pg_receivewal gains a new option, --compression-method=lz4, available
when the code is compiled with --with-lz4. Similarly to gzip, this
gives the possibility to compress archived WAL segments with LZ4. This
option is not compatible with --compress.
The implementation uses LZ4 frames, and is compatible with simple lz4
commands. Like gzip, using --synchronous ensures that any data will be
flushed to disk within the current .partial segment, so as it is
possible to retrieve as much WAL data as possible even from a
non-completed segment (this requires completing the partial file with
zeros up to the WAL segment size supported by the backend after
decompression, but this is the same as gzip).
The calculation of the streaming start LSN is able to transparently find
and check LZ4-compressed segments. Contrary to gzip where the
uncompressed size is directly stored in the object read, the LZ4 chunk
protocol does not store the uncompressed data by default. There is
contentSize that can be used with LZ4 frames by that would not help if
using an archive that includes segments compressed with the defaults of
a "lz4" command, where this is not stored. So, this commit has taken
the most extensible approach by decompressing the already-archived
segment to check its uncompressed size, through a blank output buffer in
chunks of 64kB (no actual performance difference noticed with 8kB, 16kB
or 32kB, and the operation in itself is actually fast).
Tests have been added to verify the creation and correctness of the
generated LZ4 files. The latter is achieved by the use of command
"lz4", if found in the environment.
The tar-based WAL method in walmethods.c, used now only by
pg_basebackup, does not know yet about LZ4. Its code could be extended
for this purpose.
Author: Georgios Kokolatos
Reviewed-by: Michael Paquier, Jian Guo, Magnus Hagander, Dilip Kumar
Discussion: https://postgr.es/m/ZCm1J5vfyQ2E6dYvXz8si39HQ2gwxSZ3IpYaVgYa3lUwY88SLapx9EEnOf5uEwrddhx2twG7zYKjVeuP5MwZXCNPybtsGouDsAD1o2L_I5E=@pm.me
The option name was incorrect in one of the error messages, and the
short option 'I' was used in the code but we did not intend things to be
this way. While on it, fix the documentation to refer to a "method",
and not a "level.
Oversights in commit d62bcc8, that I have detected after more review of
the LZ4 patch for pg_receivewal.
pg_receivewal includes since cada1af the option --compress, to allow the
compression of WAL segments using gzip, with a value of 0 (the default)
meaning that no compression can be used.
This commit introduces a new option, called --compression-method, able
to use as values "none", the default, and "gzip", to make things more
extensible. The case of --compress=0 becomes fuzzy with this option
layer, so we have made the choice to make pg_receivewal return an error
when using "none" and a non-zero compression level, meaning that the
authorized values of --compress are now [1,9] instead of [0,9]. Not
specifying --compress with "gzip" as compression method makes
pg_receivewal use the default of zlib instead (Z_DEFAULT_COMPRESSION).
The code in charge of finding the streaming start LSN when scanning the
existing archives is refactored and made more extensible. While on it,
rename "compression" to "compression_level" in walmethods.c, to reduce
the confusion with the introduction of the compression method, even if
the tar method used by pg_basebackup does not rely on the compression
method (yet, at least), but just on the compression level (this area
could be improved more, actually).
This is in preparation for an upcoming patch that adds LZ4 support to
pg_receivewal.
Author: Georgios Kokolatos
Reviewed-by: Michael Paquier, Jian Guo, Magnus Hagander, Dilip Kumar,
Robert Haas
Discussion: https://postgr.es/m/ZCm1J5vfyQ2E6dYvXz8si39HQ2gwxSZ3IpYaVgYa3lUwY88SLapx9EEnOf5uEwrddhx2twG7zYKjVeuP5MwZXCNPybtsGouDsAD1o2L_I5E=@pm.me
Use verbiage like "The name of the table must be distinct from the name
of any other relation (table, sequence, index, view, materialized view,
or foreign table) in the same schema." in the reference pages for all
those object types. The main change here is to mention materialized
views explicitly; although a couple of these pages failed to say
anything at all about name conflicts.
Per suggestion from Daniel Westermann.
