This warning was disabled in meson.build (warning 4273). If you
enable it, it looks like this:
../src/backend/utils/misc/ps_status.c(27): warning C4273: '__p__environ': inconsistent dll linkage
C:\Program Files (x86)\Windows Kits\10\include\10.0.22621.0\ucrt\stdlib.h(1158): note: see previous definition of '__p__environ'
The declaration in ps_status.c was:
#if !defined(WIN32) || defined(_MSC_VER)
extern char **environ;
#endif
The declaration in the OS header file is:
_DCRTIMP char*** __cdecl __p__environ (void);
#define _environ (*__p__environ())
So it is evident that this could be problematic.
The old declaration was required by the old MSVCRT library, but we
don't support that anymore with MSVC.
To fix, disable the re-declaration in ps_status.c, and also in some
other places that use the same code pattern but didn't trigger the
warning.
Then we can also re-enable the warning (delete the disablement in
meson.build).
Reviewed-by: Bryan Green <dbryan.green@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://www.postgresql.org/message-id/flat/bf060644-47ff-441b-97cf-c685d0827757@eisentraut.org
This patch introduces sequence synchronization. Sequences that are synced
will have 2 states:
- INIT (needs [re]synchronizing)
- READY (is already synchronized)
A new sequencesync worker is launched as needed to synchronize sequences.
A single sequencesync worker is responsible for synchronizing all
sequences. It begins by retrieving the list of sequences that are flagged
for synchronization, i.e., those in the INIT state. These sequences are
then processed in batches, allowing multiple entries to be synchronized
within a single transaction. The worker fetches the current sequence
values and page LSNs from the remote publisher, updates the corresponding
sequences on the local subscriber, and finally marks each sequence as
READY upon successful synchronization.
Sequence synchronization occurs in 3 places:
1) CREATE SUBSCRIPTION
- The command syntax remains unchanged.
- The subscriber retrieves sequences associated with publications.
- Published sequences are added to pg_subscription_rel with INIT
state.
- Initiate the sequencesync worker to synchronize all sequences.
2) ALTER SUBSCRIPTION ... REFRESH PUBLICATION
- The command syntax remains unchanged.
- Dropped published sequences are removed from pg_subscription_rel.
- Newly published sequences are added to pg_subscription_rel with INIT
state.
- Initiate the sequencesync worker to synchronize only newly added
sequences.
3) ALTER SUBSCRIPTION ... REFRESH SEQUENCES
- A new command introduced for PG19 by f0b3573c3a.
- All sequences in pg_subscription_rel are reset to INIT state.
- Initiate the sequencesync worker to synchronize all sequences.
- Unlike "ALTER SUBSCRIPTION ... REFRESH PUBLICATION" command,
addition and removal of missing sequences will not be done in this
case.
Author: Vignesh C <vignesh21@gmail.com>
Reviewed-by: shveta malik <shveta.malik@gmail.com>
Reviewed-by: Hou Zhijie <houzj.fnst@fujitsu.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Reviewed-by: Dilip Kumar <dilipbalaut@gmail.com>
Reviewed-by: Peter Smith <smithpb2250@gmail.com>
Reviewed-by: Nisha Moond <nisha.moond412@gmail.com>
Reviewed-by: Shlok Kyal <shlok.kyal.oss@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://postgr.es/m/CAA4eK1LC+KJiAkSrpE_NwvNdidw9F2os7GERUeSxSKv71gXysQ@mail.gmail.com
We have never had a SET syntax that allows setting a GUC_LIST_INPUT
parameter to be an empty list. A locution such as
SET search_path = '';
doesn't mean that; it means setting the GUC to contain a single item
that is an empty string. (For search_path the net effect is much the
same, because search_path ignores invalid schema names and '' must be
invalid.) This is confusing, not least because configuration-file
entries and the set_config() function can easily produce empty-list
values.
We considered making the empty-string syntax do this, but that would
foreclose ever allowing empty-string items to be valid in list GUCs.
While there isn't any obvious use-case for that today, it feels like
the kind of restriction that might hurt someday. Instead, let's
accept the forbidden-up-to-now value NULL and treat that as meaning an
empty list. (An objection to this could be "what if we someday want
to allow NULL as a GUC value?". That seems unlikely though, and even
if we did allow it for scalar GUCs, we could continue to treat it as
meaning an empty list for list GUCs.)
