1PG_UPGRADE(1) PostgreSQL 15.4 Documentation PG_UPGRADE(1)
2
6 pg_upgrade - upgrade a PostgreSQL server instance
7
9 pg_upgrade -b oldbindir [-B newbindir] -d oldconfigdir -D newconfigdir
10 [option...]
11
13 pg_upgrade (formerly called pg_migrator) allows data stored in
14 PostgreSQL data files to be upgraded to a later PostgreSQL major
15 version without the data dump/restore typically required for major
16 version upgrades, e.g., from 9.5.8 to 9.6.4 or from 10.7 to 11.2. It is
17 not required for minor version upgrades, e.g., from 9.6.2 to 9.6.3 or
18 from 10.1 to 10.2.
19 Major PostgreSQL releases regularly add new features that often change
21 the layout of the system tables, but the internal data storage format
22 rarely changes. pg_upgrade uses this fact to perform rapid upgrades by
23 creating new system tables and simply reusing the old user data files.
24 If a future major release ever changes the data storage format in a way
25 that makes the old data format unreadable, pg_upgrade will not be
26 usable for such upgrades. (The community will attempt to avoid such
27 situations.)
28 pg_upgrade does its best to make sure the old and new clusters are
30 binary-compatible, e.g., by checking for compatible compile-time
31 settings, including 32/64-bit binaries. It is important that any
32 external modules are also binary compatible, though this cannot be
33 checked by pg_upgrade.
34 pg_upgrade supports upgrades from 9.2.X and later to the current major
36 release of PostgreSQL, including snapshot and beta releases.
37
39 pg_upgrade accepts the following command-line arguments:
40 -b bindir
42 --old-bindir=bindir
43 the old PostgreSQL executable directory; environment variable
44 PGBINOLD
45 -B bindir
47 --new-bindir=bindir
48 the new PostgreSQL executable directory; default is the directory
49 where pg_upgrade resides; environment variable PGBINNEW
50 -c
52 --check
53 check clusters only, don't change any data
54 -d configdir
56 --old-datadir=configdir
57 the old database cluster configuration directory; environment
58 variable PGDATAOLD
59 -D configdir
61 --new-datadir=configdir
62 the new database cluster configuration directory; environment
63 variable PGDATANEW
64 -j njobs
66 --jobs=njobs
67 number of simultaneous processes or threads to use
68 -k
70 --link
71 use hard links instead of copying files to the new cluster
72 -N
74 --no-sync
75 By default, pg_upgrade will wait for all files of the upgraded
76 cluster to be written safely to disk. This option causes pg_upgrade
77 to return without waiting, which is faster, but means that a
78 subsequent operating system crash can leave the data directory
79 corrupt. Generally, this option is useful for testing but should
80 not be used on a production installation.
81 -o options
83 --old-options options
84 options to be passed directly to the old postgres command; multiple
85 option invocations are appended
86 -O options
88 --new-options options
89 options to be passed directly to the new postgres command; multiple
90 option invocations are appended
91 -p port
93 --old-port=port
94 the old cluster port number; environment variable PGPORTOLD
95 -P port
97 --new-port=port
98 the new cluster port number; environment variable PGPORTNEW
99 -r
101 --retain
102 retain SQL and log files even after successful completion
103 -s dir
105 --socketdir=dir
106 directory to use for postmaster sockets during upgrade; default is
107 current working directory; environment variable PGSOCKETDIR
108 -U username
110 --username=username
111 cluster's install user name; environment variable PGUSER
112 -v
114 --verbose
115 enable verbose internal logging
116 -V
118 --version
119 display version information, then exit
120 --clone
122 Use efficient file cloning (also known as “reflinks” on some
123 systems) instead of copying files to the new cluster. This can
124 result in near-instantaneous copying of the data files, giving the
125 speed advantages of -k/--link while leaving the old cluster
126 untouched.
127 File cloning is only supported on some operating systems and file
129 systems. If it is selected but not supported, the pg_upgrade run
130 will error. At present, it is supported on Linux (kernel 4.5 or
131 later) with Btrfs and XFS (on file systems created with reflink
132 support), and on macOS with APFS.
133 -?
135 --help
136 show help, then exit
137
139 These are the steps to perform an upgrade with pg_upgrade:
140 1. Optionally move the old cluster: If you are using a
142 version-specific installation directory, e.g., /opt/PostgreSQL/15,
143 you do not need to move the old cluster. The graphical installers
144 all use version-specific installation directories.
145 If your installation directory is not version-specific, e.g.,
147 /usr/local/pgsql, it is necessary to move the current PostgreSQL
148 install directory so it does not interfere with the new PostgreSQL
149 installation. Once the current PostgreSQL server is shut down, it
150 is safe to rename the PostgreSQL installation directory; assuming
151 the old directory is /usr/local/pgsql, you can do:
152 mv /usr/local/pgsql /usr/local/pgsql.old
154 to rename the directory.