Discussion: https://postgr.es/m/ZR0P278MB0920D0946509233459AF0DEFD2889@ZR0P278MB0920.CHEP278.PROD.OUTLOOK.COM
Previously failures of initial connection and logfile open caused pgbench
to proceed the benchmarking, report the incomplete results and exit with
status 2. It didn't make sense to proceed the benchmarking even when
pgbench could not start as prescribed.
This commit improves pgbench so that early errors that occur when
starting benchmark such as those failures should make pgbench exit
immediately with status 1.
Author: Yugo Nagata
Reviewed-by: Fabien COELHO, Kyotaro Horiguchi, Fujii Masao
Discussion: https://postgr.es/m/TYCPR01MB5870057375ACA8A73099C649F5349@TYCPR01MB5870.jpnprd01.prod.outlook.com
A new option "FOR ALL TABLES IN SCHEMA" in Create/Alter Publication allows
one or more schemas to be specified, whose tables are selected by the
publisher for sending the data to the subscriber.
The new syntax allows specifying both the tables and schemas. For example:
CREATE PUBLICATION pub1 FOR TABLE t1,t2,t3, ALL TABLES IN SCHEMA s1,s2;
OR
ALTER PUBLICATION pub1 ADD TABLE t1,t2,t3, ALL TABLES IN SCHEMA s1,s2;
A new system table "pg_publication_namespace" has been added, to maintain
the schemas that the user wants to publish through the publication.
Modified the output plugin (pgoutput) to publish the changes if the
relation is part of schema publication.
Updates pg_dump to identify and dump schema publications. Updates the \d
family of commands to display schema publications and \dRp+ variant will
now display associated schemas if any.
Author: Vignesh C, Hou Zhijie, Amit Kapila
Syntax-Suggested-by: Tom Lane, Alvaro Herrera
Reviewed-by: Greg Nancarrow, Masahiko Sawada, Hou Zhijie, Amit Kapila, Haiying Tang, Ajin Cherian, Rahila Syed, Bharath Rupireddy, Mark Dilger
Tested-by: Haiying Tang
Discussion: https://www.postgresql.org/message-id/CALDaNm0OANxuJ6RXqwZsM1MSY4s19nuH3734j4a72etDwvBETQ@mail.gmail.com
Prior to this patch, when running pg_receivewal, the streaming start
point would be the current location of the archives if anything is
found in the local directory where WAL segments are written, and
pg_receivewal would fall back to the current WAL flush location if there
are no archives, as of the result of an IDENTIFY_SYSTEM command.
If for some reason the WAL files from pg_receivewal were moved, it is
better to try a restart where we left at, which is the replication
slot's restart_lsn instead of skipping right to the current flush
location, to avoid holes in the WAL backed up. This commit changes
pg_receivewal to use the following sequence of methods to determine the
starting streaming LSN:
- Scan the local archives.
- Use the slot's restart_lsn, if supported by the backend and if a slot
is defined.
- Fallback to the current flush LSN as reported by IDENTIFY_SYSTEM.
To keep compatibility with older server versions, we only attempt to use
READ_REPLICATION_SLOT if the backend version is at least 15, and
fallback to the older behavior of streaming from the current flush
LSN if the command is not supported.
Some TAP tests are added to cover this feature.
Author: Ronan Dunklau
Reviewed-by: Kyotaro Horiguchi, Michael Paquier, Bharath Rupireddy
Discussion: https://postgr.es/m/18708360.4lzOvYHigE@aivenronan
The documentation was imprecise about the starting LSN used for WAL
streaming if nothing can be found in the local archive directory
defined with the pg_receivewal command, so be more talkative on this
matter.
Extracted from a larger patch by the same author.
Author: Ronan Dunklau, Michael Paquier
Discussion: https://postgr.es/m/18708360.4lzOvYHigE@aivenronan
Backpatch-through: 10
Avoid calling contrib/amcheck functions with relations that are
unsuitable for checking. Specifically, don't attempt verification of
temporary relations, or indexes whose pg_index entry indicates that the
index is invalid, or not ready.
