PG_DUMP(1)

📦 version 18.6, in opensuse-leap 16.0

💾 pg_dump — the PostgreSQL 18.6 documentation era 💅

okay bestie, sit DOWN, because we are talking about pg_dump, which is genuinely one of the most iconic little programs in the whole PostgreSQL situation. it exports a PostgreSQL database as an SQL script or to other formats. that's the whole name. that's the whole job. and it does it SO well.

✨ NAME ✨

pg_dump — export a PostgreSQL database as an SQL script or to other formats 🫶

🌸 SYNOPSIS 🌸

pg_dump [connection-option...] [option...] [dbname]

💖 DESCRIPTION 💖

okay so. pg_dump is a utility for exporting a PostgreSQL database. it makes consistent exports even if the database is being used concurrently — like, people can be reading AND writing while you dump, and it still holds it together. no cap. and it does not block other users accessing the database (readers or writers, either one, it's not gatekeeping). 💅

but bestie, real talk: except in simple cases, pg_dump is generally not the right choice for taking regular backups of production databases. see Chapter 25 for further discussion. i know, i know, shocking. but that's what it says and i'm not gonna lie to you. 🙈

pg_dump only dumps a single database. that's the whole vibe. it's a one-database girlie. if you want to export an entire cluster, or export global objects that are common to all databases in a cluster (such as roles and tablespaces), you need pg_dumpall(1) instead. different tool, different job, still iconic.

dumps can be output in script or archive file formats:

  • script dumps are plain-text files containing the SQL commands required to reconstruct the database to the state it was in at the time it was saved. to restore from such a script, feed it to psql(1). and here's the slay: script files can be used to reconstruct the database even on other machines and other architectures — and with some modifications, even on other SQL database products. ✨
  • the alternative archive file formats must be used with pg_restore(1) to rebuild the database. they let pg_restore be selective about what is restored, or even to reorder the items prior to being restored. the archive file formats are designed to be portable across architectures. 🌈

when used with one of the archive file formats and combined with pg_restore, pg_dump provides a flexible archival and transfer mechanism. you can export an entire database with pg_dump, then use pg_restore to examine the archive and/or select which parts of the database are to be restored. and the most flexible output file formats are the "custom" format (-Fc) and the "directory" format (-Fd). they allow for selection and reordering of all archived items, they support parallel restoration, and they are compressed by default. and the "directory" format is the only format that supports parallel dumps. periodt. 👑

while running pg_dump, one should examine the output for any warnings (printed on standard error), especially in light of the limitations listed below. read your warnings, bestie, they're trying to tell you something.

⚠️ Warning ⚠️ restoring a dump causes the destination to execute arbitrary code of the source superusers' choice. and partial dumps and partial restores do not limit that. if the source superusers are not trusted, the dumped SQL statements must be inspected before restoring. non-plain-text dumps can be inspected by using pg_restore's --file option. note that the client running the dump and restore need not trust the source or destination superusers. this one is genuinely serious, okay? i'm being cute but i'm not being cute about this. 💀

🎀 OPTIONS 🎀

the following command-line options control the content and format of the output. (the connection options are a whole other situation, we'll get there.)

  • dbname — specifies the name of the database to be dumped. if this is not specified, the environment variable PGDATABASE is used. and if that is not set, the user name specified for the connection is used. it's a whole fallback chain, she's resourceful. 💁
  • -a, --data-only — dump only the data, not the schema (data definitions) or statistics. table data, large objects, and sequence values are dumped. this option is similar to, but for historical reasons not identical to, specifying --section=data. (yes, they're cousins, not twins. don't @ me.)
  • -b, --large-objects, --blobs (deprecated) — include large objects in the dump. this is the default behavior except when --schema, --table, --schema-only, --statistics-only, or --no-data is specified. so the -b switch is therefore only useful to add large objects to dumps where a specific schema or table has been requested. note that large objects are considered data and therefore will be included when --data-only is used, but not when --schema-only or --statistics-only is. 🧠
  • -B, --no-large-objects, --no-blobs (deprecated) — exclude large objects in the dump. and when both -b and -B are given, the behavior is to output large objects, when data is being dumped — see the -b documentation. so -b wins that fight, fyi.
  • -c, --clean — output commands to DROP all the dumped database objects prior to outputting the commands for creating them. this option is useful when the restore is to overwrite an existing database. if any of the objects do not exist in the destination database, ignorable error messages will be reported during restore, unless --if-exists is also specified. (so if you hate the error spam, bring --if-exists along.)
  • -C, --create — begin the output with a command to create the database itself and reconnect to the created database. (with a script of this form, it doesn't matter which database in the destination installation you connect to before running the script. love that for you.) if --clean is also specified, the script drops and recreates the target database before reconnecting to it.