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Andrei Klychkov <andrew.a.klychkov@gmail.com>
Reviewed-by: Jim Jones <jim.jones@uni-muenster.de>
Discussion: https://postgr.es/m/CA+mfrmwsBmYsJayWjc8bJmicxc3phZcHHY=yW5aYe=P-1d_4bg@mail.gmail.com
The order in this list was previously pretty random and had grown
organically over time. This made it unnecessarily cumbersome to
maintain these lists, as there was no clear guidelines about where to
put new entries. Also, after the merger of the type-specific GUC
structs, the list still reflected the previous type-specific
super-order.
By using alphabetical order, the place for new entries becomes clear,
and often related entries will be listed close together.
This patch reorders the existing entries in guc_parameters.dat, and it
also augments the generation script to error if an entry is found at
the wrong place.
Note: The order is actually checked after lower-casing, to handle the
likes of "DateStyle".
Reviewed-by: John Naylor <johncnaylorls@gmail.com>
Reviewed-by: Álvaro Herrera <alvherre@kurilemu.de>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://www.postgresql.org/message-id/flat/8fdfb91e-60fb-44fa-8df6-f5dea47353c9@eisentraut.org
Instead of having five separate GUC structs, one for each type, with
the generic part contained in each of them, flip it around and have
one common struct, with the type-specific part has a subfield.
The very original GUC design had type-specific structs and
type-specific lists, and the membership in one of the lists defined
the type. But now the structs themselves know the type (from the
.vartype field), and they are all loaded into a common hash table at
run time, and so this original separation no longer makes sense. It
creates a bunch of inconsistencies in the code about whether the
type-specific or the generic struct is the primary struct, and a lot
of casting in between, which makes certain assumptions about the
struct layouts.
After the change, all these casts are gone and all the data is
accessed via normal field references. Also, various code is
simplified because only one kind of struct needs to be processed.
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://www.postgresql.org/message-id/flat/8fdfb91e-60fb-44fa-8df6-f5dea47353c9@eisentraut.org
The log output functionality of log_autovacuum_min_duration applies to
both VACUUM and ANALYZE, so it is not possible to separate the VACUUM
and ANALYZE log output thresholds. Logs are likely to be output only for
VACUUM and not for ANALYZE.
Therefore, we decided to separate the threshold for log output of VACUUM
by autovacuum (log_autovacuum_min_duration) and the threshold for log
output of ANALYZE by autovacuum (log_autoanalyze_min_duration).
Author: Shinya Kato <shinya11.kato@gmail.com>
Reviewed-by: Kasahara Tatsuhito <kasaharatt@oss.nttdata.com>
Discussion: https://www.postgresql.org/message-id/flat/CAOzEurQtfV4MxJiWT-XDnimEeZAY+rgzVSLe8YsyEKhZcajzSA@mail.gmail.com
An error context callback function might leak some memory into
ErrorContext, since those functions are run with ErrorContext as
current context. In the case where the elevel is ERROR, this is
no problem since the code level that catches the error should do
FlushErrorState to clean up, and that will reset ErrorContext.
However, if the elevel is less than ERROR then no such cleanup occurs.
In principle, repeated leaks while emitting log messages or client
notices could accumulate arbitrarily much leaked data, if no ERROR
occurs in the session.
To fix, let errfinish() perform an ErrorContext reset if it is
at the outermost error nesting level. (If it isn't, we'll delay
cleanup until the outermost nesting level is exited.)
The only actual leakage of this sort that I've been able to observe
within our regression tests was recently introduced by commit
f727b63e8. While it seems plausible that there are other such
leaks not reached in the regression tests, the lack of field
reports suggests that they're not a big problem. Accordingly,
I won't take the risk of back-patching this now. We can always
back-patch later if we get field reports of leaks.
Reported-by: Andres Freund <andres@anarazel.de>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/jngsjonyfscoont4tnwi2qoikatpd5hifsg373vmmjvugwiu6g@m6opxh7uisgd
Eager aggregation is a query optimization technique that partially
pushes aggregation past a join, and finalizes it once all the
relations are joined. Eager aggregation may reduce the number of
input rows to the join and thus could result in a better overall plan.
In the current planner architecture, the separation between the
scan/join planning phase and the post-scan/join phase means that
aggregation steps are not visible when constructing the join tree,
limiting the planner's ability to exploit aggregation-aware
optimizations. To implement eager aggregation, we collect information
about aggregate functions in the targetlist and HAVING clause, along
with grouping expressions from the GROUP BY clause, and store it in
the PlannerInfo node. During the scan/join planning phase, this
information is used to evaluate each base or join relation to
determine whether eager aggregation can be applied. If applicable, we
create a separate RelOptInfo, referred to as a grouped relation, to
represent the partially-aggregated version of the relation and
generate grouped paths for it.