These relations are not supported by any of the contrib/amcheck
functions, for reasons that are pretty fundamental. For example, the
implementation of REINDEX CONCURRENTLY can add its own "transient"
pg_index entries, which has rather unclear implications for the B-Tree
verification functions, at least in the general case -- so they just
treat it as an error. It falls to the amcheck caller (in this case
pg_amcheck) to deal with the situation at a higher level.
pg_amcheck now simply treats these conditions as additional "visibility
concerns" when it queries system catalogs. This is a little arbitrary.
It seems to have the least problems among any of the available
alternatives.
Author: Mark Dilger <mark.dilger@enterprisedb.com>
Reported-By: Alexander Lakhin <exclusion@gmail.com>
Reviewed-By: Peter Geoghegan <pg@bowt.ie>
Reviewed-By: Robert Haas <robertmhaas@gmail.com>
Bug: #17212
Discussion: https://postgr.es/m/17212-34dd4a1d6bba98bf@postgresql.org
Backpatch: 14-, where pg_amcheck was introduced.
This runtime-computed GUC shows the number of huge pages required
for the server's main shared memory area, taking advantage of the
work done in 0c39c29 and 0bd305e. This is useful for users to estimate
the amount of huge pages required for a server as it becomes possible to
do an estimation without having to start the server and potentially
allocate a large chunk of shared memory.
The number of huge pages is calculated based on the existing GUC
huge_page_size if set, or by using the system's default by looking at
/proc/meminfo on Linux. There is nothing new here as this commit reuses
the existing calculation methods, and just exposes this information
directly to the user. The routine calculating the huge page size is
refactored to limit the number of files with platform-specific flags.
This new GUC's name was the most popular choice based on the discussion
done. This is only supported on Linux.
I have taken the time to test the change on Linux, Windows and MacOS,
though for the last two ones large pages are not supported. The first
one calculates correctly the number of pages depending on the existing
GUC huge_page_size or the system's default.
Thanks to Andres Freund, Robert Haas, Kyotaro Horiguchi, Tom Lane,
Justin Pryzby (and anybody forgotten here) for the discussion.
Author: Nathan Bossart
Discussion: https://postgr.es/m/F2772387-CE0F-46BF-B5F1-CC55516EB885@amazon.com
Until now, the -C option of postgres was handled before a small subset
of GUCs computed at runtime are initialized, leading to incorrect
results as GUC machinery would fall back to default values for such
parameters.
For example, data_checksums could report "off" for a cluster as the
control file is not loaded yet. Or wal_segment_size would show a
segment size at 16MB even if initdb --wal-segsize used something else.
Worse, the command would fail to properly report the recently-introduced
shared_memory, that requires to load shared_preload_libraries as these
could ask for a chunk of shared memory.
Support for runtime GUCs comes with a limitation, as the operation is
now allowed on a running server. One notable reason for this is that
_PG_init() functions of loadable libraries are called before all
runtime-computed GUCs are initialized, and this is not guaranteed to be
safe to do on running servers. For the case of shared_memory_size,
where we want to know how much memory would be used without allocating
it, this limitation is fine. Another case where this will help is for
huge pages, with the introduction of a different GUC to evaluate the
amount of huge pages required for a server before starting it, without
having to allocate large chunks of memory.
This feature is controlled with a new GUC flag, and four parameters are
classified as runtime-computed as of this change:
- data_checksums
- shared_memory_size
- data_directory_mode
- wal_segment_size
Some TAP tests are added to provide some coverage here, using
data_checksums in the tests of pg_checksums.
Per discussion with Andres Freund, Justin Pryzby, Magnus Hagander and
more.
Author: Nathan Bossart
Discussion: https://postgr.es/m/F2772387-CE0F-46BF-B5F1-CC55516EB885@amazon.com
The subscription parameters are rearranged into two groups:
a) those that control what happens during Create Subscription
b) those that control the replication behavior
This makes the documentation of Create Subscription easier to follow.
Author: Peter Smith
Reviewed-by: Amit Kapila
Discussion: https://postgr.es/m/CAHut+PtPJDSOxtuMGpO2yDrRPKxcYGL4n7HqJP9HernZE=Cj+g@mail.gmail.com
While populating the pgbench_accounts table, plain COPY was
unconditionally used. By changing it to COPY FREEZE, the time for
VACUUM is significantly reduced, thus the total time of "pgbench -i"
is also reduced. This only happens if pgbench runs against PostgreSQL
14 or later because COPY FREEZE in previous versions of PostgreSQL does
not bring the benefit. Also if partitioning is used, COPY FREEZE
cannot be used. In this case plain COPY will be used too.