with --create, the output also includes the database's comment if any, and any configuration variable settings that are specific to this database — that is, any ALTER DATABASE ... SET ... and ALTER ROLE ... IN DATABASE ... SET ... commands that mention this database. access privileges for the database itself are also dumped, unless --no-acl is specified.

this option is ignored when emitting an archive (non-text) output file. for the archive formats, you can specify the option when you call pg_restore.
  • -e pattern, --extension=pattern — dump only extensions matching pattern. when this option is not specified, all non-system extensions in the target database will be dumped. multiple extensions can be selected by writing multiple -e switches. the pattern parameter is interpreted as a pattern according to the same rules used by psql's \d commands (see Patterns), so multiple extensions can also be selected by writing wildcard characters in the pattern. when using wildcards, be careful to quote the pattern if needed to prevent the shell from expanding the wildcards — your shell is a menace and it WILL eat your asterisks. 😭

any configuration relation registered by pg_extension_config_dump is included in the dump if its extension is specified by --extension.

Note 💜 when -e is specified, pg_dump makes no attempt to dump any other database objects that the selected extension(s) might depend upon. therefore, there is no guarantee that the results of a specific-extension dump can be successfully restored by themselves into a clean database. it's giving "you're on your own, bestie."
  • -E encoding, --encoding=encoding — create the dump in the specified character set encoding. by default, the dump is created in the database encoding. (another way to get the same result is to set the PGCLIENTENCODING environment variable to the desired dump encoding.) the supported encodings are described in Section 23.3.1. 🌐
  • -f file, --file=file — send output to the specified file. this parameter can be omitted for file based output formats, in which case the standard output is used. it must be given for the directory output format however, where it specifies the target directory instead of a file. in this case the directory is created by pg_dump and must not exist before. (so don't go pre-making the folder. she wants to do it herself. main character energy.) 🎀
  • -F format, --format=format — selects the format of the output. format can be one of the following:
    • p, plain — output a plain-text SQL script file (the default).
    • c, custom — output a custom-format archive suitable for input into pg_restore. together with the directory output format, this is the most flexible output format in that it allows manual selection and reordering of archived items during restore. this format is also compressed by default.
    • d, directory — output a directory-format archive suitable for input into pg_restore. this will create a directory with one file for each table and large object being dumped, plus a so-called Table of Contents file describing the dumped objects in a machine-readable format that pg_restore can read. a directory format archive can be manipulated with standard Unix tools; for example, files in an uncompressed archive can be compressed with the gzip, lz4, or zstd tools. this format is compressed by default using gzip and also supports parallel dumps. 🔥
    • t, tar — output a tar-format archive suitable for input into pg_restore. the tar format is compatible with the directory format: extracting a tar-format archive produces a valid directory-format archive. however, the tar format does not support compression. also, when using tar format the relative order of table data items cannot be changed during restore. 🙃
  • -j njobs, --jobs=njobs — run the dump in parallel by dumping njobs tables simultaneously. this option may reduce the time needed to perform the dump but it also increases the load on the database server. you can only use this option with the directory output format because this is the only output format where multiple processes can write their data at the same time. (it's the only format that can share, okay?)

pg_dump will open njobs + 1 connections to the database, so make sure your max_connections setting is high enough to accommodate all connections. do the math before you commit, bestie. 🧮

requesting exclusive locks on database objects while running a parallel dump could cause the dump to fail. the reason is that the pg_dump leader process requests shared locks (ACCESS SHARE) on the objects that the worker processes are going to dump later in order to make sure that nobody deletes them and makes them go away while the dump is running. if another client then requests an exclusive lock on a table, that lock will not be granted but will be queued waiting for the shared lock of the leader process to be released. consequently any other access to the table will not be granted either and will queue after the exclusive lock request. this includes the worker process trying to dump the table. without any precautions this would be a classic deadlock situation. 💀 to detect this conflict, the pg_dump worker process requests another shared lock using the NOWAIT option. if the worker process is not granted this shared lock, somebody else must have requested an exclusive lock in the meantime and there is no way to continue with the dump, so pg_dump has no choice but to abort the dump.

to be continued, bestie… 🌸

So, about pg_dump -j real quick, because this is important and also lowkey iconic: parallel dump only works if the database server supports synchronized snapshots, a feature that landed in PostgreSQL 9.2 for primary servers and 10 for standbys. With it, database clients can make sure they see the exact same data set even though they're using different connections. pg_dump -j uses multiple database connections — it connects to the database once with the leader process, and once again for each worker job. Without the synchronized snapshot feature, those different worker jobs wouldn't be guaranteed to see the same data on each connection, which could lead to an inconsistent backup. 💀 Not the vibe.

