2 $PostgreSQL: pgsql/doc/src/sgml/backup.sgml,v 2.75 2005/11/04 23:13:59 petere Exp $
5 <title>Backup and Restore</title>
7 <indexterm zone="backup"><primary>backup</></>
10 As with everything that contains valuable data, <productname>PostgreSQL</>
11 databases should be backed up regularly. While the procedure is
12 essentially simple, it is important to have a basic understanding of
13 the underlying techniques and assumptions.
17 There are three fundamentally different approaches to backing up
18 <productname>PostgreSQL</> data:
20 <listitem><para><acronym>SQL</> dump</para></listitem>
21 <listitem><para>File system level backup</para></listitem>
22 <listitem><para>On-line backup</para></listitem>
24 Each has its own strengths and weaknesses.
27 <sect1 id="backup-dump">
28 <title><acronym>SQL</> Dump</title>
31 The idea behind the SQL-dump method is to generate a text file with SQL
32 commands that, when fed back to the server, will recreate the
33 database in the same state as it was at the time of the dump.
34 <productname>PostgreSQL</> provides the utility program
35 <xref linkend="app-pgdump"> for this purpose. The basic usage of this
38 pg_dump <replaceable class="parameter">dbname</replaceable> > <replaceable class="parameter">outfile</replaceable>
40 As you see, <application>pg_dump</> writes its results to the
41 standard output. We will see below how this can be useful.
45 <application>pg_dump</> is a regular <productname>PostgreSQL</>
46 client application (albeit a particularly clever one). This means
47 that you can do this backup procedure from any remote host that has
48 access to the database. But remember that <application>pg_dump</>
49 does not operate with special permissions. In particular, it must
50 have read access to all tables that you want to back up, so in
51 practice you almost always have to run it as a database superuser.
55 To specify which database server <application>pg_dump</> should
56 contact, use the command line options <option>-h
57 <replaceable>host</></> and <option>-p <replaceable>port</></>. The
58 default host is the local host or whatever your
59 <envar>PGHOST</envar> environment variable specifies. Similarly,
60 the default port is indicated by the <envar>PGPORT</envar>
61 environment variable or, failing that, by the compiled-in default.
62 (Conveniently, the server will normally have the same compiled-in
67 As any other <productname>PostgreSQL</> client application,
68 <application>pg_dump</> will by default connect with the database
69 user name that is equal to the current operating system user name. To override
70 this, either specify the <option>-U</option> option or set the
71 environment variable <envar>PGUSER</envar>. Remember that
72 <application>pg_dump</> connections are subject to the normal
73 client authentication mechanisms (which are described in <xref
74 linkend="client-authentication">).
78 Dumps created by <application>pg_dump</> are internally consistent,
79 that is, updates to the database while <application>pg_dump</> is
80 running will not be in the dump. <application>pg_dump</> does not
81 block other operations on the database while it is working.
82 (Exceptions are those operations that need to operate with an
83 exclusive lock, such as <command>VACUUM FULL</command>.)
88 When your database schema relies on OIDs (for instance as foreign
89 keys) you must instruct <application>pg_dump</> to dump the OIDs
90 as well. To do this, use the <option>-o</option> command line
95 <sect2 id="backup-dump-restore">
96 <title>Restoring the dump</title>
99 The text files created by <application>pg_dump</> are intended to
100 be read in by the <application>psql</application> program. The
101 general command form to restore a dump is
103 psql <replaceable class="parameter">dbname</replaceable> < <replaceable class="parameter">infile</replaceable>
105 where <replaceable class="parameter">infile</replaceable> is what
106 you used as <replaceable class="parameter">outfile</replaceable>
107 for the <application>pg_dump</> command. The database <replaceable
108 class="parameter">dbname</replaceable> will not be created by this
109 command, you must create it yourself from <literal>template0</> before executing
110 <application>psql</> (e.g., with <literal>createdb -T template0
111 <replaceable class="parameter">dbname</></literal>).
112 <application>psql</> supports options similar to <application>pg_dump</>
113 for controlling the database server location and the user name. See
114 <xref linkend="app-psql">'s reference page for more information.
118 Not only must the target database already exist before starting to
119 run the restore, but so must all the users who own objects in the
120 dumped database or were granted permissions on the objects. If they
121 do not, then the restore will fail to recreate the objects with the
122 original ownership and/or permissions. (Sometimes this is what you want,
123 but usually it is not.)
127 Once restored, it is wise to run <xref linkend="sql-analyze"
128 endterm="sql-analyze-title"> on each database so the optimizer has
129 useful statistics. An easy way to do this is to run
130 <command>vacuumdb -a -z</> to
131 <command>VACUUM ANALYZE</> all databases; this is equivalent to
132 running <command>VACUUM ANALYZE</command> manually.
136 The ability of <application>pg_dump</> and <application>psql</> to
137 write to or read from pipes makes it possible to dump a database
138 directly from one server to another; for example:
140 pg_dump -h <replaceable>host1</> <replaceable>dbname</> | psql -h <replaceable>host2</> <replaceable>dbname</>
146 The dumps produced by <application>pg_dump</> are relative to
147 <literal>template0</>. This means that any languages, procedures,
148 etc. added to <literal>template1</> will also be dumped by
149 <application>pg_dump</>. As a result, when restoring, if you are
150 using a customized <literal>template1</>, you must create the
151 empty database from <literal>template0</>, as in the example
157 For advice on how to load large amounts of data into
158 <productname>PostgreSQL</productname> efficiently, refer to <xref
163 <sect2 id="backup-dump-all">
164 <title>Using <application>pg_dumpall</></title>
167 The above mechanism is cumbersome and inappropriate when backing
168 up an entire database cluster. For this reason the <xref
169 linkend="app-pg-dumpall"> program is provided.