Author: Tatsuo Ishii
Discussion: https://postgr.es/m/20210308.143907.2014279678657453983.t-ishii@gmail.com
Reviewed-by: Fabien COELHO, Laurenz Albe, Peter Geoghegan, Dean Rasheed
The current refresh behavior tries to just refresh added/dropped
publications but that leads to removing wrong tables from subscription. We
can't refresh just the dropped publication because it is quite possible
that some of the tables are removed from publication by that time and now
those will remain as part of the subscription. Also, there is a chance
that the tables that were part of the publication being dropped are also
part of another publication, so we can't remove those.
So, we decided that by default, add/drop commands will also act like
REFRESH PUBLICATION which means they will refresh all the publications. We
can keep the old behavior for "add publication" but it is better to be
consistent with "drop publication".
Author: Hou Zhijie
Reviewed-by: Masahiko Sawada, Amit Kapila
Backpatch-through: 14, where it was introduced
Discussion: https://postgr.es/m/OS0PR01MB5716935D4C2CC85A6143073F94EF9@OS0PR01MB5716.jpnprd01.prod.outlook.com
Using --quiet in combination with --no-strict-names didn't work as
documented, a warning message was still emitted. Since the --quiet
flag was working in an unconventional way to other utilities, fix
by removing the functionality instead.
Backpatch through 14 where pg_amcheck was introduced.
Bug: 17148
Reported-by: Chen Jiaoqian <chenjq.jy@fujitsu.com>
Reviewed-by: Julien Rouhaud <rjuju123@gmail.com>
Discussion: https://postgr.es/m/17148-b5087318e2b04fc6@postgresql.org
Backpatch-through: 14
This reverts the following commits:
1b5617eb84 Describe (auto-)analyze behavior for partitioned tables
0e69f705cc Set pg_class.reltuples for partitioned tables
41badeaba8 Document ANALYZE storage parameters for partitioned tables
0827e8af70 autovacuum: handle analyze for partitioned tables
There are efficiency issues in this code when handling databases with
large numbers of partitions, and it doesn't look like there isn't any
trivial way to handle those. There are some other issues as well. It's
now too late in the cycle for nontrivial fixes, so we'll have to let
Postgres 14 users continue to manually deal with ANALYZE their
partitioned tables, and hopefully we can fix the issues for Postgres 15.
I kept [most of] be280cdad2 ("Don't reset relhasindex for partitioned
tables on ANALYZE") because while we added it due to 0827e8af70, it is
a good bugfix in its own right, since it affects manual analyze as well
as autovacuum-induced analyze, and there's no reason to revert it.
I retained the addition of relkind 'p' to tables included by
pg_stat_user_tables, because reverting that would require a catversion
bump.
Also, in pg14 only, I keep a struct member that was added to
PgStat_TabStatEntry to avoid breaking compatibility with existing stat
files.
Backpatch to 14.
Discussion: https://postgr.es/m/20210722205458.f2bug3z6qzxzpx2s@alap3.anarazel.de
The initdb help message for --sync-only was a bit terse, and not
really self-explanatory. Make it clearer that initdb --sync-only
will exit after syncing, and expand the docs with a note on when
the option can be useful. Also align the help output with others
that exit immediately.
Author: Nathan Bossart, Gurjeet Singh
Discussion: https://postgr.es/m/CABwTF4U6hbNNE1bv=LxQdJybmUdZ5NJQ9rKY9tN82NXM8QH+iQ@mail.gmail.com
Since commit e462856a7a, pg_upgrade automatically creates a script to
update extensions, so mention that instead of ALTER EXTENSION.
Backpatch-through: 9.6
Commit a8fd13cab0 added support for prepared transactions to built-in
logical replication via a new option "two_phase" for a subscription. The
"two_phase" option was not allowed with the existing streaming option.