156 2. For source installs, build the new version: Build the new
158 PostgreSQL source with configure flags that are compatible with the
159 old cluster. pg_upgrade will check pg_controldata to make sure all
160 settings are compatible before starting the upgrade.
161 3. Install the new PostgreSQL binaries: Install the new server's
163 binaries and support files. pg_upgrade is included in a default
164 installation.
165 For source installs, if you wish to install the new server in a
167 custom location, use the prefix variable:
168 make prefix=/usr/local/pgsql.new install
170 4. Initialize the new PostgreSQL cluster: Initialize the new cluster
172 using initdb. Again, use compatible initdb flags that match the old
173 cluster. Many prebuilt installers do this step automatically. There
174 is no need to start the new cluster.
175 5. Install extension shared object files: Many extensions and custom
177 modules, whether from contrib or another source, use shared object
178 files (or DLLs), e.g., pgcrypto.so. If the old cluster used these,
179 shared object files matching the new server binary must be
180 installed in the new cluster, usually via operating system
181 commands. Do not load the schema definitions, e.g., CREATE
182 EXTENSION pgcrypto, because these will be duplicated from the old
183 cluster. If extension updates are available, pg_upgrade will report
184 this and create a script that can be run later to update them.
185 6. Copy custom full-text search files: Copy any custom full text
187 search files (dictionary, synonym, thesaurus, stop words) from the
188 old to the new cluster.
189 7. Adjust authentication: pg_upgrade will connect to the old and new
191 servers several times, so you might want to set authentication to
192 peer in pg_hba.conf or use a ~/.pgpass file (see Section 34.16).
193 8. Stop both servers: Make sure both database servers are stopped
195 using, on Unix, e.g.:
196 pg_ctl -D /opt/PostgreSQL/9.6 stop
198 pg_ctl -D /opt/PostgreSQL/15 stop
199 or on Windows, using the proper service names:
201 NET STOP postgresql-9.6
203 NET STOP postgresql-15
204 Streaming replication and log-shipping standby servers can remain
206 running until a later step.
207 9. Prepare for standby server upgrades: If you are upgrading standby
209 servers using methods outlined in section Step 11, verify that the
210 old standby servers are caught up by running pg_controldata against
211 the old primary and standby clusters. Verify that the “Latest
212 checkpoint location” values match in all clusters. (There will be a
213 mismatch if old standby servers were shut down before the old
214 primary or if the old standby servers are still running.) Also,
215 make sure wal_level is not set to minimal in the postgresql.conf
216 file on the new primary cluster.
217 10. Run pg_upgrade: Always run the pg_upgrade binary of the new
219 server, not the old one. pg_upgrade requires the specification of
220 the old and new cluster's data and executable (bin) directories.
221 You can also specify user and port values, and whether you want the
222 data files linked or cloned instead of the default copy behavior.
223 If you use link mode, the upgrade will be much faster (no file
225 copying) and use less disk space, but you will not be able to
226 access your old cluster once you start the new cluster after the
227 upgrade. Link mode also requires that the old and new cluster data
228 directories be in the same file system. (Tablespaces and pg_wal can
229 be on different file systems.) Clone mode provides the same speed
230 and disk space advantages but does not cause the old cluster to be
231 unusable once the new cluster is started. Clone mode also requires
232 that the old and new data directories be in the same file system.
233 This mode is only available on certain operating systems and file
234 systems.
235 The --jobs option allows multiple CPU cores to be used for
237 copying/linking of files and to dump and restore database schemas
238 in parallel; a good place to start is the maximum of the number of
239 CPU cores and tablespaces. This option can dramatically reduce the
240 time to upgrade a multi-database server running on a multiprocessor
241 machine.
242 For Windows users, you must be logged into an administrative
244 account, and then start a shell as the postgres user and set the
245 proper path:
246 RUNAS /USER:postgres "CMD.EXE"
248 SET PATH=%PATH%;C:\Program Files\PostgreSQL\15\bin;
249 and then run pg_upgrade with quoted directories, e.g.:
251 pg_upgrade.exe
253 --old-datadir "C:/Program Files/PostgreSQL/9.6/data"
254 --new-datadir "C:/Program Files/PostgreSQL/15/data"
255 --old-bindir "C:/Program Files/PostgreSQL/9.6/bin"
256 --new-bindir "C:/Program Files/PostgreSQL/15/bin"
257 Once started, pg_upgrade will verify the two clusters are
259 compatible and then do the upgrade. You can use pg_upgrade --check
260 to perform only the checks, even if the old server is still
261 running. pg_upgrade --check will also outline any manual
262 adjustments you will need to make after the upgrade. If you are
263 going to be using link or clone mode, you should use the option
264 --link or --clone with --check to enable mode-specific checks.
