This commit extends the WAIT FOR LSN command with an optional MODE option in
the WITH clause that specifies which LSN type to wait for:
WAIT FOR LSN '<lsn>' [WITH (MODE '<mode>', ...)]
where mode can be:
- 'standby_replay' (default): Wait for WAL to be replayed to the specified
LSN,
- 'standby_write': Wait for WAL to be written (received) to the specified
LSN,
- 'standby_flush': Wait for WAL to be flushed to disk at the specified LSN,
- 'primary_flush': Wait for WAL to be flushed to disk on the primary server.
The default mode is 'standby_replay', matching the original behavior when MODE
is not specified. This follows the pattern used by COPY and EXPLAIN
commands, where options are specified as string values in the WITH clause.
Modes are explicitly named to distinguish between primary and standby
operations:
- Standby modes ('standby_replay', 'standby_write', 'standby_flush') can only
be used during recovery (on a standby server),
- Primary mode ('primary_flush') can only be used on a primary server.
The 'standby_write' and 'standby_flush' modes are useful for scenarios where
applications need to ensure WAL has been received or persisted on the standby
without necessarily waiting for replay to complete. The 'primary_flush' mode
allows waiting for WAL to be flushed on the primary server.
This commit also includes includes:
- Documentation updates for the new syntax and mode descriptions,
- Test coverage for all four modes, including error cases and concurrent
waiters,
- Wakeup logic in walreceiver for standby write/flush waiters,
- Wakeup logic in WAL writer for primary flush waiters.
Discussion: https://postgr.es/m/CABPTF7UiArgW-sXj9CNwRzUhYOQrevLzkYcgBydmX5oDes1sjg%40mail.gmail.com
Author: Xuneng Zhou <xunengzhou@gmail.com>
Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Reviewed-by: Alvaro Herrera <alvherre@kurilemu.de>
Add support for waiting on WAL write and flush LSNs in addition to the
existing replay LSN wait type. This provides the foundation for
extending the WAIT FOR command with MODE parameter.
Key changes are following.
- Add WAIT_LSN_TYPE_STANDBY_WRITE and WAIT_LSN_TYPE_STANDBY_FLUSH to
WaitLSNType.
- Add GetCurrentLSNForWaitType() to retrieve the current LSN for each wait
type.
- Add new wait events WAIT_EVENT_WAIT_FOR_WAL_WRITE and
WAIT_EVENT_WAIT_FOR_WAL_FLUSH for pg_stat_activity visibility.
- Update WaitForLSN() to use GetCurrentLSNForWaitType() internally.
Discussion: https://postgr.es/m/CABPTF7UiArgW-sXj9CNwRzUhYOQrevLzkYcgBydmX5oDes1sjg%40mail.gmail.com
Author: Xuneng Zhou <xunengzhou@gmail.com>
Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Reviewed-by: Alvaro Herrera <alvherre@kurilemu.de>
7d854bdc5b has removed two symbols from pg_config.h.in. This file is
automatically generated. The correct cleanup needs to be done in the
build scripts, instead. autoheader produces now a consistent
pg_config.h.in, without the symbols that were removed in the previous
commit.
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/1193764.1767573683@sss.pgh.pa.us
We must tell init about each role name we plan to connect as,
else SSPI auth fails. Similar to previous patches such as
da44d71e7.
Oversight in f3c9e341c, per buildfarm member drongo.
This fixes a poorly written integer comparison function which was
performing subtraction in an attempt to return a negative value when
a < b and a positive value when a > b, and 0 when the values were equal.
Unfortunately that didn't always work correctly due to two's complement
having the INT_MIN 1 further from zero than INT_MAX. This could result
in an overflow and cause the comparison function to return an incorrect
result, which would result in the binary search failing to find the
value being searched for.
This could cause poor selectivity estimates when the statistics stored
the value of INT_MAX (2147483647) and the value being searched for was
large enough to result in the binary search doing a comparison with that
INT_MAX value.
Author: Chao Li <li.evan.chao@gmail.com>
Reviewed-by: David Rowley <dgrowleyml@gmail.com>
Discussion: https://postgr.es/m/CAEoWx2ng1Ot5LoKbVU-Dh---dFTUZWJRH8wv2chBu29fnNDMaQ@mail.gmail.com
Backpatch-through: 14
Add a new "location" column to the pg_available_extensions and
pg_available_extension_versions views, exposing the directory where
the extension is located.
The default system location is shown as '$system', the same value
that can be used to configure the extension_control_path GUC.