This commit permits the combination of "streaming" and "two_phase"
subscription options. It extends the pgoutput plugin and the subscriber
side code to add the prepare API for streaming transactions which will
apply the changes accumulated in the spool-file at prepare time.
Author: Peter Smith and Ajin Cherian
Reviewed-by: Vignesh C, Amit Kapila, Greg Nancarrow
Tested-By: Haiying Tang
Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru
Discussion: https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com
It wasn't all that clear which lock levels, if any, would be held on the
DEFAULT partition during an ATTACH PARTITION operation.
Also, clarify which locks will be taken if the DEFAULT partition or the
table being attached are themselves partitioned tables.
Here I'm only backpatching to v12 as before then we obtained an ACCESS
EXCLUSIVE lock on the partitioned table. It seems much less relevant to
mention which locks are taken on other tables when the partitioned table
itself is locked with an ACCESS EXCLUSIVE lock.
Author: Matthias van de Meent, David Rowley
Discussion: https://postgr.es/m/CAEze2WiTB6iwrV8W_J=fnrnZ7fowW3qu-8iQ8zCHP3FiQ6+o-A@mail.gmail.com
Backpatch-through: 12
The logic used to support a change of access method for a table is
similar to changes for tablespace or relation persistence, requiring a
table rewrite with an exclusive lock of the relation changed. Table
rewrites done in ALTER TABLE already go through the table AM layer when
scanning tuples from the old relation and inserting them into the new
one, making this implementation straight-forward.
Note that partitioned tables are not supported as these have no access
methods defined.
Author: Justin Pryzby, Jeff Davis
Reviewed-by: Michael Paquier, Vignesh C
Discussion: https://postgr.es/m/20210228222530.GD20769@telsasoft.com
Add pg_resetxlog -u option to set the oldest xid in pg_control.
Previously -x set this value be -2 billion less than the -x value.
However, this causes the server to immediately scan all relation's
relfrozenxid so it can advance pg_control's oldest xid to be inside the
autovacuum_freeze_max_age range, which is inefficient and might disrupt
diagnostic recovery. pg_upgrade will use this option to better create
the new cluster to match the old cluster.
Reported-by: Jason Harvey, Floris Van Nee
Discussion: https://postgr.es/m/20190615183759.GB239428@rfd.leadboat.com, 87da83168c644fd9aae38f546cc70295@opammb0562.comp.optiver.com
Author: Bertrand Drouvot
Backpatch-through: 9.6
Renaming triggers on partitioned tables had two problems: first,
it did not recurse to renaming the triggers on the partitions; and
second, it failed to prohibit renaming clone triggers. Having triggers
with different names in partitions is pointless, and furthermore pg_dump
would not preserve names for partitions anyway.
Not backpatched -- making the ALTER TRIGGER throw an error in stable
versions might cause problems for existing scripts.
Co-authored-by: Arne Roland <A.Roland@index.de>
Co-authored-by: Álvaro Herrera <alvherre@alvh.no-ip.org>
Reviewed-by: Zhihong Yu <zyu@yugabyte.com>
Discussion: https://postgr.es/m/d0fd7040c2fb4de1a111b9d9ccc456b8@index.de
We had documentation of default_transaction_isolation et al,
but for some reason not of transaction_isolation et al.
AFAICS this is just an ancient oversight, so repair.
Per bug #17077 from Yanliang Lei.
Discussion: https://postgr.es/m/17077-ade8e166a01e1374@postgresql.org
To add support for streaming transactions at prepare time into the
built-in logical replication, we need to do the following things:
* Modify the output plugin (pgoutput) to implement the new two-phase API
callbacks, by leveraging the extended replication protocol.
* Modify the replication apply worker, to properly handle two-phase
transactions by replaying them on prepare.
* Add a new SUBSCRIPTION option "two_phase" to allow users to enable
two-phase transactions. We enable the two_phase once the initial data sync
is over.
We however must explicitly disable replication of two-phase transactions
during replication slot creation, even if the plugin supports it. We
don't need to replicate the changes accumulated during this phase,
and moreover, we don't have a replication connection open so we don't know
where to send the data anyway.
The streaming option is not allowed with this new two_phase option. This
can be done as a separate patch.