170 <application>pg_dumpall</> backs up each database in a given
171 cluster, and also preserves cluster-wide data such as users and
172 groups. The basic usage of this command is:
174 pg_dumpall > <replaceable>outfile</>
176 The resulting dump can be restored with <application>psql</>:
178 psql -f <replaceable class="parameter">infile</replaceable> postgres
180 (Actually, you can specify any existing database name to start from,
181 but if you are reloading in an empty cluster then <literal>postgres</>
182 should generally be used.) It is always necessary to have
183 database superuser access when restoring a <application>pg_dumpall</>
184 dump, as that is required to restore the user and group information.
188 <sect2 id="backup-dump-large">
189 <title>Handling large databases</title>
192 Since <productname>PostgreSQL</productname> allows tables larger
193 than the maximum file size on your system, it can be problematic
194 to dump such a table to a file, since the resulting file will likely
195 be larger than the maximum size allowed by your system. Since
196 <application>pg_dump</> can write to the standard output, you can
197 just use standard Unix tools to work around this possible problem.
201 <title>Use compressed dumps.</title>
203 You can use your favorite compression program, for example
204 <application>gzip</application>.
207 pg_dump <replaceable class="parameter">dbname</replaceable> | gzip > <replaceable class="parameter">filename</replaceable>.gz
213 createdb <replaceable class="parameter">dbname</replaceable>
214 gunzip -c <replaceable class="parameter">filename</replaceable>.gz | psql <replaceable class="parameter">dbname</replaceable>
220 cat <replaceable class="parameter">filename</replaceable>.gz | gunzip | psql <replaceable class="parameter">dbname</replaceable>
226 <title>Use <command>split</>.</title>
228 The <command>split</command> command
229 allows you to split the output into pieces that are
230 acceptable in size to the underlying file system. For example, to
231 make chunks of 1 megabyte:
234 pg_dump <replaceable class="parameter">dbname</replaceable> | split -b 1m - <replaceable class="parameter">filename</replaceable>
240 createdb <replaceable class="parameter">dbname</replaceable>
241 cat <replaceable class="parameter">filename</replaceable>* | psql <replaceable class="parameter">dbname</replaceable>
247 <title>Use the custom dump format.</title>
249 If <productname>PostgreSQL</productname> was built on a system with the
250 <application>zlib</> compression library installed, the custom dump
251 format will compress data as it writes it to the output file. This will
252 produce dump file sizes similar to using <command>gzip</command>, but it
253 has the added advantage that tables can be restored selectively. The
254 following command dumps a database using the custom dump format:
257 pg_dump -Fc <replaceable class="parameter">dbname</replaceable> > <replaceable class="parameter">filename</replaceable>
260 A custom-format dump is not a script for <application>psql</>, but
261 instead must be restored with <application>pg_restore</>.
262 See the <xref linkend="app-pgdump"> and <xref
263 linkend="app-pgrestore"> reference pages for details.
270 <sect1 id="backup-file">
271 <title>File system level backup</title>
274 An alternative backup strategy is to directly copy the files that
275 <productname>PostgreSQL</> uses to store the data in the database. In
276 <xref linkend="creating-cluster"> it is explained where these files
277 are located, but you have probably found them already if you are
278 interested in this method. You can use whatever method you prefer
279 for doing usual file system backups, for example
282 tar -cf backup.tar /usr/local/pgsql/data
287 There are two restrictions, however, which make this method
288 impractical, or at least inferior to the <application>pg_dump</>
294 The database server <emphasis>must</> be shut down in order to
295 get a usable backup. Half-way measures such as disallowing all
296 connections will <emphasis>not</emphasis> work
297 (mainly because <command>tar</command> and similar tools do not take an
298 atomic snapshot of the state of the file system at a point in
299 time). Information about stopping the server can be found in
300 <xref linkend="postmaster-shutdown">. Needless to say that you
301 also need to shut down the server before restoring the data.
307 If you have dug into the details of the file system layout of the
308 database, you may be tempted to try to back up or restore only certain
309 individual tables or databases from their respective files or
310 directories. This will <emphasis>not</> work because the
311 information contained in these files contains only half the
312 truth. The other half is in the commit log files
313 <filename>pg_clog/*</filename>, which contain the commit status of
314 all transactions. A table file is only usable with this
315 information. Of course it is also impossible to restore only a
316 table and the associated <filename>pg_clog</filename> data
317 because that would render all other tables in the database
318 cluster useless. So file system backups only work for complete
319 restoration of an entire database cluster.
326 An alternative file-system backup approach is to make a
327 <quote>consistent snapshot</quote> of the data directory, if the
328 file system supports that functionality (and you are willing to
329 trust that it is implemented correctly). The typical procedure is
330 to make a <quote>frozen snapshot</> of the volume containing the
331 database, then copy the whole data directory (not just parts, see
332 above) from the snapshot to a backup device, then release the frozen
333 snapshot. This will work even while the database server is running.
334 However, a backup created in this way saves
335 the database files in a state where the database server was not
336 properly shut down; therefore, when you start the database server
337 on the backed-up data, it will think the server had crashed
338 and replay the WAL log. This is not a problem, just be aware of
339 it (and be sure to include the WAL files in your backup).