  • -n pattern, --schema=pattern — dump only schemas matching pattern; this selects both the schema itself and all the objects chillin' inside it. If you don't specify this option, all non-system schemas in the target database get dumped. Wanna pick multiple schemas? Just write multiple -n switches, bestie. The pattern parameter is interpreted as a pattern using the exact same rules as psql's \d commands (see Patterns), so you can also select multiple schemas by throwing wildcard characters into the pattern. When you use wildcards, be careful to quote the pattern if needed so the shell doesn't expand the wildcards first — see Examples below. 🐚
Note ✨ When -n is specified, pg_dump makes no attempt to dump any other database objects that the selected schema(s) might depend upon. So there's genuinely no guarantee that the results of a specific-schema dump can be successfully restored by themselves into a clean database. It's giving "self-contained but maybe not", fr.
Note 🎀 Non-schema objects such as large objects are not dumped when -n is specified. You can add large objects back to the dump with the --large-objects switch.
  • -N pattern, --exclude-schema=pattern — do not dump any schemas matching pattern. The pattern is interpreted using the same rules as for -n. -N can be given more than once, so you can exclude schemas matching any of several patterns. When both -n and -N are given, the behavior is to dump just the schemas that match at least one -n switch but no -N switches. If -N shows up without -n, then schemas matching -N are excluded from what would otherwise be a normal dump. 💅
  • -O, --no-owner — don't output commands to set ownership of objects to match the original database. By default, pg_dump issues ALTER OWNER or SET SESSION AUTHORIZATION statements to set ownership of created database objects. Those statements will fail when the script is run unless it's started by a superuser (or the same user that owns all the objects in the script). To make a script that can be restored by any user — but will give that user ownership of all the objects — specify -O. This option is ignored when emitting an archive (non-text) output file. For the archive formats, you can specify the option when you call pg_restore. 👑
  • -R, --no-reconnect — this option is obsolete but still accepted for backwards compatibility. She's retired but still shows up to the function, iykyk. 🙈
  • -s, --schema-only — dump only the object definitions (schema), not data or statistics. This option cannot be used with --data-only or --statistics-only. It's similar to — but for historical reasons not identical to — specifying --section=pre-data --section=post-data. (Do not confuse this with the --schema option, which uses the word “schema” in a whole different meaning. Same word, different girl. 😭) To exclude table data for only a subset of tables in the database, see --exclude-table-data.
  • -S username, --superuser=username — specify the superuser user name to use when disabling triggers. This is relevant only if --disable-triggers is used. (Usually, it's better to leave this out, and instead start the resulting script as superuser. Just saying. 💁)
  • -t pattern, --table=pattern — dump only tables with names matching pattern. Multiple tables can be selected by writing multiple -t switches. The pattern parameter is interpreted as a pattern using the same rules as psql's \d commands (see Patterns), so multiple tables can also be selected by writing wildcard characters in the pattern. When using wildcards, be careful to quote the pattern if needed so the shell doesn't expand the wildcards — see Examples below. 🌸

As well as tables, this option can be used to dump the definition of matching views, materialized views, foreign tables, and sequences. It will not dump the contents of views or materialized views, and the contents of foreign tables will only be dumped if the corresponding foreign server is specified with --include-foreign-data.

The -n and -N switches have no effect when -t is used, because tables selected by -t will be dumped regardless of those switches, and non-table objects will not be dumped. Periodt.

Note 🦋 When -t is specified, pg_dump makes no attempt to dump any other database objects that the selected table(s) might depend upon. Therefore, there is no guarantee that the results of a specific-table dump can be successfully restored by themselves into a clean database. Same energy as the -n warning, different day.
  • -T pattern, --exclude-table=pattern — do not dump any tables matching pattern. The pattern is interpreted using the same rules as for -t. -T can be given more than once to exclude tables matching any of several patterns. When both -t and -T are given, the behavior is to dump just the tables that match at least one -t switch but no -T switches. If -T appears without -t, then tables matching -T are excluded from what is otherwise a normal dump. ✨
  • -v, --verbose — specifies verbose mode. This makes pg_dump output detailed object comments and start/stop times to the dump file, plus progress messages to standard error. Repeating the option causes additional debug-level messages to appear on standard error. Spam me with the receipts, honestly. 🔥
  • -V, --version — print the pg_dump version and exit. Short and sweet. 💅
  • -x, --no-privileges, --no-acl — prevent dumping of access privileges (grant/revoke commands). No permissions drama in your dump file. 🚫
  • -Z level, -Z method[:detail], --compress=level, --compress=method[:detail] — specify the compression method and/or the compression level to use. The compression method can be set to gzip, lz4, zstd, or none for no compression. A compression detail string can optionally be specified. If the detail string is an integer, it specifies the compression level. Otherwise, it should be a comma-separated list of items, each of the form keyword or keyword=value. Currently, the supported keywords are level and long.