We don't allow to toggle two_phase option of a subscription because it can
lead to an inconsistent replica. For the same reason, we don't allow to
refresh the publication once the two_phase is enabled for a subscription
unless copy_data option is false.
Author: Peter Smith, Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich
Reviewed-by: Amit Kapila, Sawada Masahiko, Vignesh C, Dilip Kumar, Takamichi Osumi, Greg Nancarrow
Tested-By: Haiying Tang
Discussion: https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru
Discussion: https://postgr.es/m/CAA4eK1+opiV4aFTmWWUF9h_32=HfPOW9vZASHarT0UA5oBrtGw@mail.gmail.com
Allow a pager to be used by the \watch command. This works but isn't
very useful with traditional pagers like "less", so use a different
environment variable. The popular open source tool "pspg" (also by
Pavel) knows how to display the output if you set PSQL_WATCH_PAGER="pspg
--stream".
To make \watch react quickly when the user quits the pager or presses
^C, and also to increase the accuracy of its timing and decrease the
rate of useless context switches, change the main loop of the \watch
command to use sigwait() rather than a sleeping/polling loop, on Unix.
Supported on Unix only for now (like pspg).
Author: Pavel Stehule <pavel.stehule@gmail.com>
Author: Thomas Munro <thomas.munro@gmail.com>
Discussion: https://postgr.es/m/CAFj8pRBfzUUPz-3gN5oAzto9SDuRSq-TQPfXU_P6h0L7hO%2BEhg%40mail.gmail.com
Commit 4656e3d66 replaced the "#define CLOBBER_CACHE_ALWAYS"
testing mechanism with a GUC, which has been a great help for
doing cache-clobber testing in more efficient ways; but there
is a gap in the implementation. The only way to do cache-clobber
testing during an initdb run is to use the old method with #define,
because one can't set the GUC from outside. Improve this by
adding a switch to initdb for the purpose.
(Perhaps someday we should let initdb pass through arbitrary
"-c NAME=VALUE" switches. Quoting difficulties dissuaded me
from attempting that right now, though.)
Back-patch to v14 where 4656e3d66 came in.
Discussion: https://postgr.es/m/1582507.1624227029@sss.pgh.pa.us
Extend the replication command CREATE_REPLICATION_SLOT to support the
TWO_PHASE option. This will allow decoding commands like PREPARE
TRANSACTION, COMMIT PREPARED and ROLLBACK PREPARED for slots created with
this option. The decoding of the transaction happens at prepare command.
This patch also adds support of two-phase in pg_recvlogical via a new
option --two-phase.
This option will also be used by future patches that allow streaming of
transactions at prepare time for built-in logical replication. With this,
the out-of-core logical replication solutions can enable replication of
two-phase transactions via replication protocol.
Author: Ajin Cherian
Reviewed-By: Jeff Davis, Vignesh C, Amit Kapila
Discussion:
https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ruhttps://postgr.es/m/64b9f783c6e125f18f88fbc0c0234e34e71d8639.camel@j-davis.com
This commit prevents pg_checksums to do a rewrite of a block if it has
no need to, in the case where the computed checksum matches with what's
already stored in the block read. This is helpful to accelerate
successive runs of the tool when the previous ones got interrupted, for
example.
The number of blocks and files written is tracked and reported by the
tool once finished. Note that the final flush of the data folder
happens even if no blocks are written, as it could be possible that a
previous interrupted run got stopped while doing a flush.
Author: Greg Sabino Mullane
Reviewed-by: Paquier Michael, Julien Rouhaud
Discussion: https://postgr.es/m/CAKAnmmL+k6goxmVzQJB+0bAR0PN1sgo6GDUXJhyhUmVMze1QAw@mail.gmail.com
The syntax summaries for CREATE FUNCTION and allied commands
made it look like LEAKPROOF is an alternative to
IMMUTABLE/STABLE/VOLATILE, when of course it is an orthogonal
option. Improve that.
Per gripe from aazamrafeeque0. Thanks to David Johnston for
suggestions.
Discussion: https://postgr.es/m/162444349581.694.5818572718530259025@wrigleys.postgresql.org