Grouped relation paths can be generated in two ways. The first method
involves adding sorted and hashed partial aggregation paths on top of
the non-grouped paths. To limit planning time, we only consider the
cheapest or suitably-sorted non-grouped paths in this step.
Alternatively, grouped paths can be generated by joining a grouped
relation with a non-grouped relation. Joining two grouped relations
is currently not supported.
To further limit planning time, we currently adopt a strategy where
partial aggregation is pushed only to the lowest feasible level in the
join tree where it provides a significant reduction in row count.
This strategy also helps ensure that all grouped paths for the same
grouped relation produce the same set of rows, which is important to
support a fundamental assumption of the planner.
For the partial aggregation that is pushed down to a non-aggregated
relation, we need to consider all expressions from this relation that
are involved in upper join clauses and include them in the grouping
keys, using compatible operators. This is essential to ensure that an
aggregated row from the partial aggregation matches the other side of
the join if and only if each row in the partial group does. This
ensures that all rows within the same partial group share the same
"destiny", which is crucial for maintaining correctness.
One restriction is that we cannot push partial aggregation down to a
relation that is in the nullable side of an outer join, because the
NULL-extended rows produced by the outer join would not be available
when we perform the partial aggregation, while with a
non-eager-aggregation plan these rows are available for the top-level
aggregation. Pushing partial aggregation in this case may result in
the rows being grouped differently than expected, or produce incorrect
values from the aggregate functions.
If we have generated a grouped relation for the topmost join relation,
we finalize its paths at the end. The final paths will compete in the
usual way with paths built from regular planning.
The patch was originally proposed by Antonin Houska in 2017. This
commit reworks various important aspects and rewrites most of the
current code. However, the original patch and reviews were very
useful.
Author: Richard Guo <guofenglinux@gmail.com>
Author: Antonin Houska <ah@cybertec.at> (in an older version)
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Jian He <jian.universality@gmail.com>
Reviewed-by: Tender Wang <tndrwang@gmail.com>
Reviewed-by: Matheus Alcantara <matheusssilv97@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: David Rowley <dgrowleyml@gmail.com>
Reviewed-by: Tomas Vondra <tomas@vondra.me> (in an older version)
Reviewed-by: Andy Fan <zhihuifan1213@163.com> (in an older version)
Reviewed-by: Ashutosh Bapat <ashutosh.bapat.oss@gmail.com> (in an older version)
Discussion: https://postgr.es/m/CAMbWs48jzLrPt1J_00ZcPZXWUQKawQOFE8ROc-ADiYqsqrpBNw@mail.gmail.com
Without this, rebuilds can malfunction unless --enable-depend is used.
Historically we've expected that you can get away without
--enable-depend as long as you manually clean after changing *.h
files; the makefiles are supposed to handle other sorts of
dependencies. So add this one.
Follow-on to 635998965, so no need for back-patch.
Discussion: https://postgr.es/m/3121329.1758650878@sss.pgh.pa.us
This commit allows to log the raw parse tree in the same way we
currently log the parse tree, rewritten tree, and plan tree.
To avoid unnecessary log noise for users not interested in this
detail, a new GUC option, "debug_print_raw_parse", has been added.
When starting the PostgreSQL process with "-d N", and N is 3 or higher,
debug_print_raw_parse is enabled automatically, alongside
debug_print_parse.
Author: Chao Li <lic@highgo.com>
Reviewed-by: Tender Wang <tndrwang@gmail.com>
Reviewed-by: Tatsuo Ishii <ishii@postgresql.org>
Reviewed-by: John Naylor <johncnaylorls@gmail.com>
Discussion: https://postgr.es/m/CAEoWx2mcO0Gpo4vd8kPMAFWeJLSp0MeUUnaLdE1x0tSVd-VzUw%40mail.gmail.com
Commit 6359989654 had it so that the parameter "debug_discard_caches"
did not exist unless DISCARD_CACHES_ENABLED was defined (typically via
enabling asserts). This was a mistake, it did not correspond to the
prior setup. Several tests use this parameter, so they were now
failing if you did not have asserts enabled.
Store the information in guc_tables.c in a .dat file similar to the
catalog data in src/include/catalog/, and generate a part of
guc_tables.c from that. The goal is to make it easier to edit that
information, and to be able to make changes to the downstream data
structures more easily. (Essentially, those are the same reasons as
for the original adoption of the .dat format.)