User-defined locations are only visible for super users, otherwise
'<insufficient privilege>' is returned as a column value, the same
behaviour that we already use in pg_stat_activity.
I failed to resist the temptation to do a little extra editorializing of
the TAP test script.
Catalog version bumped.
Author: Matheus Alcantara <mths.dev@pm.me>
Reviewed-By: Chao Li <li.evan.chao@gmail.com>
Reviewed-By: Rohit Prasad <rohit.prasad@arm.com>
Reviewed-By: Michael Banck <mbanck@gmx.net>
Reviewed-By: Manni Wood <manni.wood@enterprisedb.com>
Reviewed-By: Euler Taveira <euler@eulerto.com>
Reviewed-By: Quan Zongliang <quanzongliang@yeah.net>
Commit f54af9f267 added a check for
io_uring_queue_init_mem(). However, it used the macro name
HAVE_LIBURING_QUEUE_INIT_MEM in both meson.build and the C code, while
the Autotools build script defined HAVE_IO_URING_QUEUE_INIT_MEM. As a
result, the optimization was never enabled in builds configured with
Autotools, as the C code checked for the wrong macro name.
This commit changes the macro name to HAVE_IO_URING_QUEUE_INIT_MEM in
meson.build and the C code. This matches the actual function
name (io_uring_queue_init_mem), following the standard HAVE_<FUNCTION>
convention.
Backpatch to 18, where the macro was introduced.
Bug: #19368
Reported-by: Evan Si <evsi@amazon.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/19368-016d79a7f3a1c599@postgresql.org
Backpatch-through: 18
This <note> was originally written to describe the double levels
of de-backslashing encountered when a backslash-aware string
literal is used to hold the text representation of a composite
value. It still made sense when we switched to mostly using E'...'
syntax for that type of literal. However, commit f77de4b0c mangled
it completely by changing the example literal to be SQL-standard.
The extra pass of de-backslashing described in the text doesn't
actually occur with the example as written, unless you happen to
be using standard_conforming_strings = off.
We could restore this <note> to self-consistency by reverting the
change from f77de4b0c, but on the whole I judge that its time has
passed. standard_conforming_strings = off is nearly obsolete,
and may soon be fully so. But without that, the behavior isn't
so complicated as to justify a discursive note. I observe that
the nearby section about array I/O syntax has no equivalent text,
although that syntax is equally subject to this issue.
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/2998401.1767038920@sss.pgh.pa.us
Discussion: https://postgr.es/m/3279216.1767072538@sss.pgh.pa.us
jit_profiling_support=true captures profile data for Linux perf. On
other platforms, LLVMCreatePerfJITEventListener() returns NULL and the
attempt to register the listener would crash.
Fix by ignoring the setting in that case. The documentation already
says that it only has an effect if perf support is present, and we
already did the same for older LLVM versions that lacked support.
No field reports, unsurprisingly for an obscure developer-oriented
setting. Noticed in passing while working on commit 1a28b4b4.
Backpatch-through: 14
Reviewed-by: Matheus Alcantara <matheusssilv97@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/CA%2BhUKGJgB6gvrdDohgwLfCwzVQm%3DVMtb9m0vzQn%3DCwWn-kwG9w%40mail.gmail.com
Up to now, index amhandlers were expected to produce a new, palloc'd
struct on each call. That requires palloc/pfree overhead, and creates
a risk of memory leaks if the caller fails to pfree, and the time
taken to fill such a large structure isn't nil. Moreover, we were
storing these things in the relcache, eating several hundred bytes for
each cached index. There is not anything in these structs that needs
to vary at runtime, so let's change the definition so that an
amhandler can return a pointer to a "static const" struct of which
there's only one copy per index AM. Mark all the core code's
IndexAmRoutine pointers const so that we catch anyplace that might
still try to change or pfree one.
(This is similar to the way we were already handling TableAmRoutine
structs. This commit does fix one comment that was infelicitously
copied-and-pasted into tableamapi.c.)
This commit needs to be called out in the v19 release notes as an API
change for extension index AMs. An un-updated AM will still work
(as of now, anyway) but it risks memory leaks and will be slower than
necessary.
Author: Matthias van de Meent <boekewurm+postgres@gmail.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAEoWx2=vApYk2LRu8R0DdahsPNEhWUxGBZ=rbZo1EXE=uA+opQ@mail.gmail.com
Previously, ReplicationSlotsComputeRequiredXmin() computed the oldest
xmin across all slots without holding ProcArrayLock (when
already_locked is false), acquiring the lock just before updating the
replication slot xmin.