343 If your database is spread across multiple file systems, there may not
344 be any way to obtain exactly-simultaneous frozen snapshots of all
345 the volumes. For example, if your data files and WAL log are on different
346 disks, or if tablespaces are on different file systems, it might
347 not be possible to use snapshot backup because the snapshots must be
349 Read your file system documentation very carefully before trusting
350 to the consistent-snapshot technique in such situations. The safest
351 approach is to shut down the database server for long enough to
352 establish all the frozen snapshots.
356 Another option is to use <application>rsync</> to perform a file
357 system backup. This is done by first running <application>rsync</>
358 while the database server is running, then shutting down the database
359 server just long enough to do a second <application>rsync</>. The
360 second <application>rsync</> will be much quicker than the first,
361 because it has relatively little data to transfer, and the end result
362 will be consistent because the server was down. This method
363 allows a file system backup to be performed with minimal downtime.
367 Note that a file system backup will not necessarily be
368 smaller than an SQL dump. On the contrary, it will most likely be
369 larger. (<application>pg_dump</application> does not need to dump
370 the contents of indexes for example, just the commands to recreate
375 <sect1 id="backup-online">
376 <title>On-line backup and point-in-time recovery (PITR)</title>
378 <indexterm zone="backup">
379 <primary>on-line backup</primary>
382 <indexterm zone="backup">
383 <primary>point-in-time recovery</primary>
386 <indexterm zone="backup">
387 <primary>PITR</primary>
391 At all times, <productname>PostgreSQL</> maintains a
392 <firstterm>write ahead log</> (WAL) in the <filename>pg_xlog/</>
393 subdirectory of the cluster's data directory. The log describes
394 every change made to the database's data files. This log exists
395 primarily for crash-safety purposes: if the system crashes, the
396 database can be restored to consistency by <quote>replaying</> the
397 log entries made since the last checkpoint. However, the existence
398 of the log makes it possible to use a third strategy for backing up
399 databases: we can combine a file-system-level backup with backup of
400 the WAL files. If recovery is needed, we restore the backup and
401 then replay from the backed-up WAL files to bring the backup up to
402 current time. This approach is more complex to administer than
403 either of the previous approaches, but it has some significant
408 We do not need a perfectly consistent backup as the starting point.
409 Any internal inconsistency in the backup will be corrected by log
410 replay (this is not significantly different from what happens during
411 crash recovery). So we don't need file system snapshot capability,
412 just <application>tar</> or a similar archiving tool.
417 Since we can string together an indefinitely long sequence of WAL files
418 for replay, continuous backup can be achieved simply by continuing to archive
419 the WAL files. This is particularly valuable for large databases, where
420 it may not be convenient to take a full backup frequently.
425 There is nothing that says we have to replay the WAL entries all the
426 way to the end. We could stop the replay at any point and have a
427 consistent snapshot of the database as it was at that time. Thus,
428 this technique supports <firstterm>point-in-time recovery</>: it is
429 possible to restore the database to its state at any time since your base
435 If we continuously feed the series of WAL files to another
436 machine that has been loaded with the same base backup file, we
437 have a <quote>hot standby</> system: at any point we can bring up
438 the second machine and it will have a nearly-current copy of the
446 As with the plain file-system-backup technique, this method can only
447 support restoration of an entire database cluster, not a subset.
448 Also, it requires a lot of archival storage: the base backup may be bulky,
449 and a busy system will generate many megabytes of WAL traffic that
450 have to be archived. Still, it is the preferred backup technique in
451 many situations where high reliability is needed.
455 To recover successfully using an on-line backup, you need a continuous
456 sequence of archived WAL files that extends back at least as far as the
457 start time of your backup. So to get started, you should set up and test
458 your procedure for archiving WAL files <emphasis>before</> you take your
459 first base backup. Accordingly, we first discuss the mechanics of
463 <sect2 id="backup-archiving-wal">
464 <title>Setting up WAL archiving</title>
467 In an abstract sense, a running <productname>PostgreSQL</> system
468 produces an indefinitely long sequence of WAL records. The system
469 physically divides this sequence into WAL <firstterm>segment
470 files</>, which are normally 16MB apiece (although the size can be
471 altered when building <productname>PostgreSQL</>). The segment
472 files are given numeric names that reflect their position in the
473 abstract WAL sequence. When not using WAL archiving, the system
474 normally creates just a few segment files and then
475 <quote>recycles</> them by renaming no-longer-needed segment files
476 to higher segment numbers. It's assumed that a segment file whose
477 contents precede the checkpoint-before-last is no longer of
478 interest and can be recycled.
482 When archiving WAL data, we want to capture the contents of each segment
483 file once it is filled, and save that data somewhere before the segment
484 file is recycled for reuse. Depending on the application and the
485 available hardware, there could be many different ways of <quote>saving
486 the data somewhere</>: we could copy the segment files to an NFS-mounted
487 directory on another machine, write them onto a tape drive (ensuring that
488 you have a way of restoring the file with its original file name), or batch
489 them together and burn them onto CDs, or something else entirely. To
490 provide the database administrator with as much flexibility as possible,
491 <productname>PostgreSQL</> tries not to make any assumptions about how
492 the archiving will be done. Instead, <productname>PostgreSQL</> lets
493 the administrator specify a shell command to be executed to copy a
494 completed segment file to wherever it needs to go. The command could be
495 as simple as a <application>cp</>, or it could invoke a complex shell
496 script — it's all up to you.
500 The shell command to use is specified by the <xref
501 linkend="guc-archive-command"> configuration parameter, which in practice
502 will always be placed in the <filename>postgresql.conf</filename> file.