Reviewed-by: John Naylor <johncnaylorls@gmail.com>
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Reviewed-by: David E. Wheeler <david@justatheory.com>
Discussion: https://www.postgresql.org/message-id/flat/dae6fe89-1e0c-4c3f-8d92-19d23374fb10%40eisentraut.org
These changes don't actually fix any leaks. They just make sure that
Valgrind will find pointers to data structures that remain allocated
at process exit, and thus not falsely complain about leaks. In
particular, we are trying to avoid situations where there is no
pointer to the beginning of an allocated block (except possibly
within the block itself, which Valgrind won't count).
* Because dynahash.c never frees hashtable storage except by deleting
the whole hashtable context, it doesn't bother to track the individual
blocks of elements allocated by element_alloc(). This results in
"possibly lost" complaints from Valgrind except when the first element
of each block is actively in use. (Otherwise it'll be on a freelist,
but very likely only reachable via "interior pointers" within element
blocks, which doesn't satisfy Valgrind.)
To fix, if we're building with USE_VALGRIND, expend an extra pointer's
worth of space in each element block so that we can chain them all
together from the HTAB header. Skip this in shared hashtables though:
Valgrind doesn't track those, and we'd need additional locking to make
it safe to manipulate a shared chain.
While here, update a comment obsoleted by 9c911ec06.
* Put the dlist_node fields of catctup and catclist structs first.
This ensures that the dlist pointers point to the starts of these
palloc blocks, and thus that Valgrind won't consider them
"possibly lost".
* The postmaster's PMChild structs and the autovac launcher's
avl_dbase structs also have the dlist_node-is-not-first problem,
but putting it first still wouldn't silence the warning because we
bulk-allocate those structs in an array, so that Valgrind sees a
single allocation. Commonly the first array element will be pointed
to only from some later element, so that the reference would be an
interior pointer even if it pointed to the array start. (This is the
same issue as for dynahash elements.) Since these are pretty simple
data structures, I don't feel too bad about faking out Valgrind by
just keeping a static pointer to the array start.
(This is all quite hacky, and it's not hard to imagine usages where
we'd need some other idea in order to have reasonable leak tracking of
structures that are only accessible via dlist_node lists. But these
changes seem to be enough to silence this class of leakage complaints
for the moment.)
* Free a couple of data structures manually near the end of an
autovacuum worker's run when USE_VALGRIND, and ensure that the final
vac_update_datfrozenxid() call is done in a non-permanent context.
This doesn't have any real effect on the process's total memory
consumption, since we're going to exit as soon as that last
transaction is done. But it does pacify Valgrind.
* Valgrind complains about the postmaster's socket-files and
lock-files lists being leaked, which we can silence by just
not nulling out the static pointers to them.
* Valgrind seems not to consider the global "environ" variable as
a valid root pointer; so when we allocate a new environment array,
it claims that data is leaked. To fix that, keep our own
statically-allocated copy of the pointer, similarly to the previous
item.
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/285483.1746756246@sss.pgh.pa.us
In the current system architecture, none of these are worth obsessing
over; most are once-per-process leaks. However, Valgrind complains
about all of them, and if we get to using threads rather than
processes for backend sessions, it will become more interesting to
avoid per-session leaks.
* Fix leaks in StartupXLOG() and ShutdownWalRecovery().
* Fix leakage of pq_mq_handle in a parallel worker.
While at it, move mq_putmessage's "Assert(pq_mq_handle != NULL)"
to someplace where it's not trivially useless.
* Fix leak in logicalrep_worker_detach().
* Don't leak the startup-packet buffer in ProcessStartupPacket().
* Fix leak in evtcache.c's DecodeTextArrayToBitmapset().
If the presented array is toasted, this neglected to free the
detoasted copy, which was then leaked into EventTriggerCacheContext.
* I'm distressed by the amount of code that BuildEventTriggerCache
is willing to run while switched into a long-lived cache context.
Although the detoasted array is the only leak that Valgrind reports,
let's tighten things up while we're here. (DecodeTextArrayToBitmapset
is still run in the cache context, so doing this doesn't remove the
need for the detoast fix. But it reduces the surface area for other
leaks.)
* load_domaintype_info() intentionally leaked some intermediate cruft
into the long-lived DomainConstraintCache's memory context, reasoning
that the amount of leakage will typically not be much so it's not
worth doing a copyObject() of the final tree to avoid that. But
Valgrind knows nothing of engineering tradeoffs and complains anyway.