265 pg_upgrade requires write permission in the current directory.
266 Obviously, no one should be accessing the clusters during the
268 upgrade. pg_upgrade defaults to running servers on port 50432 to
269 avoid unintended client connections. You can use the same port
270 number for both clusters when doing an upgrade because the old and
271 new clusters will not be running at the same time. However, when
272 checking an old running server, the old and new port numbers must
273 be different.
274 If an error occurs while restoring the database schema, pg_upgrade
276 will exit and you will have to revert to the old cluster as
277 outlined in Step 17 below. To try pg_upgrade again, you will need
278 to modify the old cluster so the pg_upgrade schema restore
279 succeeds. If the problem is a contrib module, you might need to
280 uninstall the contrib module from the old cluster and install it in
281 the new cluster after the upgrade, assuming the module is not being
282 used to store user data.
283 11. Upgrade streaming replication and log-shipping standby servers: If
285 you used link mode and have Streaming Replication (see
286 Section 27.2.5) or Log-Shipping (see Section 27.2) standby servers,
287 you can follow these steps to quickly upgrade them. You will not be
288 running pg_upgrade on the standby servers, but rather rsync on the
289 primary. Do not start any servers yet.
290 If you did not use link mode, do not have or do not want to use
292 rsync, or want an easier solution, skip the instructions in this
293 section and simply recreate the standby servers once pg_upgrade
294 completes and the new primary is running.
295 1. Install the new PostgreSQL binaries on standby servers: Make
297 sure the new binaries and support files are installed on all
298 standby servers.
299 2. Make sure the new standby data directories do not exist: Make
301 sure the new standby data directories do not exist or are
302 empty. If initdb was run, delete the standby servers' new data
303 directories.
304 3. Install extension shared object files: Install the same
306 extension shared object files on the new standbys that you
307 installed in the new primary cluster.
308 4. Stop standby servers: If the standby servers are still running,
310 stop them now using the above instructions.
311 5. Save configuration files: Save any configuration files from the
313 old standbys' configuration directories you need to keep, e.g.,
314 postgresql.conf (and any files included by it),
315 postgresql.auto.conf, pg_hba.conf, because these will be
316 overwritten or removed in the next step.
317 6. Run rsync: When using link mode, standby servers can be quickly
319 upgraded using rsync. To accomplish this, from a directory on
320 the primary server that is above the old and new database
321 cluster directories, run this on the primary for each standby
322 server:
323 rsync --archive --delete --hard-links --size-only --no-inc-recursive old_cluster new_cluster remote_dir
325 where old_cluster and new_cluster are relative to the current
327 directory on the primary, and remote_dir is above the old and
328 new cluster directories on the standby. The directory structure
329 under the specified directories on the primary and standbys
330 must match. Consult the rsync manual page for details on
331 specifying the remote directory, e.g.,
332 rsync --archive --delete --hard-links --size-only --no-inc-recursive /opt/PostgreSQL/9.5 \
334 /opt/PostgreSQL/9.6 standby.example.com:/opt/PostgreSQL
335 You can verify what the command will do using rsync's --dry-run
337 option. While rsync must be run on the primary for at least one
338 standby, it is possible to run rsync on an upgraded standby to
339 upgrade other standbys, as long as the upgraded standby has not
340 been started.
341 What this does is to record the links created by pg_upgrade's
343 link mode that connect files in the old and new clusters on the
344 primary server. It then finds matching files in the standby's
345 old cluster and creates links for them in the standby's new
346 cluster. Files that were not linked on the primary are copied
347 from the primary to the standby. (They are usually small.) This
348 provides rapid standby upgrades. Unfortunately, rsync
349 needlessly copies files associated with temporary and unlogged
350 tables because these files don't normally exist on standby
351 servers.
352 If you have tablespaces, you will need to run a similar rsync
354 command for each tablespace directory, e.g.:
355 rsync --archive --delete --hard-links --size-only --no-inc-recursive /vol1/pg_tblsp/PG_9.5_201510051 \
357 /vol1/pg_tblsp/PG_9.6_201608131 standby.example.com:/vol1/pg_tblsp
358 If you have relocated pg_wal outside the data directories,
360 rsync must be run on those directories too.
361 7. Configure streaming replication and log-shipping standby
363 servers: Configure the servers for log shipping. (You do not
364 need to run pg_backup_start() and pg_backup_stop() or take a
365 file system backup as the standbys are still synchronized with
366 the primary.) Replication slots are not copied and must be
367 recreated.
368 12. Restore pg_hba.conf: If you modified pg_hba.conf, restore its
370 original settings. It might also be necessary to adjust other
371 configuration files in the new cluster to match the old cluster,
372 e.g., postgresql.conf (and any files included by it),
373 postgresql.auto.conf.