504 any <literal>%p</> is replaced by the absolute path of the file to
505 archive, while any <literal>%f</> is replaced by the file name only.
506 Write <literal>%%</> if you need to embed an actual <literal>%</>
507 character in the command. The simplest useful command is something
510 archive_command = 'cp -i %p /mnt/server/archivedir/%f </dev/null'
512 which will copy archivable WAL segments to the directory
513 <filename>/mnt/server/archivedir</>. (This is an example, not a
514 recommendation, and may not work on all platforms.)
518 The archive command will be executed under the ownership of the same
519 user that the <productname>PostgreSQL</> server is running as. Since
520 the series of WAL files being archived contains effectively everything
521 in your database, you will want to be sure that the archived data is
522 protected from prying eyes; for example, archive into a directory that
523 does not have group or world read access.
527 It is important that the archive command return zero exit status if and
528 only if it succeeded. Upon getting a zero result,
529 <productname>PostgreSQL</> will assume that the WAL segment file has been
530 successfully archived, and will remove or recycle it.
531 However, a nonzero status tells
532 <productname>PostgreSQL</> that the file was not archived; it will try
533 again periodically until it succeeds.
537 The archive command should generally be designed to refuse to overwrite
538 any pre-existing archive file. This is an important safety feature to
539 preserve the integrity of your archive in case of administrator error
540 (such as sending the output of two different servers to the same archive
542 It is advisable to test your proposed archive command to ensure that it
543 indeed does not overwrite an existing file, <emphasis>and that it returns
544 nonzero status in this case</>. We have found that <literal>cp -i</> does
545 this correctly on some platforms but not others. If the chosen command
546 does not itself handle this case correctly, you should add a command
547 to test for pre-existence of the archive file. For example, something
550 archive_command = 'test ! -f .../%f && cp %p .../%f'
552 works correctly on most Unix variants.
556 While designing your archiving setup, consider what will happen if
557 the archive command fails repeatedly because some aspect requires
558 operator intervention or the archive runs out of space. For example, this
559 could occur if you write to tape without an autochanger; when the tape
560 fills, nothing further can be archived until the tape is swapped.
561 You should ensure that any error condition or request to a human operator
562 is reported appropriately so that the situation can be
563 resolved relatively quickly. The <filename>pg_xlog/</> directory will
564 continue to fill with WAL segment files until the situation is resolved.
568 The speed of the archiving command is not important, so long as it can keep up
569 with the average rate at which your server generates WAL data. Normal
570 operation continues even if the archiving process falls a little behind.
571 If archiving falls significantly behind, this will increase the amount of
572 data that would be lost in the event of a disaster. It will also mean that
573 the <filename>pg_xlog/</> directory will contain large numbers of
574 not-yet-archived segment files, which could eventually exceed available
575 disk space. You are advised to monitor the archiving process to ensure that
576 it is working as you intend.
580 If you are concerned about being able to recover right up to the
581 current instant, you may want to take additional steps to ensure that
582 the current, partially-filled WAL segment is also copied someplace.
583 This is particularly important if your server generates only little WAL
584 traffic (or has slack periods where it does so), since it could take a
585 long time before a WAL segment file is completely filled and ready to
586 archive. One possible way to handle this is to set up a
587 <application>cron</> job that periodically (once a minute, perhaps)
588 identifies the current WAL segment file and saves it someplace safe.
589 Then the combination of the archived WAL segments and the saved current
590 segment will be enough to ensure you can always restore to within a
591 minute of current time. This behavior is not presently built into
592 <productname>PostgreSQL</> because we did not want to complicate the
593 definition of the <xref linkend="guc-archive-command"> by requiring it
594 to keep track of successively archived, but different, copies of the
595 same WAL file. The <xref linkend="guc-archive-command"> is only
596 invoked on completed WAL segments. Except in the case of retrying a
597 failure, it will be called only once for any given file name.
601 In writing your archive command, you should assume that the file names to
602 be archived may be up to 64 characters long and may contain any
603 combination of ASCII letters, digits, and dots. It is not necessary to
604 remember the original full path (<literal>%p</>) but it is necessary to
605 remember the file name (<literal>%f</>).
609 Note that although WAL archiving will allow you to restore any
610 modifications made to the data in your <productname>PostgreSQL</> database
611 it will not restore changes made to configuration files (that is,
612 <filename>postgresql.conf</>, <filename>pg_hba.conf</> and
613 <filename>pg_ident.conf</>), since those are edited manually rather
614 than through SQL operations.
615 You may wish to keep the configuration files in a location that will
616 be backed up by your regular file system backup procedures. See
617 <xref linkend="runtime-config-file-locations"> for how to relocate the
622 <sect2 id="backup-base-backup">
623 <title>Making a Base Backup</title>
626 The procedure for making a base backup is relatively simple:
630 Ensure that WAL archiving is enabled and working.
635 Connect to the database as a superuser, and issue the command
637 SELECT pg_start_backup('label');
639 where <literal>label</> is any string you want to use to uniquely
640 identify this backup operation. (One good practice is to use the
641 full path where you intend to put the backup dump file.)
642 <function>pg_start_backup</> creates a <firstterm>backup label</> file,
643 called <filename>backup_label</>, in the cluster directory with
644 information about your backup.
648 It does not matter which database within the cluster you connect to to
649 issue this command. You can ignore the result returned by the function;
650 but if it reports an error, deal with that before proceeding.