On the whole, the copyObject doesn't cost that much and this is surely
not a performance-critical code path, so let's do it the clean way.
* MarkGUCPrefixReserved didn't bother to clean up removed placeholder
GUCs at all, which shows up as a leak in one regression test.
It seems appropriate for it to do as much cleanup as
define_custom_variable does when replacing placeholders, so factor
that code out into a helper function. define_custom_variable's logic
was one brick shy of a load too: it forgot to free the separate
allocation for the placeholder's name.
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Richard Guo <guofenglinux@gmail.com>
Discussion: https://postgr.es/m/285483.1746756246@sss.pgh.pa.us
Currently, ALTER DATABASE/ROLE/SYSTEM RESET [ALL] with an unknown
custom GUC with a prefix reserved by MarkGUCPrefixReserved() errors
(unless a superuser runs a RESET ALL variant). This is problematic
for cases such as an extension library upgrade that removes a GUC.
To fix, simply make sure the relevant code paths explicitly allow
it. Note that we require superuser or privileges on the parameter
to reset it. This is perhaps a bit more restrictive than is
necessary, but it's not clear whether further relaxing the
requirements is safe.
Oversight in commit 88103567cb. The ALTER SYSTEM fix is dependent
on commit 2d870b4aef, which first appeared in v17. Unfortunately,
back-patching that commit would introduce ABI breakage, and while
that breakage seems unlikely to bother anyone, it doesn't seem
worth the risk. Hence, the ALTER SYSTEM part of this commit is
omitted on v15 and v16.
Reported-by: Mert Alev <mert@futo.org>
Reviewed-by: Laurenz Albe <laurenz.albe@cybertec.at>
Discussion: https://postgr.es/m/18964-ba09dea8c98fccd6%40postgresql.org
Backpatch-through: 15
Previously, the check_hook functions for max_slot_wal_keep_size and
idle_replication_slot_timeout would incorrectly raise an ERROR for values
set in postgresql.conf during upgrade, even though those values were not
actively used in the upgrade process.
To prevent logical slot invalidation during upgrade, we used to set
special values for these GUCs. Now, instead of relying on those values, we
directly prevent WAL removal and logical slot invalidation caused by
max_slot_wal_keep_size and idle_replication_slot_timeout.
Note: PostgreSQL 17 does not include the idle_replication_slot_timeout
GUC, so related changes were not backported.
BUG #18979
Reported-by: jorsol <jorsol@gmail.com>
Author: Dilip Kumar <dilipbalaut@gmail.com>
Reviewed by: vignesh C <vignesh21@gmail.com>
Reviewed by: Alvaro Herrera <alvherre@alvh.no-ip.org>
Backpatch-through: 17, where it was introduced
Discussion: https://postgr.es/m/219561.1751826409@sss.pgh.pa.us
Discussion: https://postgr.es/m/18979-a1b7fdbb7cd181c6@postgresql.org
Previously, the idle_replication_slot_timeout parameter used minutes
as its unit, based on the assumption that values would typically exceed
one minute in production environments. However, this caused unexpected
behavior: specifying a value below 30 seconds would round down to 0,
effectively disabling the timeout. This could be surprising to users.
To allow finer-grained control and avoid such confusion, this commit changes
the unit of idle_replication_slot_timeout to seconds. Larger values can
still be specified easily using standard time suffixes, for example,
'24h' for 24 hours.
Back-patch to v18 where idle_replication_slot_timeout was added.
Reported-by: Gunnar Morling <gunnar.morling@googlemail.com>
Author: Fujii Masao <masao.fujii@gmail.com>
Reviewed-by: Laurenz Albe <laurenz.albe@cybertec.at>
Reviewed-by: David G. Johnston <david.g.johnston@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Reviewed-by: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CADGJaX_0+FTguWpNSpgVWYQP_7MhoO0D8=cp4XozSQgaZ40Odw@mail.gmail.com
Backpatch-through: 18
This routine has come as a useful piece to be able to know the list of
injection points currently attached in a system. One area would be to
use it in a set-returning function, or just let out-of-core code play
with it.
This hides the internals of the shared memory array lookup holding the
information about the injection points (point name, library and function
name), allocating the result in a palloc'd List consumable by the
caller.