If no compression level is specified, the default compression level will be used. If only a level is specified without mentioning an algorithm, gzip compression will be used if the level is greater than 0, and no compression will be used if the level is 0.

For the custom and directory archive formats, this specifies compression of individual table-data segments, and the default is to compress using gzip at a moderate level. For plain text output, setting a nonzero compression level causes the entire output file to be compressed, as though it had been fed through gzip, lz4, or zstd — but the default is not to compress. With zstd compression, long mode may improve the compression ratio, at the cost of increased memory use. And the tar archive format currently does not support compression at all. Not a single bit. 😭

  • --binary-upgrade — this option is for use by in-place upgrade utilities. Its use for other purposes is not recommended or supported. The behavior of the option may change in future releases without notice. It's giving "for internal use only, don't touch". 💀
  • --column-inserts, --attribute-inserts — dump data as INSERT commands with explicit column names (INSERT INTO table (column, ...) VALUES ...). This will make restoration very slow; it's mainly useful for making dumps that can be loaded into non-PostgreSQL databases. The upside? Any error during restoring will cause only the rows that are part of the problematic INSERT to be lost, rather than the entire table contents. Silver lining moment. 🌈
  • --disable-dollar-quoting — this option disables the use of dollar quoting for function bodies, and forces them to be quoted using SQL standard string syntax. 🎀
  • --disable-triggers — this option is relevant only when creating a dump that includes data but does not include schema. It instructs pg_dump to include commands to temporarily disable triggers on the target tables while the data is restored. Use this if you have referential integrity checks or other triggers on the tables that you do not want to invoke during data restore. Presently, the commands emitted for --disable-triggers must be done as superuser. So you should also specify a superuser name with -S, or preferably be careful to start the resulting script as a superuser. This option is ignored when emitting an archive (non-text) output file. For the archive formats, you can specify the option when you call pg_restore.
  • --enable-row-security — this option is relevant only when dumping the contents of a table which has row security. By default, pg_dump will set row_security to off, to ensure that all data is dumped from the table. If the user does not have sufficient privileges to bypass row security, then an error is thrown. This parameter instructs pg_dump to set row_security to on instead, allowing the user to dump the parts of the contents of the table that they have access to. Note that if you use this option currently, you probably also want the dump to be in INSERT format, as the COPY FROM during restore does not support row security. 🙈
  • --exclude-extension=pattern — do not dump any extensions matching pattern. The pattern is interpreted using the same rules as for -e. --exclude-extension can be given more than once to exclude extensions matching any of several patterns. When both -e and --exclude-extension are given, the behavior is to dump just the extensions that match at least one -e switch but no --exclude-extension switches. If --exclude-extension appears without -e, then extensions matching --exclude-extension are excluded from what is otherwise a normal dump. 💜
  • --exclude-table-and-children=pattern — bestie, this is literally the same as -T/--exclude-table, except it also excludes any partitions or inheritance child tables of the table(s) matching pattern. 🫶
  • --exclude-table-data=pattern — no data for you for any tables matching pattern. The pattern gets interpreted by the exact same rules as -t. You can drop --exclude-table-data more than once and it'll exclude tables matching any of several patterns. Iconic use case: you need a particular table's definition even though you do not need the data inside it. 💅

Wanna exclude data for every single table in the database? Go look at --schema-only or --statistics-only.

  • --exclude-table-data-and-children=pattern — same energy as --exclude-table-data, except it also excludes the data of any partitions or inheritance child tables of the table(s) matching pattern. 🎀
  • --extra-float-digits=ndigits — use the value you give here as extra_float_digits when dumping floating-point data, instead of the maximum available precision. Heads up bestie: routine dumps made for backup purposes should not use this option. 🙈
  • --filter=filename — give it a filename and it'll read patterns from there for objects to include or exclude from the dump. Those patterns get interpreted by the same rules as their corresponding options: -t/--table, --table-and-children, -T/--exclude-table, and --exclude-table-and-children for tables; -n/--schema and -N/--exclude-schema for schemas; --include-foreign-data for data on foreign servers; --exclude-table-data and --exclude-table-data-and-children for table data; and -e/--extension and --exclude-extension for extensions. Wanna read from STDIN? Use - as the filename. 🌈 You can run --filter right alongside all those include/exclude options above, and you can also specify it more than once for multiple filter files. Periodt.