655 Perform the backup, using any convenient file-system-backup tool
656 such as <application>tar</> or <application>cpio</>. It is neither
657 necessary nor desirable to stop normal operation of the database
663 Again connect to the database as a superuser, and issue the command
665 SELECT pg_stop_backup();
667 This should return successfully.
672 Once the WAL segment files used during the backup are archived as part
673 of normal database activity, you are done.
680 Some backup tools that you might wish to use emit warnings or errors
681 if the files they are trying to copy change while the copy proceeds.
682 This situation is normal, and not an error, when taking a base backup of
683 an active database; so you need to ensure that you can distinguish
684 complaints of this sort from real errors. For example, some versions
685 of <application>rsync</> return a separate exit code for <quote>vanished
686 source files</>, and you can write a driver script to accept this exit
687 code as a non-error case. Also,
688 some versions of GNU <application>tar</> consider it an error if a file
689 is changed while <application>tar</> is copying it. There does not seem
690 to be any very convenient way to distinguish this error from other types
691 of errors, other than manual inspection of <application>tar</>'s messages.
692 GNU <application>tar</> is therefore not the best tool for making base
697 It is not necessary to be very concerned about the amount of time elapsed
698 between <function>pg_start_backup</> and the start of the actual backup,
699 nor between the end of the backup and <function>pg_stop_backup</>; a
700 few minutes' delay won't hurt anything. You
701 must however be quite sure that these operations are carried out in
702 sequence and do not overlap.
706 Be certain that your backup dump includes all of the files underneath
707 the database cluster directory (e.g., <filename>/usr/local/pgsql/data</>).
708 If you are using tablespaces that do not reside underneath this directory,
709 be careful to include them as well (and be sure that your backup dump
710 archives symbolic links as links, otherwise the restore will mess up
715 You may, however, omit from the backup dump the files within the
716 <filename>pg_xlog/</> subdirectory of the cluster directory. This
717 slight complication is worthwhile because it reduces the risk
718 of mistakes when restoring. This is easy to arrange if
719 <filename>pg_xlog/</> is a symbolic link pointing to someplace outside
720 the cluster directory, which is a common setup anyway for performance
725 To make use of this backup, you will need to keep around all the WAL
726 segment files generated during and after the file system backup.
727 To aid you in doing this, the <function>pg_stop_backup</> function
728 creates a <firstterm>backup history file</> that is immediately
729 stored into the WAL archive area. This file is named after the first
730 WAL segment file that you need to have to make use of the backup.
731 For example, if the starting WAL file is
732 <literal>0000000100001234000055CD</> the backup history file will be
734 <literal>0000000100001234000055CD.007C9330.backup</>. (The second
735 number in the file name stands for an exact position within the WAL
736 file, and can ordinarily be ignored.) Once you have safely archived
737 the file system backup and the WAL segment files used during the
738 backup (as specified in the backup history file), all archived WAL
739 segments with names numerically less are no longer needed to recover
740 the file system backup and may be deleted. However, you should
741 consider keeping several backup sets to be absolutely certain that
742 you can recover your data. Keep in mind that only completed WAL
743 segment files are archived, so there will be delay between running
744 <function>pg_stop_backup</> and the archiving of all WAL segment
745 files needed to make the file system backup consistent.
748 The backup history file is just a small text file. It contains the
749 label string you gave to <function>pg_start_backup</>, as well as
750 the starting and ending times of the backup. If you used the label
751 to identify where the associated dump file is kept, then the
752 archived history file is enough to tell you which dump file to
753 restore, should you need to do so.
757 Since you have to keep around all the archived WAL files back to your
758 last base backup, the interval between base backups should usually be
759 chosen based on how much storage you want to expend on archived WAL
760 files. You should also consider how long you are prepared to spend
761 recovering, if recovery should be necessary — the system will have to
762 replay all those WAL segments, and that could take awhile if it has
763 been a long time since the last base backup.
767 It's also worth noting that the <function>pg_start_backup</> function
768 makes a file named <filename>backup_label</> in the database cluster
769 directory, which is then removed again by <function>pg_stop_backup</>.
770 This file will of course be archived as a part of your backup dump file.
771 The backup label file includes the label string you gave to
772 <function>pg_start_backup</>, as well as the time at which
773 <function>pg_start_backup</> was run, and the name of the starting WAL
774 file. In case of confusion it will
775 therefore be possible to look inside a backup dump file and determine
776 exactly which backup session the dump file came from.
780 It is also possible to make a backup dump while the postmaster is
781 stopped. In this case, you obviously cannot use
782 <function>pg_start_backup</> or <function>pg_stop_backup</>, and
783 you will therefore be left to your own devices to keep track of which
784 backup dump is which and how far back the associated WAL files go.
785 It is generally better to follow the on-line backup procedure above.
789 <sect2 id="backup-pitr-recovery">
790 <title>Recovering with an On-line Backup</title>
793 Okay, the worst has happened and you need to recover from your backup.
794 Here is the procedure:
798 Stop the postmaster, if it's running.
803 If you have the space to do so,
804 copy the whole cluster data directory and any tablespaces to a temporary
805 location in case you need them later. Note that this precaution will
806 require that you have enough free space on your system to hold two
807 copies of your existing database. If you do not have enough space,
808 you need at the least to copy the contents of the <filename>pg_xlog</>
809 subdirectory of the cluster data directory, as it may contain logs which
810 were not archived before the system went down.
815 Clean out all existing files and subdirectories under the cluster data
816 directory and under the root directories of any tablespaces you are using.
821 Restore the database files from your backup dump. Be careful that they
822 are restored with the right ownership (the database system user, not
823 root!) and with the right permissions. If you are using tablespaces,
824 you may want to verify that the symbolic links in <filename>pg_tblspc/</>
825 were correctly restored.