Reviewed-by: Jeff Davis <pgsql@j-davis.com>
Reviewed-by: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Reviewed-by: Rahila Syed <rahilasyed90@gmail.com>
Discussion: https://postgr.es/m/Z_xYkA21KyLEHvWR@paquier.xyz
Discussion: https://postgr.es/m/aBG2rPwl3GE7m1-Q@paquier.xyz
Move oauth_validator_libraries in postgresql.conf.sample to be grouped
with the other CONN_AUTH_AUTH settings, rather than making up a new
ad-hoc category. This matches the internal categorization and also
how it is listed in the documentation.
The macros INJECTION_POINT() and INJECTION_POINT_CACHED() are extended
with an optional argument that can be passed down to the callback
attached when an injection point is run, giving to callbacks the
possibility to manipulate a stack state given by the caller. The
existing callbacks in modules injection_points and test_aio have their
declarations adjusted based on that.
da7226993f (core AIO infrastructure) and 93bc3d75d8 (test_aio) and
been relying on a set of workarounds where a static variable called
pgaio_inj_cur_handle is used as runtime argument in the injection point
callbacks used by the AIO tests, in combination with a TRY/CATCH block
to reset the argument value. The infrastructure introduced in this
commit will be reused for the AIO tests, simplifying them.
Reviewed-by: Greg Burd <greg@burd.me>
Discussion: https://postgr.es/m/Z_y9TtnXubvYAApS@paquier.xyz
10f6646847 intended to limit the value of io_combine_limit to the minimum of
io_combine_limit and io_max_combine_limit. To avoid issues with interdependent
GUCs, it introduced io_combine_limit_guc and set io_combine_limit in assign
hooks. That plan was thwarted by guc_tables.c accidentally still referencing
io_combine_limit, instead of io_combine_limit_guc. That lead to the GUC
machinery overriding the work done in the assign hooks, potentially leaving
io_combine_limit with a too high value.
The consequence of this bug was that when running with io_combine_limit >
io_combine_limit_guc the AIO machinery would not have reserved large enough
iovec and IO data arrays, with one IO's arrays overlapping with another IO's,
leading to total confusion.
To make such a problem easier to detect in the future, add assertions to
pgaio_io_set_handle_data_* checking the length is smaller than
io_max_combine_limit (not just PG_IOV_MAX).
It'd be nice to have a few tests for this, but it's not entirely obvious how
to do so portably.
As remarked upon by Tom, the GUC assignment hooks really shouldn't set the
underlying variable, that's the job of the GUC machinery. Change that as well.
Discussion: https://postgr.es/m/c5jyqnuwrpigd35qe7xdypxsisdjrdba5iw63mhcse4mzjogxo@qdjpv22z763f
It can be set to either COPY (the default) or CLONE if the system
supports it. CLONE causes callers of copydir(), currently CREATE
DATABASE ... STRATEGY=FILE_COPY and ALTER DATABASE ... SET TABLESPACE =
..., to use copy_file_range (Linux, FreeBSD) or copyfile (macOS) to copy
files instead of a read-write loop over the contents.
CLONE gives the kernel the opportunity to share block ranges on
copy-on-write file systems and push copying down to storage on others,
depending on configuration. On some systems CLONE can be used to clone
large databases quickly with CREATE DATABASE ... TEMPLATE=source
STRATEGY=FILE_COPY.
Other operating systems could be supported; patches welcome.
Co-authored-by: Nazir Bilal Yavuz <byavuz81@gmail.com>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Ranier Vilela <ranier.vf@gmail.com>
Discussion: https://postgr.es/m/CA%2BhUKGLM%2Bt%2BSwBU-cHeMUXJCOgBxSHLGZutV5zCwY4qrCcE02w%40mail.gmail.com
Add basic NUMA awareness routines, using a minimal src/port/pg_numa.c
portability wrapper and an optional build dependency, enabled by
--with-libnuma configure option. For now this is Linux-only, other
platforms may be supported later.
A built-in SQL function pg_numa_available() allows checking NUMA
support, i.e. that the server was built/linked with the NUMA library.
The main function introduced is pg_numa_query_pages(), which allows
determining the NUMA node for individual memory pages. Internally the
function uses move_pages(2) syscall, as it allows batching, and is more
efficient than get_mempolicy(2).
Author: Jakub Wartak <jakub.wartak@enterprisedb.com>
Co-authored-by: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Álvaro Herrera <alvherre@alvh.no-ip.org>
Reviewed-by: Tomas Vondra <tomas@vondra.me>
Discussion: https://postgr.es/m/CAKZiRmxh6KWo0aqRqvmcoaX2jUxZYb4kGp3N%3Dq1w%2BDiH-696Xw%40mail.gmail.com
This escape shows the numeric server IP address that the client
has connected to. Unix-socket connections will show "[local]".