The file lists one object pattern per row, formatted like this:

{ include | exclude } { extension | foreign_data | table | table_and_children | table_data | table_data_and_children | schema } PATTERN

The first keyword says whether the objects matched by the pattern get included or excluded. The second keyword says what type of object the pattern is filtering: ✨

  • extension: extensions. Works like the -e/--extension or --exclude-extension option.
  • foreign_data: data on foreign servers. Works like the --include-foreign-data option. This keyword can only be used with the include keyword.
  • table: tables. Works like the -t/--table or -T/--exclude-table option.
  • table_and_children: tables including any partitions or inheritance child tables. Works like the --table-and-children or --exclude-table-and-children option.
  • table_data: table data of any tables matching pattern. Works like the --exclude-table-data option. This keyword can only be used with the exclude keyword.
  • table_data_and_children: table data of any tables matching pattern as well as any partitions or inheritance children of the table(s). Works like the --exclude-table-data-and-children option. This keyword can only be used with the exclude keyword.
  • schema: schemas. Works like the -n/--schema or -N/--exclude-schema option.

Lines starting with # are treated as comments and ignored — and yes, you can also stick a comment after an object pattern row. Blank lines get ignored too. For how to do quoting inside patterns, check out Patterns. 📖

Example files are listed below in the Examples section.

  • --if-exists — use DROP ... IF EXISTS commands to drop objects while in --clean mode. This shuts up the “does not exist” errors that might otherwise get reported. Not valid unless --clean is also specified. 🚨
  • --include-foreign-data=foreignserver — dumps the data for any foreign table whose foreign server matches the foreignserver pattern. Wanna grab multiple foreign servers? Just write multiple --include-foreign-data switches. Also, foreignserver is interpreted as a pattern by the same rules psql's \d commands use (see Patterns), so you can also select multiple foreign servers by throwing wildcard characters in the pattern. When you use wildcards, quote the pattern if you need to so the shell doesn't expand them first — see Examples below. The one and only exception: an empty pattern is disallowed. 🚫
Note — using wildcards in --include-foreign-data may result in access to unexpected foreign servers. And to use this option securely, make sure the named server has a trusted owner. 🔐
Note — when --include-foreign-data is specified, pg_dump does not check that the foreign table is writable. So there's no guarantee that the results of a foreign table dump can be successfully restored. 😭
  • --inserts — dump data as INSERT commands (instead of COPY). This makes restoration very slow; it's mainly useful for making dumps that can be loaded into non-PostgreSQL databases. The upside: any error during restoring only loses the rows that are part of the problematic INSERT, rather than the entire table contents. But bestie, note the restore might fail altogether if you've rearranged column order. The --column-inserts option is safe against column order changes, though it's even slower. 💀
  • --load-via-partition-root — when dumping data for a table partition, make the COPY or INSERT statements target the root of the partitioning hierarchy that contains it, instead of the partition itself. That way the appropriate partition gets re-determined for each row when the data is loaded. Super useful when you're restoring data on a server where rows don't always land in the same partitions they did on the original server. When does that even happen? Say the partitioning column is of type text and the two systems have different definitions of the collation used to sort the partitioning column. 🦋
  • --lock-wait-timeout=timeout — do not wait forever to acquire shared table locks at the beginning of the dump. Instead it fails if it can't lock a table within the specified timeout. The timeout can be given in any of the formats accepted by SET statement_timeout. (Allowed formats vary depending on the server version you're dumping from, but an integer number of milliseconds is accepted by all versions.) ⏳
  • --no-comments — don't dump COMMENT commands. 🙅
  • --no-data — don't dump data. 🙅
  • --no-policies — don't dump row security policies. 🙅
  • --no-publications — don't dump publications. 🙅
  • --no-schema — don't dump schema (data definitions). 🙅
  • --no-security-labels — don't dump security labels. 🙅
  • --no-statistics — don't dump statistics. This is the default. ✨
  • --no-subscriptions — don't dump subscriptions. 🙅
  • --no-sync — by default, pg_dump waits for all files to be written safely to disk. This option makes pg_dump return without waiting, which is faster, but means a subsequent operating system crash can leave the dump corrupt. Generally this option is useful for testing but should not be used when dumping data from a production installation. 😬
  • --no-table-access-method — don't output commands to select table access methods. With this option, all objects get created with whichever table access method is the default during restore. 🎀

This option is ignored when emitting an archive (non-text) output file. For the archive formats, you can specify the option when you call pg_restore.