830 Remove any files present in <filename>pg_xlog/</>; these came from the
831 backup dump and are therefore probably obsolete rather than current.
832 If you didn't archive <filename>pg_xlog/</> at all, then re-create it,
833 and be sure to re-create the subdirectory
834 <filename>pg_xlog/archive_status/</> as well.
839 If you had unarchived WAL segment files that you saved in step 2,
840 copy them into <filename>pg_xlog/</>. (It is best to copy them,
841 not move them, so that you still have the unmodified files if a
842 problem occurs and you have to start over.)
847 Create a recovery command file <filename>recovery.conf</> in the cluster
848 data directory (see <xref linkend="recovery-config-settings">). You may
849 also want to temporarily modify <filename>pg_hba.conf</> to prevent
850 ordinary users from connecting until you are sure the recovery has worked.
855 Start the postmaster. The postmaster will go into recovery mode and
856 proceed to read through the archived WAL files it needs. Upon completion
857 of the recovery process, the postmaster will rename
858 <filename>recovery.conf</> to <filename>recovery.done</> (to prevent
859 accidentally re-entering recovery mode in case of a crash later) and then
860 commence normal database operations.
865 Inspect the contents of the database to ensure you have recovered to
866 where you want to be. If not, return to step 1. If all is well,
867 let in your users by restoring <filename>pg_hba.conf</> to normal.
874 The key part of all this is to set up a recovery command file that
875 describes how you want to recover and how far the recovery should
876 run. You can use <filename>recovery.conf.sample</> (normally
877 installed in the installation <filename>share/</> directory) as a
878 prototype. The one thing that you absolutely must specify in
879 <filename>recovery.conf</> is the <varname>restore_command</>,
880 which tells <productname>PostgreSQL</> how to get back archived
881 WAL file segments. Like the <varname>archive_command</>, this is
882 a shell command string. It may contain <literal>%f</>, which is
883 replaced by the name of the desired log file, and <literal>%p</>,
884 which is replaced by the absolute path to copy the log file to.
885 Write <literal>%%</> if you need to embed an actual <literal>%</>
886 character in the command. The simplest useful command is
889 restore_command = 'cp /mnt/server/archivedir/%f %p'
891 which will copy previously archived WAL segments from the directory
892 <filename>/mnt/server/archivedir</>. You could of course use something
893 much more complicated, perhaps even a shell script that requests the
894 operator to mount an appropriate tape.
898 It is important that the command return nonzero exit status on failure.
899 The command <emphasis>will</> be asked for log files that are not present
900 in the archive; it must return nonzero when so asked. This is not an
901 error condition. Be aware also that the base name of the <literal>%p</>
902 path will be different from <literal>%f</>; do not expect them to be
907 WAL segments that cannot be found in the archive will be sought in
908 <filename>pg_xlog/</>; this allows use of recent un-archived segments.
909 However segments that are available from the archive will be used in
910 preference to files in <filename>pg_xlog/</>. The system will not
911 overwrite the existing contents of <filename>pg_xlog/</> when retrieving
916 Normally, recovery will proceed through all available WAL segments,
917 thereby restoring the database to the current point in time (or as
918 close as we can get given the available WAL segments). But if you want
919 to recover to some previous point in time (say, right before the junior
920 DBA dropped your main transaction table), just specify the required
921 stopping point in <filename>recovery.conf</>. You can specify the stop
922 point, known as the <quote>recovery target</>, either by date/time or
923 by completion of a specific transaction ID. As of this writing only
924 the date/time option is very usable, since there are no tools to help
925 you identify with any accuracy which transaction ID to use.
930 The stop point must be after the ending time of the base backup (the
931 time of <function>pg_stop_backup</>). You cannot use a base backup
932 to recover to a time when that backup was still going on. (To
933 recover to such a time, you must go back to your previous base backup
934 and roll forward from there.)
938 <sect3 id="recovery-config-settings" xreflabel="Recovery Settings">
939 <title>Recovery Settings</title>
942 These settings can only be made in the <filename>recovery.conf</>
943 file, and apply only for the duration of the recovery. They must be
944 reset for any subsequent recovery you wish to perform. They cannot be
945 changed once recovery has begun.
950 <varlistentry id="restore-command" xreflabel="restore_command">
951 <term><varname>restore_command</varname> (<type>string</type>)</term>
954 The shell command to execute to retrieve an archived segment of
955 the WAL file series. This parameter is required.
956 Any <literal>%f</> in the string is
957 replaced by the name of the file to retrieve from the archive,
958 and any <literal>%p</> is replaced by the absolute path to copy
960 Write <literal>%%</> to embed an actual <literal>%</> character
964 It is important for the command to return a zero exit status if and
965 only if it succeeds. The command <emphasis>will</> be asked for file
966 names that are not present in the archive; it must return nonzero
967 when so asked. Examples:
969 restore_command = 'cp /mnt/server/archivedir/%f "%p"'
970 restore_command = 'copy /mnt/server/archivedir/%f "%p"' # Windows
976 <varlistentry id="recovery-target-time" xreflabel="recovery_target_time">
977 <term><varname>recovery_target_time</varname>
978 (<type>timestamp</type>)
982 This parameter specifies the time stamp up to which recovery
984 At most one of <varname>recovery_target_time</> and
985 <xref linkend="recovery-target-xid"> can be specified.
986 The default is to recover to the end of the WAL log.
987 The precise stopping point is also influenced by
988 <xref linkend="recovery-target-inclusive">.