Non-client processes (e.g. background processes) will show "[none]".
We expect that this option will be of interest to only a fairly
small number of users. Therefore the implementation is optimized
for the case where it's not used (that is, we don't do the string
conversion until we have to), and we've not added the field to
csvlog or jsonlog formats.
Author: Greg Sabino Mullane <htamfids@gmail.com>
Reviewed-by: Cary Huang <cary.huang@highgo.ca>
Reviewed-by: David Steele <david@pgmasters.net>
Reviewed-by: Jim Jones <jim.jones@uni-muenster.de>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAKAnmmK-U+UicE-qbNU23K--Q5XTLdM6bj+gbkZBZkjyjrd3Ow@mail.gmail.com
Previously effective_io_concurrency and maintenance_io_concurrency could not
be set above 0 on machines without fadvise support. AIO enables IO concurrency
without such support, via io_method=worker.
Currently only subsystems using the read stream API will take advantage of
this. Other users of maintenance_io_concurrency (like recovery prefetching)
which leverage OS advice directly will not benefit from this change. In those
cases, maintenance_io_concurrency will have no effect on I/O behavior.
Author: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Noah Misch <noah@leadboat.com>
Discussion: https://postgr.es/m/CAAKRu_atGgZePo=_g6T3cNtfMf0QxpvoUh5OUqa_cnPdhLd=gw@mail.gmail.com
check_createrole_self_grant and check_synchronized_standby_slots
were allocating memory on a LOG elevel without checking if the
allocation succeeded or not, which would have led to a segfault
on allocation failure.
On top of that, a number of callsites were using the ERROR level,
relying on erroring out rather than returning false to allow the
GUC machinery handle it gracefully. Other callsites used WARNING
instead of LOG. While neither being not wrong, this changes all
check_ functions do it consistently with LOG.
init_custom_variable gets a promoted elevel to FATAL to keep
the guc_malloc error handling in line with the rest of the
error handling in that function which already call FATAL. If
we encounter an OOM in this callsite there is no graceful
handling to be had, better to error out hard.
Backpatch the fix to check_createrole_self_grant down to v16
and the fix to check_synchronized_standby_slots down to v17
where they were introduced.
Author: Daniel Gustafsson <daniel@yesql.se>
Reported-by: Nikita <pm91.arapov@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Bug: #18845
Discussion: https://postgr.es/m/18845-582c6e10247377ec@postgresql.org
Backpatch-through: 16
Until now max_files_per_process=N limited each backend to open N files in
total (minus a safety factor), even if there were already more files opened in
postmaster and inherited by backends. Change max_files_per_process to control
how many additional files each process is allowed to open.
The main motivation for this is the patch to add io_method=io_uring, which
needs to open one file for each backend. Without this patch, even if
RLIMIT_NOFILE is high enough, postmaster will fail in set_max_safe_fds() if
started with a high max_connections. The cause of the failure is that, until
now, set_max_safe_fds() subtracted the already open files from
max_files_per_process.
Reviewed-by: Noah Misch <noah@leadboat.com>
Discussion: https://postgr.es/m/w6uiicyou7hzq47mbyejubtcyb2rngkkf45fk4q7inue5kfbeo@bbfad3qyubvs
Discussion: https://postgr.es/m/CAGECzQQh6VSy3KG4pN1d=h9J=D1rStFCMR+t7yh_Kwj-g87aLQ@mail.gmail.com
This commit introduces a new GUC option max_active_replication_origins
to control the maximum number of active replication
origins. Previously, this was controlled by
'max_replication_slots'. Having a separate GUC option provides better
flexibility for setting up subscribers, as they may not require
replication slots (for cascading replication) but always require
replication origins.
Author: Euler Taveira <euler@eulerto.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Reviewed-by: vignesh C <vignesh21@gmail.com>
Discussion: https://postgr.es/m/b81db436-8262-4575-b7c4-bc0c1551000b@app.fastmail.com
This new parameter works just like the storage parameter of the
same name: if set to true (which is the default), autovacuum and
VACUUM attempt to truncate any empty pages at the end of the table.
It is primarily intended to help users avoid locking issues on hot
standbys. The setting can be overridden with the storage parameter
or VACUUM's TRUNCATE option.