This could lead to a race condition: if a backend created a new slot
and updates the global replication slot xmin, another backend
concurrently running ReplicationSlotsComputeRequiredXmin() could
overwrite that update with an invalid or stale value. This happens
because the concurrent backend might have computed the aggregate xmin
before the new slot was accounted for, but applied the update after
the new slot had already updated the global value.
In the reported failure, a walsender for an apply worker computed
InvalidTransactionId as the oldest xmin and overwrote a valid
replication slot xmin value computed by a walsender for a tablesync
worker. Consequently, the tablesync worker computed a transaction ID
via GetOldestSafeDecodingTransactionId() effectively without
considering the replication slot xmin. This led to the error "cannot
build an initial slot snapshot as oldest safe xid %u follows
snapshot's xmin %u", which was an assertion failure prior to commit
240e0dbacd.
To fix this, we acquire ReplicationSlotControlLock in exclusive mode
during slot creation to perform the initial update of the slot
xmin. In ReplicationSlotsComputeRequiredXmin(), we hold
ReplicationSlotControlLock in shared mode until the global slot xmin
is updated in ProcArraySetReplicationSlotXmin(). This prevents
concurrent computations and updates of the global xmin by other
backends during the initial slot xmin update process, while still
permitting concurrent calls to ReplicationSlotsComputeRequiredXmin().
Backpatch to all supported versions.
Author: Zhijie Hou <houzj.fnst@fujitsu.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Reviewed-by: Pradeep Kumar <spradeepkumar29@gmail.com>
Reviewed-by: Hayato Kuroda (Fujitsu) <kuroda.hayato@fujitsu.com>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://postgr.es/m/CAA4eK1L8wYcyTPxNzPGkhuO52WBGoOZbT0A73Le=ZUWYAYmdfw@mail.gmail.com
Backpatch-through: 14
We currently require LLVM 14, so these probes for LLVM 9 functions
always succeeded. Even when the features aren't enabled in an LLVM
build, dummy functions are defined (a problem for a later commit).
The whole PGAC_CHECK_LLVM_FUNCTIONS macro and Meson equivalent are
removed, because we switched to testing LLVM_VERSION_MAJOR at compile
time in subsequent work and these were the last holdouts. That suits
the nature of LLVM API evolution better, and also allows for strictly
mechanical pruning in future commits like 820b5af7 and 972c2cd2. They
advanced the minimum LLVM version but failed to spot these.
Reviewed-by: Matheus Alcantara <matheusssilv97@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/CA%2BhUKGJgB6gvrdDohgwLfCwzVQm%3DVMtb9m0vzQn%3DCwWn-kwG9w%40mail.gmail.com
This new function exposes at SQL level some information related to
multixacts, not available until now. This data is useful for monitoring
purposes, especially for workloads that make a heavy use of multixacts:
- num_mxids, number of MultiXact IDs in use.
- num_members, number of member entries in use.
- members_size, bytes used by num_members in pg_multixact/members/.
- oldest_multixact: oldest MultiXact still needed.
This patch has been originally proposed when MultiXactOffset was still
32 bits, to monitor wraparound. This part is not relevant anymore since
bd8d9c9bdf that has widen MultiXactOffset to 64 bits. The monitoring
of disk space usage for the members is still relevant.
Some tests are added to check this function, in the shape of one
isolation test with concurrent transactions that take a ROW SHARE lock,
and some SQL tests for pg_read_all_stats. Some documentation is added
to explain some patterns that can come from the information provided by
the function.
Bump catalog version.
Author: Naga Appani <nagnrik@gmail.com>
Reviewed-by: Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Reviewed-by: Atsushi Torikoshi <torikoshia@oss.nttdata.com>
Discussion: https://postgr.es/m/CA+QeY+AAsYK6WvBW4qYzHz4bahHycDAY_q5ECmHkEV_eB9ckzg@mail.gmail.com
This function calculates in bytes the storage taken between two
multixact offsets. This will be used in an upcoming patch, introduced
separately here as this piece can be useful on its own.
Author: Naga Appani <nagnrik@gmail.com>
Co-authored-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/aUyTvZMq2CLgNEB4@paquier.xyz
This routine returned a number of members as a MultiXactOffset,
calculated based on the difference between the next-to-be-assigned
offset and the oldest offset. However, this number is not actually an
offset but a number.
This type confusion comes from the original implementation of
MultiXactMemberFreezeThreshold(), in 53bb309d2d. The number of
members is now defined as a uint64, large enough for MultiXactOffset.
This change will be used in a follow-up patch.