993 <varlistentry id="recovery-target-xid" xreflabel="recovery_target_xid">
994 <term><varname>recovery_target_xid</varname> (<type>string</type>)</term>
997 This parameter specifies the transaction ID up to which recovery
998 will proceed. Keep in mind
999 that while transaction IDs are assigned sequentially at transaction
1000 start, transactions can complete in a different numeric order.
1001 The transactions that will be recovered are those that committed
1002 before (and optionally including) the specified one.
1003 At most one of <varname>recovery_target_xid</> and
1004 <xref linkend="recovery-target-time"> can be specified.
1005 The default is to recover to the end of the WAL log.
1006 The precise stopping point is also influenced by
1007 <xref linkend="recovery-target-inclusive">.
1012 <varlistentry id="recovery-target-inclusive"
1013 xreflabel="recovery_target_inclusive">
1014 <term><varname>recovery_target_inclusive</varname>
1015 (<type>boolean</type>)
1019 Specifies whether we stop just after the specified recovery target
1020 (<literal>true</literal>), or just before the recovery target
1021 (<literal>false</literal>).
1022 Applies to both <xref linkend="recovery-target-time">
1023 and <xref linkend="recovery-target-xid">, whichever one is
1024 specified for this recovery. This indicates whether transactions
1025 having exactly the target commit time or ID, respectively, will
1026 be included in the recovery. Default is <literal>true</>.
1031 <varlistentry id="recovery-target-timeline"
1032 xreflabel="recovery_target_timeline">
1033 <term><varname>recovery_target_timeline</varname>
1034 (<type>string</type>)
1038 Specifies recovering into a particular timeline. The default is
1039 to recover along the same timeline that was current when the
1040 base backup was taken. You would only need to set this parameter
1041 in complex re-recovery situations, where you need to return to
1042 a state that itself was reached after a point-in-time recovery.
1043 See <xref linkend="backup-timelines"> for discussion.
1054 <sect2 id="backup-timelines">
1055 <title>Timelines</title>
1057 <indexterm zone="backup">
1058 <primary>timelines</primary>
1062 The ability to restore the database to a previous point in time creates
1063 some complexities that are akin to science-fiction stories about time
1064 travel and parallel universes. In the original history of the database,
1065 perhaps you dropped a critical table at 5:15PM on Tuesday evening.
1066 Unfazed, you get out your backup, restore to the point-in-time 5:14PM
1067 Tuesday evening, and are up and running. In <emphasis>this</> history of
1068 the database universe, you never dropped the table at all. But suppose
1069 you later realize this wasn't such a great idea after all, and would like
1070 to return to some later point in the original history. You won't be able
1071 to if, while your database was up-and-running, it overwrote some of the
1072 sequence of WAL segment files that led up to the time you now wish you
1073 could get back to. So you really want to distinguish the series of
1074 WAL records generated after you've done a point-in-time recovery from
1075 those that were generated in the original database history.
1079 To deal with these problems, <productname>PostgreSQL</> has a notion
1080 of <firstterm>timelines</>. Each time you recover to a point-in-time
1081 earlier than the end of the WAL sequence, a new timeline is created
1082 to identify the series of WAL records generated after that recovery.
1083 (If recovery proceeds all the way to the end of WAL, however, we do not
1084 start a new timeline: we just extend the existing one.) The timeline
1085 ID number is part of WAL segment file names, and so a new timeline does
1086 not overwrite the WAL data generated by previous timelines. It is
1087 in fact possible to archive many different timelines. While that might
1088 seem like a useless feature, it's often a lifesaver. Consider the
1089 situation where you aren't quite sure what point-in-time to recover to,
1090 and so have to do several point-in-time recoveries by trial and error
1091 until you find the best place to branch off from the old history. Without
1092 timelines this process would soon generate an unmanageable mess. With
1093 timelines, you can recover to <emphasis>any</> prior state, including
1094 states in timeline branches that you later abandoned.
1098 Each time a new timeline is created, <productname>PostgreSQL</> creates
1099 a <quote>timeline history</> file that shows which timeline it branched
1100 off from and when. These history files are necessary to allow the system
1101 to pick the right WAL segment files when recovering from an archive that
1102 contains multiple timelines. Therefore, they are archived into the WAL
1103 archive area just like WAL segment files. The history files are just
1104 small text files, so it's cheap and appropriate to keep them around
1105 indefinitely (unlike the segment files which are large). You can, if
1106 you like, add comments to a history file to make your own notes about
1107 how and why this particular timeline came to be. Such comments will be
1108 especially valuable when you have a thicket of different timelines as
1109 a result of experimentation.
1113 The default behavior of recovery is to recover along the same timeline
1114 that was current when the base backup was taken. If you want to recover
1115 into some child timeline (that is, you want to return to some state that
1116 was itself generated after a recovery attempt), you need to specify the
1117 target timeline ID in <filename>recovery.conf</>. You cannot recover into
1118 timelines that branched off earlier than the base backup.
1122 <sect2 id="backup-online-caveats">
1123 <title>Caveats</title>
1126 At this writing, there are several limitations of the on-line backup
1127 technique. These will probably be fixed in future releases:
1132 Operations on hash and R-tree indexes are
1133 not presently WAL-logged, so replay will not update these index types.
1134 The recommended workaround is to manually <command>REINDEX</> each
1135 such index after completing a recovery operation.
1141 If a <command>CREATE DATABASE</> command is executed while a base
1142 backup is being taken, and then the template database that the
1143 <command>CREATE DATABASE</> copied is modified while the base backup
1144 is still in progress, it is possible that recovery will cause those
1145 modifications to be propagated into the created database as well.