  • --no-tablespaces — don't output commands to select tablespaces. With this option, all objects get created in whichever tablespace is the default during restore. 🎀

This option is ignored when emitting an archive (non-text) output file. For the archive formats, you can specify the option when you call pg_restore.

  • --no-toast-compression — don't output commands to set TOAST compression methods. With this option, all columns get restored with the default compression setting. 🫶
  • --no-unlogged-table-data — don't dump the contents of unlogged tables and sequences. This option has no effect on whether the table and sequence definitions (schema) get dumped; it only suppresses dumping the table and sequence data. FYI: data in unlogged tables and sequences is always excluded when dumping from a standby server. 💅
  • --on-conflict-do-nothing — adds ON CONFLICT DO NOTHING to INSERT commands. Not valid unless --inserts, --column-inserts or --rows-per-insert is also specified. 🚨
  • --quote-all-identifiers — force quoting of all identifiers. This one is recommended when you're dumping a database from a server whose PostgreSQL major version differs from pg_dump's, or when the output is meant to be loaded into a server of a different major version. By default, pg_dump only quotes identifiers that are reserved words in its own major version — which sometimes causes compatibility issues when you're dealing with servers of other versions that may have slightly different sets of reserved words. Using --quote-all-identifiers prevents those issues, at the price of a harder-to-read dump script. 📜
  • --restrict-key=restrict_key — use the string you provide as the psql \restrict key in the dump output. This can only be specified for plain-text dumps, i.e., when --format is set to plain or the --format option is omitted. If no restrict key is specified, pg_dump will generate a random one as needed. Keys may contain only alphanumeric characters. 🔑

This option is primarily intended for testing purposes and other scenarios that require repeatable output (e.g., comparing dump files). It is not recommended for general use, because a malicious server with advance knowledge of the key may be able to inject arbitrary code that will be executed on the machine that runs psql with the dump output. 😱

  • --rows-per-insert=nrows — okay so this one dumps the data as INSERT commands instead of COPY, and it lets you cap how many rows go in each INSERT. The value you pass must be a number bigger than zero, no zero, no negative, don't try it 💅. The cute part? If restoring hits an error, only the rows chillin' inside that one problematic INSERT get lost — not the whole entire table's contents. glow-up fr fr ✨
  • --section=sectionname — only dump the section you name. The section name can be pre-data, data, or post-data. You can pass this option more than once to grab multiple sections at once, slay. Default vibe is dump all the sections.
    • The data section holds the actual table data, large-object contents, sequence values, and statistics for tables, materialized views, and foreign tables.
    • Post-data items include definitions of indexes, triggers, rules, statistics for indexes, and constraints other than validated check and not-null constraints.
    • Pre-data items include all other data definition items.
  • --sequence-data — include sequence data in the dump. This is the default behavior ✨ except when --no-data, --schema-only, or --statistics-only is specified.
  • --serializable-deferrable — use a serializable transaction for the dump so the snapshot used is consistent with later database states 🦋. But it does this by waiting for a point in the transaction stream where no anomalies can be present, so there's no risk of the dump failing or making other transactions roll back with a serialization_failure. See Chapter 13 for more on transaction isolation and concurrency control.
Heads up bestie: this option is not beneficial for a dump that's only meant for disaster recovery.

It could be useful for a dump used to load a copy of the database for reporting or other read-only load sharing while the original database keeps getting updated. Without it, the dump may reflect a state that isn't consistent with any serial execution of the transactions that eventually commit. Example: if batch processing techniques are used, a batch may show as closed in the dump without all of the items that are in the batch appearing. 😭

This option makes no difference if there are no read-write transactions active when pg_dump is started. If read-write transactions are active, the start of the dump may be delayed for an indeterminate length of time. Once running, performance with or without the switch is the same. periodt.

  • --snapshot=snapshotname — use the specified synchronized snapshot when making a dump of the database (see Table 9.100 for more details) 🎀.

This is the move when you need to synchronize the dump with a logical replication slot (see Chapter 47) or with a concurrent session.

And if it's a parallel dump? The snapshot name defined by this option is used instead of taking a new snapshot. iconic.