Reviewed-by: Naga Appani <nagnrik@gmail.com>
Discussion: https://postgr.es/m/aUyTvZMq2CLgNEB4@paquier.xyz
REL_18_STABLE and master have commit ee485912, so they always use the
newer LLVM opaque pointer functions. Drop -Wno-deprecated-declarations
(commit a56e7b660) for code under jit/llvm in those branches, to catch
any new deprecation warnings that arrive in future version of LLVM.
Older branches continued to use functions marked deprecated in LLVM 14
and 15 (ie switched to the newer functions only for LLVM 16+), as a
precaution against unforeseen compatibility problems with bitcode
already shipped. In those branches, the comment about warning
suppression is updated to explain that situation better. In theory we
could suppress warnings only for LLVM 14 and 15 specifically, but that
isn't done here.
Backpatch-through: 14
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/1407185.1766682319%40sss.pgh.pa.us
estimate_hash_bucket_stats is defined to return zero to *mcv_freq if
it cannot obtain a value for the frequency of the most common value.
Its sole caller final_cost_hashjoin ignored this provision and would
blindly believe the zero value, resulting in computing zero for the
largest bucket size. In consequence, the safety check that intended
to prevent the largest bucket from exceeding get_hash_memory_limit()
was ineffective, allowing very silly plans to be chosen if statistics
were missing.
After fixing final_cost_hashjoin to disregard zero results for
mcv_freq, a second problem appeared: some cases that should use hash
joins failed to. This is because estimate_hash_bucket_stats was
unaware of the fact that ANALYZE won't store MCV statistics if it
doesn't find any multiply-occurring values. Thus the lack of an MCV
stats entry doesn't necessarily mean that we know nothing; we may
well know that the column is unique. The former coding returned zero
for *mcv_freq in this case, which was pretty close to correct, but now
final_cost_hashjoin doesn't believe it and disables the hash join.
So check to see if there is a HISTOGRAM stats entry; if so, ANALYZE
has in fact run for this column and must have found it to be unique.
In that case report the MCV frequency as 1 / rows, instead of claiming
ignorance.
Reporting a more accurate *mcv_freq in this case can also affect the
bucket-size skew adjustment further down in estimate_hash_bucket_stats,
causing hash-join cost estimates to change slightly. This affects
some plan choices in the core regression tests. The first diff in
join.out corresponds to a case where we have no stats and should not
risk a hash join, but the remaining changes are caused by producing
a better bucket-size estimate for unique join columns. Those are all
harmless changes so far as I can tell.
The existing behavior was introduced in commit 4867d7f62 in v11.
It appears from the commit log that disabling the bucket-size safety
check in the absence of statistics was intentional; but we've now seen
a case where the ensuing behavior is bad enough to make that seem like
a poor decision. In any case the lack of other problems with that
safety check after several years helps to justify enforcing it more
strictly. However, we won't risk back-patching this, in case any
applications are depending on the existing behavior.
Bug: #19363
Reported-by: Jinhui Lai <jinhui.lai@qq.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://postgr.es/m/2380165.1766871097@sss.pgh.pa.us
Discussion: https://postgr.es/m/19363-8dd32fc7600a1153@postgresql.org
This patch causes one postgres_fdw test case to revert to the plan
it used before aa86129e1, i.e., using a remote sort in preference to
local sort. That decision is actually a coin-flip because cost_sort()
will give the same answer on both sides, so that the plan choice comes
down to little more than roundoff error. In consequence, the test
output can change as a result of even minor changes in nearby costs,
as we saw in aa86129e1 (compare also b690e5fac and 4b14e1871).
b690e5fac's solution to stabilizing the adjacent test case was to
disable sorting locally, and here we extend that to the currently-
problematic case. Without this, the following patch would cause this
plan choice to change back in this same way, for even less apparent
reason.
Author: Tom Lane <tgl@sss.pgh.pa.us>
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://postgr.es/m/2551253.1766952956@sss.pgh.pa.us
Mkvcbuild.pm scrapes Makefile contents, but couldn't understand the
change made by commit bec2a0aa. Revealed by BF animal hamerkop in
branch REL_16_STABLE.
1. It used += instead of =, which didn't match the pattern that
Mkvcbuild.pm looks for. Drop the +.
2. Mkvcbuild.pm doesn't link PROGRAM executables with libpgport. Apply
a local workaround to REL_16_STABLE only (later branches dropped
Mkvcbuild.pm).