Since there's presently no way to determine whether a Boolean
storage parameter is explicitly set or has just picked up the
default value, this commit also introduces an isset_offset member
to relopt_parse_elt.
Suggested-by: Will Storey <will@summercat.com>
Author: Nathan Bossart <nathandbossart@gmail.com>
Co-authored-by: Gurjeet Singh <gurjeet@singh.im>
Reviewed-by: Laurenz Albe <laurenz.albe@cybertec.at>
Reviewed-by: Fujii Masao <masao.fujii@oss.nttdata.com>
Reviewed-by: Robert Treat <rob@xzilla.net>
Discussion: https://postgr.es/m/Z2DE4lDX4tHqNGZt%40dev.null
The new GUC extension_control_path specifies a path to look for
extension control files. The default value is $system, which looks in
the compiled-in location, as before.
The path search uses the same code and works in the same way as
dynamic_library_path.
Some use cases of this are: (1) testing extensions during package
builds, (2) installing extensions outside security-restricted
containers like Python.app (on macOS), (3) adding extensions to
PostgreSQL running in a Kubernetes environment using operators such as
CloudNativePG without having to rebuild the base image for each new
extension.
There is also a tweak in Makefile.global so that it is possible to
install extensions using PGXS into an different directory than the
default, using 'make install prefix=/else/where'. This previously
only worked when specifying the subdirectories, like 'make install
datadir=/else/where/share pkglibdir=/else/where/lib', for purely
implementation reasons. (Of course, without the path feature,
installing elsewhere was rarely useful.)
Author: Peter Eisentraut <peter@eisentraut.org>
Co-authored-by: Matheus Alcantara <matheusssilv97@gmail.com>
Reviewed-by: David E. Wheeler <david@justatheory.com>
Reviewed-by: Gabriele Bartolini <gabriele.bartolini@enterprisedb.com>
Reviewed-by: Marco Nenciarini <marco.nenciarini@enterprisedb.com>
Reviewed-by: Niccolò Fei <niccolo.fei@enterprisedb.com>
Discussion: https://www.postgresql.org/message-id/flat/E7C7BFFB-8857-48D4-A71F-88B359FADCFD@justatheory.com
The default of 128kB is unchanged, but the upper limit is changed from
32 blocks to 128 blocks, unless the operating system's IOV_MAX is too
low. Some other RDBMSes seem to cap their multi-block buffer pool I/O
around this number, and it seems useful to allow experimentation.
The concrete change is to our definition of PG_IOV_MAX, which provides
the maximum for io_combine_limit and io_max_combine_limit. It also
affects a couple of other places that work with arrays of struct iovec
or smaller objects on the stack, so we still don't want to use the
system IOV_MAX directly without a clamp: it is not under our control and
likely to be 1024. 128 seems acceptable for our current usage.
For Windows, we can't use real scatter/gather yet, so we continue to
define our own IOV_MAX value of 16 and emulate preadv()/pwritev() with
loops. Someone would need to research the trade-offs of raising that
number.
NB if trying to see this working: you might temporarily need to hack
BAS_BULKREAD to be bigger, since otherwise the obvious way of "a very
big SELECT" is limited by that for now.
Suggested-by: Tomas Vondra <tomas@vondra.me>
Discussion: https://postgr.es/m/CA%2BhUKG%2B2T9p-%2BzM6Eeou-RAJjTML6eit1qn26f9twznX59qtCA%40mail.gmail.com
The existing io_combine_limit can be changed by users. The new
io_max_combine_limit is fixed at server startup time, and functions as a
silent clamp on the user setting. That in itself is probably quite
useful, but the primary motivation is:
aio_init.c allocates shared memory for all asynchronous IOs including
some per-block data, and we didn't want to waste memory you'd never used
by assuming they could be up to PG_IOV_MAX. This commit already halves
the size of 'AioHandleIov' and 'AioHandleData'. A follow-up commit can
now expand PG_IOV_MAX without affecting that.
Since our GUC system doesn't support dependencies or cross-checks
between GUCs, the user-settable one now assigns a "raw" value to
io_combine_limit_guc, and the lower of io_combine_limit_guc and
io_max_combine_limit is maintained in io_combine_limit.
Reviewed-by: Andres Freund <andres@anarazel.de> (earlier version)
Discussion: https://postgr.es/m/CA%2BhUKG%2B2T9p-%2BzM6Eeou-RAJjTML6eit1qn26f9twznX59qtCA%40mail.gmail.com
