E.

To run the tests, one need only run:

$ tox

However, if one does not have both Python 2.6 and 2.7 installed simultaneously (WAL-E supports both and tests both), there will be errors in running tox as seen previously. One can restrict the test to the Python of one's choice to avoid that:

$ tox -e py27

To run a somewhat more lengthy suite of integration tests that communicate with AWS S3, one might run tox like this:

$ WALE_S3_INTEGRATION_TESTS=TRUE      \
  AWS_ACCESS_KEY_ID=[AKIA...]         \
  AWS_SECRET_ACCESS_KEY=[...]         \
  WALE_WABS_INTEGRATION_TESTS=TRUE    \
  WABS_ACCOUNT_NAME=[...]             \
  WABS_ACCESS_KEY=[...]               \
  tox -- -n 8

Looking carefully at the above, notice the -n 8 added the tox invocation. This -n 8 is after a -- that indicates to tox that the subsequent arguments are for the underlying test program, not tox itself.

This is to enable parallel test execution, which makes the integration tests complete a small fraction of the time it would take otherwise. It is a design requirement of new tests that parallel execution not be sacrificed.

The above invocation tests WAL-E with every test environment defined in tox.ini. When iterating, testing all of those is typically not a desirable use of time, so one can restrict the integration test to one virtual environment, in a combination of features seen in all the previous examples:

$ WALE_S3_INTEGRATION_TESTS=TRUE      \
  AWS_ACCESS_KEY_ID=[AKIA...]         \
  AWS_SECRET_ACCESS_KEY=[...]         \
  WALE_WABS_INTEGRATION_TESTS=TRUE    \
  WABS_ACCOUNT_NAME=[...]             \
  WABS_ACCESS_KEY=[...]               \
  tox -e py27 -- -n 8

Coverage testing can be used by combining any of these using pytest-cov, e.g.: tox -- --cov wal_e and tox -- --cov wal_e --cov-report html; see htmlcov/index.html.

Finally, the test framework used is pytest. If possible, do not submit Python unittest style tests: those tend to be more verbose and anemic in power; however, any automated testing is better than a lack thereof, so if you are familiar with unittest, do not let the preference for pytest idiom be an impediment to submitting code.