Backpatch-through: 16
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/175163.1766357334%40sss.pgh.pa.us
When looking up statistical data about an expression, we failed to
look through PlaceHolderVar nodes, treating them as opaque. This
could prevent us from matching an expression to base columns, index
expressions, or extended statistics, as examine_variable() relies on
strict structural matching.
As a result, queries involving PlaceHolderVar nodes often fell back to
default selectivity estimates, potentially leading to poor plan
choices.
This patch updates examine_variable() to strip PlaceHolderVars before
analysis. This is safe during estimation because PlaceHolderVars are
transparent for the purpose of statistics lookup: they do not alter
the value distribution of the underlying expression.
To minimize performance overhead on this hot path, a lightweight
walker first checks for the presence of PlaceHolderVars. The more
expensive mutator is invoked only when necessary.
There is one ensuing plan change in the regression tests, which is
expected and demonstrates the fix: the rowcount estimate becomes much
more accurate with this patch.
Back-patch to v18. Although this issue exists before that, changes in
this version made it common enough to notice. Given the lack of field
reports for older versions, I am not back-patching further.
Reported-by: Haowu Ge <gehaowu@bitmoe.com>
Author: Richard Guo <guofenglinux@gmail.com>
Discussion: https://postgr.es/m/62af586c-c270-44f3-9c5e-02c81d537e3d.gehaowu@bitmoe.com
Backpatch-through: 18
When pulling up a subquery, we may need to wrap its targetlist items
in PlaceHolderVars to enforce separate identity or as a result of
outer joins. However, this causes any upper-level WHERE clauses
referencing these outputs to contain PlaceHolderVars, which prevents
indxpath.c from recognizing that they could be matched to index
columns or index expressions, potentially affecting the planner's
ability to use indexes.
To fix, explicitly strip PlaceHolderVars from index operands. A
PlaceHolderVar appearing in a relation-scan-level expression is
effectively a no-op. Nevertheless, to play it safe, we strip only
PlaceHolderVars that are not marked nullable.
The stripping is performed recursively to handle cases where
PlaceHolderVars are nested or interleaved with other node types. To
minimize performance impact, we first use a lightweight walker to
check for the presence of strippable PlaceHolderVars. The expensive
mutator is invoked only if a candidate is found, avoiding unnecessary
memory allocation and tree copying in the common case where no
PlaceHolderVars are present.
Back-patch to v18. Although this issue exists before that, changes in
this version made it common enough to notice. Given the lack of field
reports for older versions, I am not back-patching further.
Reported-by: Haowu Ge <gehaowu@bitmoe.com>
Author: Richard Guo <guofenglinux@gmail.com>
Discussion: https://postgr.es/m/62af586c-c270-44f3-9c5e-02c81d537e3d.gehaowu@bitmoe.com
Backpatch-through: 18
Presumably, the C type MsgType was meant to hold the protocol message
type in the pre-version-3 era, but this was never fully developed even
then, and the name is pretty confusing nowadays. It has only one
vestigial use for cancel requests that we can get rid of. Since a
cancel request is indicated by a special protocol version number, we
can use the ProtocolVersion type, which MsgType was based on.
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/505e76cb-0ca2-4e22-ba0f-772b5dc3f230%40eisentraut.org
pg_stat_get_backend_activity() calls pgstat_clip_activity() to ensure
that the reported query string is correctly truncated when it finishes
with an incomplete multi-byte sequence. However, the result returned by
the function was not what pgstat_clip_activity() generated, but the
non-truncated, original, contents from PgBackendStatus.st_activity_raw.
Oversight in 54b6cd589a, so backpatch all the way down.
Author: Chao Li <li.evan.chao@gmail.com>
Discussion: https://postgr.es/m/CAEoWx2mDzwc48q2EK9tSXS6iJMJ35wvxNQnHX+rXjy5VgLvJQw@mail.gmail.com
Backpatch-through: 14
This change has the advantage of removing some weird type casts, caused
by offset calculations based on pgoff_t but saved as int (on older
branches we use off_t, which could be 4 or 8 bytes depending on the
environment). These are safe currently because capped by
MAX_PHYSICAL_FILESIZE, but we would run into problems when to make
MAX_PHYSICAL_FILESIZE larger or allow callers of these routines to use a
larger physical max size on demand.
While on it, this improves BufFileDumpBuffer() so as we do not use an
offset for "availbytes". It is not a file offset per-set, but a number
of available bytes.
This change should lead to no functional changes.
Author: Chao Li <li.evan.chao@gmail.com>
Discussion: https://postgr.es/m/aUStrqoOCDRFAq1M@paquier.xyz