  • --statistics — dump optimizer statistics. 📊
  • --statistics-only — dump only the statistics, not the schema (data definitions) or data. Optimizer statistics for tables, materialized views, foreign tables, and indexes are dumped. ✨
  • --strict-names — require that each extension (-e/--extension), schema (-n/--schema) and table (-t/--table) pattern match at least one extension/schema/table in the database being dumped. This also applies to filters used with --filter. Note that if none of the extension/schema/table patterns find matches, pg_dump will generate an error even without --strict-names. 💀

This option has no effect on --exclude-extension, -N/--exclude-schema, -T/--exclude-table, or --exclude-table-data. An exclude pattern failing to match any objects is not considered an error. 🙈

  • --sync-method=method — when set to fsync, which is the default, pg_dump --format=directory will recursively open and synchronize all files in the archive directory.

On Linux, syncfs may be used instead to ask the operating system to synchronize the whole file system that contains the archive directory. See recovery_init_sync_method for info about the caveats to be aware of when using syncfs. 🫶

This option has no effect when --no-sync is used or --format is not set to directory.

  • --table-and-children=pattern — this is the same as the -t/--table option, except it also includes any partitions or inheritance child tables of the table(s) matching the pattern. family reunion energy 👑
  • --use-set-session-authorization — output SQL-standard SET SESSION AUTHORIZATION commands instead of ALTER OWNER commands to determine object ownership. This makes the dump more standards-compatible, but depending on the history of the objects in the dump, might not restore properly 😭. Also, a dump using SET SESSION AUTHORIZATION will certainly require superuser privileges to restore correctly, whereas ALTER OWNER requires lesser privileges.
  • -?, --help — show help about pg_dump command line arguments, and exit. 💖

The following command-line options control the database connection parameters.

  • -d dbname, --dbname=dbname — specifies the name of the database to connect to. This is equivalent to specifying dbname as the first non-option argument on the command line. The dbname can be a connection string. If so, connection string parameters will override any conflicting command line options. sneaky but valid 💅
  • -h host, --host=host — specifies the host name of the machine on which the server is running. If the value begins with a slash, it is used as the directory for the Unix domain socket. The default is taken from the PGHOST environment variable, if set, else a Unix domain socket connection is attempted. 🌸
  • -p port, --port=port — specifies the TCP port or local Unix domain socket file extension on which the server is listening for connections. Defaults to the PGPORT environment variable, if set, or a compiled-in default. ✨
  • -U username, --username=username — user name to connect as.
  • -w, --no-password — never issue a password prompt. If the server requires password authentication and a password is not available by other means such as a .pgpass file, the connection attempt will fail 💀. This option can be useful in batch jobs and scripts where no user is present to enter a password.
  • -W, --password — force pg_dump to prompt for a password before connecting to a database.

This option is never essential, since pg_dump will automatically prompt for a password if the server demands password authentication. However, pg_dump will waste a connection attempt finding out that the server wants a password. In some cases it is worth typing -W to avoid the extra connection attempt. lowkey efficient 🎀

  • --role=rolename — specifies a role name to be used to create the dump. This option causes pg_dump to issue a SET ROLE rolename command after connecting to the database. It's useful when the authenticated user (specified by -U) lacks privileges needed by pg_dump, but can switch to a role with the required rights. Some installations have a policy against logging in directly as a superuser, and use of this option allows dumps to be made without violating the policy. ✨

🌈 ENVIRONMENT 🌈

  • PGDATABASE
  • PGHOST
  • PGOPTIONS
  • PGPORT
  • PGUSER

These are default connection parameters. 💖

  • PG_COLOR — specifies whether to use color in diagnostic messages. Possible values are always, auto and never. 🌸

This utility, like most other PostgreSQL utilities, also uses the environment variables supported by libpq (see Section 32.15).

💅 DIAGNOSTICS 💅

pg_dump internally executes SELECT statements. So if you're having problems running pg_dump, make sure you're able to select information from the database using, for example, psql(1). Also, any default connection settings and environment variables used by the libpq front-end library will apply. 🦋

The database activity of pg_dump is normally collected by the cumulative statistics system. If that's not the vibe, you can set parameter track_counts to false via PGOPTIONS or the ALTER USER command. ✨

💅 NOTES 💅

ok so real talk, bestie — this section is the "read before you wreck your whole weekend" part, and honestly it's giving ✨wisdom✨. buckle up 🎀

If your database cluster has any local additions to the template1 database, be careful when restoring the output of pg_dump into a truly empty database — otherwise you're lowkey likely to get errors because the added objects get defined twice. 🙈 To make an empty database with no local additions, copy from template0 and not template1, like this:

CREATE DATABASE foo WITH TEMPLATE template0;

When you pick a dump without schema and also use the --disable-triggers option, pg_dump emits commands to disable triggers on user tables before it inserts the data, and then emits commands to re-enable them after the data has been inserted. 🫶 So far so cute — but if the restore gets stopped in the middle, the system catalogs might be left in the wrong state. 💀 not the vibe.