374 13. Start the new server: The new server can now be safely started,
376 and then any rsync'ed standby servers.
377 14. Post-upgrade processing: If any post-upgrade processing is
379 required, pg_upgrade will issue warnings as it completes. It will
380 also generate script files that must be run by the administrator.
381 The script files will connect to each database that needs
382 post-upgrade processing. Each script should be run using:
383 psql --username=postgres --file=script.sql postgres
385 The scripts can be run in any order and can be deleted once they
387 have been run.
388 Caution
390 In general it is unsafe to access tables referenced in rebuild
391 scripts until the rebuild scripts have run to completion; doing
392 so could yield incorrect results or poor performance. Tables
393 not referenced in rebuild scripts can be accessed immediately.
394 15. Statistics: Because optimizer statistics are not transferred by
396 pg_upgrade, you will be instructed to run a command to regenerate
397 that information at the end of the upgrade. You might need to set
398 connection parameters to match your new cluster.
399 16. Delete old cluster: Once you are satisfied with the upgrade, you
401 can delete the old cluster's data directories by running the script
402 mentioned when pg_upgrade completes. (Automatic deletion is not
403 possible if you have user-defined tablespaces inside the old data
404 directory.) You can also delete the old installation directories
405 (e.g., bin, share).
406 17. Reverting to old cluster: If, after running pg_upgrade, you wish
408 to revert to the old cluster, there are several options:
409 • If the --check option was used, the old cluster was unmodified;
411 it can be restarted.
412 • If the --link option was not used, the old cluster was
414 unmodified; it can be restarted.
415 • If the --link option was used, the data files might be shared
417 between the old and new cluster:
418 • If pg_upgrade aborted before linking started, the old
420 cluster was unmodified; it can be restarted.
421 • If you did not start the new cluster, the old cluster was
423 unmodified except that, when linking started, a .old suffix
424 was appended to $PGDATA/global/pg_control. To reuse the old
425 cluster, remove the .old suffix from
426 $PGDATA/global/pg_control; you can then restart the old
427 cluster.
428 • If you did start the new cluster, it has written to shared
430 files and it is unsafe to use the old cluster. The old
431 cluster will need to be restored from backup in this case.
432
434 pg_upgrade creates various working files, such as schema dumps, stored
435 within pg_upgrade_output.d in the directory of the new cluster. Each
436 run creates a new subdirectory named with a timestamp formatted as per
437 ISO 8601 (%Y%m%dT%H%M%S), where all its generated files are stored.
438 pg_upgrade_output.d and its contained files will be removed
439 automatically if pg_upgrade completes successfully; but in the event of
440 trouble, the files there may provide useful debugging information.
441 pg_upgrade launches short-lived postmasters in the old and new data
443 directories. Temporary Unix socket files for communication with these
444 postmasters are, by default, made in the current working directory. In
445 some situations the path name for the current directory might be too
446 long to be a valid socket name. In that case you can use the -s option
447 to put the socket files in some directory with a shorter path name. For
448 security, be sure that that directory is not readable or writable by
449 any other users. (This is not supported on Windows.)
450 All failure, rebuild, and reindex cases will be reported by pg_upgrade
452 if they affect your installation; post-upgrade scripts to rebuild
453 tables and indexes will be generated automatically. If you are trying
454 to automate the upgrade of many clusters, you should find that clusters
455 with identical database schemas require the same post-upgrade steps for
456 all cluster upgrades; this is because the post-upgrade steps are based
457 on the database schemas, and not user data.
458 For deployment testing, create a schema-only copy of the old cluster,
460 insert dummy data, and upgrade that.
461 pg_upgrade does not support upgrading of databases containing table
463 columns using these reg* OID-referencing system data types:
464 regcollation
465 regconfig
466 regdictionary
467 regnamespace
468 regoper
469 regoperator
470 regproc
471 regprocedure
472 (regclass, regrole, and regtype can be upgraded.)
473 If you want to use link mode and you do not want your old cluster to be
475 modified when the new cluster is started, consider using the clone
476 mode. If that is not available, make a copy of the old cluster and
477 upgrade that in link mode. To make a valid copy of the old cluster, use
478 rsync to create a dirty copy of the old cluster while the server is
479 running, then shut down the old server and run rsync --checksum again
480 to update the copy with any changes to make it consistent. (--checksum
481 is necessary because rsync only has file modification-time granularity
482 of one second.) You might want to exclude some files, e.g.,
483 postmaster.pid, as documented in Section 26.3.3. If your file system
484 supports file system snapshots or copy-on-write file copies, you can
485 use that to make a backup of the old cluster and tablespaces, though
486 the snapshot and copies must be created simultaneously or while the
487 database server is down.
488
490 initdb(1), pg_ctl(1), pg_dump(1), postgres(1)
491PostgreSQL 15.4 2023 PG_UPGRADE(1)