1146 This is of course undesirable. To avoid this risk, it is best not to
1147 modify any template databases while taking a base backup.
1153 <command>CREATE TABLESPACE</> commands are WAL-logged with the literal
1154 absolute path, and will therefore be replayed as tablespace creations
1155 with the same absolute path. This might be undesirable if the log is
1156 being replayed on a different machine. It can be dangerous even if
1157 the log is being replayed on the same machine, but into a new data
1158 directory: the replay will still overwrite the contents of the original
1159 tablespace. To avoid potential gotchas of this sort, the best practice
1160 is to take a new base backup after creating or dropping tablespaces.
1167 It should also be noted that the default <acronym>WAL</acronym>
1168 format is fairly bulky since it includes many disk page snapshots.
1169 These page snapshots are designed to support crash recovery,
1170 since we may need to fix partially-written disk pages. Depending
1171 on your system hardware and software, the risk of partial writes may
1172 be small enough to ignore, in which case you can significantly reduce
1173 the total volume of archived logs by turning off page snapshots
1174 using the <xref linkend="guc-full-page-writes"> parameter.
1175 (Read the notes and warnings in
1176 <xref linkend="wal"> before you do so.)
1177 Turning off page snapshots does not prevent use of the logs for PITR
1179 An area for future development is to compress archived WAL data by
1180 removing unnecessary page copies even when <varname>full_page_writes</>
1181 is on. In the meantime, administrators
1182 may wish to reduce the number of page snapshots included in WAL by
1183 increasing the checkpoint interval parameters as much as feasible.
1188 <sect1 id="migration">
1189 <title>Migration Between Releases</title>
1191 <indexterm zone="migration">
1192 <primary>upgrading</primary>
1195 <indexterm zone="migration">
1196 <primary>version</primary>
1197 <secondary>compatibility</secondary>
1201 This section discusses how to migrate your database data from one
1202 <productname>PostgreSQL</> release to a newer one.
1203 The software installation procedure <foreignphrase>per se</> is not the
1204 subject of this section; those details are in <xref linkend="installation">.
1208 As a general rule, the internal data storage format is subject to
1209 change between major releases of <productname>PostgreSQL</> (where
1210 the number after the first dot changes). This does not apply to
1211 different minor releases under the same major release (where the
1212 number after the second dot changes); these always have compatible
1213 storage formats. For example, releases 7.0.1, 7.1.2, and 7.2 are
1214 not compatible, whereas 7.1.1 and 7.1.2 are. When you update
1215 between compatible versions, you can simply replace the executables
1216 and reuse the data directory on disk. Otherwise you need to back
1217 up your data and restore it on the new server. This has to be done
1218 using <application>pg_dump</>; file system level backup methods
1219 obviously won't work. There are checks in place that prevent you
1220 from using a data directory with an incompatible version of
1221 <productname>PostgreSQL</productname>, so no great harm can be done by
1222 trying to start the wrong server version on a data directory.
1226 It is recommended that you use the <application>pg_dump</> and
1227 <application>pg_dumpall</> programs from the newer version of
1228 <productname>PostgreSQL</>, to take advantage of any enhancements
1229 that may have been made in these programs. Current releases of the
1230 dump programs can read data from any server version back to 7.0.
1234 The least downtime can be achieved by installing the new server in
1235 a different directory and running both the old and the new servers
1236 in parallel, on different ports. Then you can use something like
1239 pg_dumpall -p 5432 | psql -d postgres -p 6543
1242 to transfer your data. Or use an intermediate file if you want.
1243 Then you can shut down the old server and start the new server at
1244 the port the old one was running at. You should make sure that the
1245 old database is not updated after you run <application>pg_dumpall</>,
1246 otherwise you will obviously lose that data. See <xref
1247 linkend="client-authentication"> for information on how to prohibit
1252 In practice you probably want to test your client
1253 applications on the new setup before switching over completely.
1254 This is another reason for setting up concurrent installations
1255 of old and new versions.
1259 If you cannot or do not want to run two servers in parallel you can
1260 do the backup step before installing the new version, bring down
1261 the server, move the old version out of the way, install the new
1262 version, start the new server, restore the data. For example:
1265 pg_dumpall > backup
1267 mv /usr/local/pgsql /usr/local/pgsql.old
1268 cd ~/postgresql-&version;
1270 initdb -D /usr/local/pgsql/data
1271 postmaster -D /usr/local/pgsql/data
1272 psql -f backup postgres
1275 See <xref linkend="runtime"> about ways to start and stop the
1276 server and other details. The installation instructions will advise
1277 you of strategic places to perform these steps.
1282 When you <quote>move the old installation out of the way</quote>
1283 it may no longer be perfectly usable. Some of the executable programs
1284 contain absolute paths to various installed programs and data files.
1285 This is usually not a big problem but if you plan on using two
1286 installations in parallel for a while you should assign them
1287 different installation directories at build time. (This problem
1288 is rectified in <productname>PostgreSQL</> 8.0 and later, but you
1289 need to be wary of moving older installations.)
1295 <!-- Keep this comment at the end of the file
1300 sgml-minimize-attributes:nil
1301 sgml-always-quote-attributes:t
1303 sgml-indent-tabs-mode:nil
1305 sgml-parent-document:nil
1306 sgml-default-dtd-file:"./reference.ced"
1307 sgml-exposed-tags:nil
1308 sgml-local-catalogs:("/usr/share/sgml/catalog")
1309 sgml-local-ecat-files:nil