When --statistics is specified, pg_dump will include most optimizer statistics in the resulting dump file. Keyword: most. It does not include all statistics — like the ones created explicitly with CREATE STATISTICS (CREATE_STATISTICS(7)), custom statistics added by an extension, or statistics collected by the cumulative statistics system. So it may still be useful to run ANALYZE after restoring from a dump file to make sure performance is optimal; see Section 24.1.3 and Section 24.1.6 for more info. 🌸

Because pg_dump is used to transfer data to newer versions of PostgreSQL, the output of pg_dump can be expected to load into PostgreSQL server versions newer than pg_dump's version. pg_dump can also dump from PostgreSQL servers older than its own version (currently, servers back to version 9.2 are supported). 👑

However — and this is the part where we pay attention — pg_dump cannot dump from PostgreSQL servers newer than its own major version. It will refuse to even try, rather than risk making an invalid dump. no cap, it just says no. 🚫

Also, it is not guaranteed that pg_dump's output can be loaded into a server of an older major version — not even if the dump was taken from a server of that version. Loading a dump file into an older server may require manual editing of the dump file to remove syntax the older server doesn't understand. Use of the --quote-all-identifiers option is recommended in cross-version cases, because it can prevent problems that arise from varying reserved-word lists in different PostgreSQL versions. 🌈

When dumping logical replication subscriptions, pg_dump will generate CREATE SUBSCRIPTION commands that use the connect = false option, so that restoring the subscription does not make remote connections for creating a replication slot or for the initial table copy. That way, the dump can be restored without requiring network access to the remote servers. ✨ It is then up to you (the user, the main character) to reactivate the subscriptions in a suitable way. If the involved hosts have changed, the connection information might have to be changed. It might also be appropriate to truncate the target tables before initiating a new full table copy. If users intend to copy initial data during refresh they must create the slot with two_phase = false. After the initial sync, the two_phase option will be automatically enabled by the subscriber if the subscription had been originally created with the two_phase = true option. 🦋

And finally, the general recommendation: use the -X (--no-psqlrc) option when restoring a database from a plain-text pg_dump script. It keeps the restore process clean and prevents potential conflicts with non-default psql configurations. chef's kiss, honestly. 💖

💖 EXAMPLES 💖

okay this is the part everyone actually reads, and honestly? iconic. here's every single one. 🎀

To dump a database called mydb into an SQL-script file:

$ pg_dump mydb > db.sql

To reload such a script into a (freshly created) database named newdb:

$ psql -X -d newdb -f db.sql

To dump a database into a custom-format archive file:

$ pg_dump -Fc mydb > db.dump

To dump a database into a directory-format archive:

$ pg_dump -Fd mydb -f dumpdir

To dump a database into a directory-format archive in parallel with 5 worker jobs:

$ pg_dump -Fd mydb -j 5 -f dumpdir

To reload an archive file into a (freshly created) database named newdb:

$ pg_restore -d newdb db.dump

To reload an archive file into the same database it was dumped from, discarding the current contents of that database:

$ pg_restore -d postgres --clean --create db.dump

To dump a single table named mytab:

$ pg_dump -t mytab mydb > db.sql

To dump all tables whose names start with emp in the detroit schema, except for the table named employee_log:

$ pg_dump -t 'detroit.emp*' -T detroit.employee_log mydb > db.sql

To dump all schemas whose names start with east or west and end in gsm, excluding any schemas whose names contain the word test:

$ pg_dump -n 'east*gsm' -n 'west*gsm' -N '*test*' mydb > db.sql

The same, using regular expression notation to consolidate the switches:

$ pg_dump -n '(east|west)*gsm' -N '*test*' mydb > db.sql

To dump all database objects except for tables whose names begin with ts_:

$ pg_dump -T 'ts_*' mydb > db.sql

To specify an upper-case or mixed-case name in -t and related switches, you need to double-quote the name; else it will be folded to lower case (see Patterns). But double quotes are special to the shell, so in turn they must be quoted. 🐚 Thus, to dump a single table with a mixed-case name, you need something like

$ pg_dump -t "\"MixedCaseName\"" mydb > mytab.sql

To dump all tables whose names start with mytable, except for table mytable2, specify a filter file filter.txt like:

include table mytable*
exclude table mytable2
$ pg_dump --filter=filter.txt mydb > db.sql

🦋 SEE ALSO 🦋

and that's the whole page bestie!! 🫶 go forth and dump responsibly, you absolute database legend. this was PostgreSQL 18.6, dated 2026 — mwah 💅✨

📦 where this page comes from