The finalfunc might return a read-write expanded object. If we
de-duplicate multiple call sites for the aggregate, any function(s)
receiving the aggregate result earlier could alter or destroy the
value that reaches the ones called later. This is a brown-paper-bag
bug in commit 42b746d4c, because we actually considered the need
for read-only-ness but failed to realize that it applied to the case
with a finalfunc as well as the case without.
Per report from Justin Pryzby. New error in HEAD,
no need for back-patch.
Discussion: https://postgr.es/m/ZDm5TuKsh3tzoEjz@telsasoft.com
Commit 3e310d837 taught isAssignmentIndirectionExpr() to look through
CoerceToDomain nodes. That's not sufficient, because since commit
04fe805a1 it's been possible for the planner to simplify
CoerceToDomain to RelabelType when the domain has no constraints
to enforce. So we need to look through RelabelType too.
Per bug #17897 from Alexander Lakhin. Although 3e310d837 was
back-patched to v11, it seems sufficient to apply this change
to v12 and later, since 04fe805a1 came in in v12.
Dmitry Dolgov
Discussion: https://postgr.es/m/17897-4216c546c3874044@postgresql.org
For some reason I had not implemented RBM_ZERO_AND_CLEANUP_LOCK support in
ExtendBufferedRelTo(), likely thinking it not being reachable. But it is
reachable, e.g. when replaying a WAL record for a page in a relation that
subsequently is truncated (likely only reachable when doing crash recovery or
PITR, not during ongoing streaming replication).
As now all of the RBM_* modes are supported, remove assertions checking mode.
As we had no test coverage for this scenario, add a new TAP test. There's
plenty more that ought to be tested in this area...
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Reported-by: Alexander Lakhin <exclusion@gmail.com>
Reviewed-by: Alexander Lakhin <exclusion@gmail.com>
Discussion: https://postgr.es/m/392271.1681238924%40sss.pgh.pa.us
Discussion: https://postgr.es/m/0b5eb82b-cb99-e0a4-b932-3dc60e2e3926@gmail.com
The partition pruning logic assumed that "b IS NOT true" was exactly the
same as "b IS FALSE". This is not the case when considering NULL values.
Fix this so we correctly include any partition which could hold NULL
values for the NOT case.
Additionally, this fixes a bug in the partition pruning code which handles
partitioned tables partitioned like ((NOT boolcol)). This is a seemingly
unlikely schema design, and it was untested and also broken.
Here we add tests for the ((NOT boolcol)) case and insert some actual data
into those tables and verify we do get the correct rows back when running
queries. I've also adjusted the existing boolpart tests to include some
data and verify we get the correct results too.
Both the bugs being fixed here could lead to incorrect query results with
fewer rows being returned than expected. No additional rows could have
been returned accidentally.
In passing, remove needless ternary expression. It's more simple just to
pass !is_not_clause to makeBoolConst(). It makes sense to do this so the
code is consistent with the bug fix in the "else if" condition just below.
David Kimura did submit a patch to fix the first of the issues here, but
that's not what's being committed here.
Reported-by: David Kimura
Reviewed-by: Richard Guo, David Kimura
Discussion: https://postgr.es/m/CAHnPFjQ5qxs6J_p+g8=ww7GQvfn71_JE+Tygj0S7RdRci1uwPw@mail.gmail.com
Backpatch-through: 11, all supported versions
Hash join tuples reuse the HOT status bit to indicate match status
during hash join execution. Correct reuse requires clearing the bit in
all tuples. Serial hash join and parallel multi-batch hash join do so
upon inserting the tuple into the hashtable. Single batch parallel hash
join and batch 0 of unexpected multi-batch hash joins forgot to do this.
It hadn't come up before because hashtable tuple match bits are only
used for right and full outer joins and parallel ROJ and FOJ were
unsupported. 11c2d6fdf5 introduced support for parallel ROJ/FOJ but
neglected to ensure the match bits were reset.
Author: Melanie Plageman <melanieplageman@gmail.com>
Reported-by: Richard Guo <guofenglinux@gmail.com>
Discussion: https://postgr.es/m/flat/CAMbWs48Nde1Mv%3DBJv6_vXmRKHMuHZm2Q_g4F6Z3_pn%2B3EV6BGQ%40mail.gmail.com
bea3d7e has redesigned the regexp engine so as all the allocations go
through palloc() with a dedicated memory context. hba.c had to cope
with the past memory management logic by going through all the HBA and
ident lines generated, then directly free all the regexps found in
AuthTokens to ensure that no leaks would happen. Such leaks could
happen for example in the postmaster after a SIGHUP, in the event of
an HBA and/or ident reload failure where all the new content parsed must
be discarded, including all the regexps that may have been compiled.
Now that regexps are palloc()'d in their own memory context,
MemoryContextDelete() is enough to ensure that all the compiled regexps
are properly gone. Simplifying this logic in hba.c has the effect to
only remove code. Most of it is new in v16, except the part for regexps
compiled in ident entries for the system username, so doing this cleanup
now rather than when v17 opens for business will reduce future diffs
with the upcoming REL_16_STABLE.
Some comments were incorrect since bea3d7e, now fixed to reflect the
reality.
Reviewed-by: Bertrand Drouvot, Álvaro Herrera
Discussion: https://postgr.es/m/ZDdJ289Ky2qEj4h+@paquier.xyz
This GUC was renamed to debug_parallel_query in 5352ca22e. That commit
added an entry into map_old_guc_names[] to allow the old name still to
work. That was done to allow a transition time where the buildfarm
configs could be swapped over to use debug_parallel_query instead. That
work is now complete.
Here we remove the old name with the intention of breaking any user code
which is using force_parallel_query. As mentioned in the commit message
for 5352ca22e, it appeared many users were misled into thinking that
setting this GUC was doing something useful for them to make queries run
more quickly.
Discussion: https://postgr.es/m/CAApHDvoR7EOz7Tvyzrd18FO-Dw2Cp4Uyq25TEWguK+XyCJtzOw@mail.gmail.com
Make sure that function declarations use names that exactly match the
corresponding names from function definitions in a few places. These
inconsistencies were all introduced relatively recently, after the code
base had parameter name mismatches fixed in bulk (see commits starting
with commits 4274dc22 and 035ce1fe).
pg_bsd_indent still has a couple of similar inconsistencies, which I
(pgeoghegan) have left untouched for now.
Like all earlier commits that cleaned up function parameter names, this
commit was written with help from clang-tidy.
WHen building with GSSAPI support, explicitly require MIT Kerberos and
check for gssapi_ext.h in configure.ac and meson.build. Also add
documentation explicitly stating that we now require MIT Kerberos when
building with GSSAPI support.
Reveiwed by: Johnathan Katz
Discussion: https://postgr.es/m/abcc73d0-acf7-6896-e0dc-f5bc12a61bb1@postgresql.org
This reverts commit 3d03b24c3 (Revert Add support for Kerberos
credential delegation) which was committed on the grounds of concern
about portability, but on further review and discussion, it's clear that
we are better off explicitly requiring MIT Kerberos as that appears to
be the only GSSAPI library currently that's under proper maintenance
and ongoing development. The API used for storing credentials was added
to MIT Kerberos over a decade ago while for the other libraries which
appear to be mainly based on Heimdal, which exists explicitly to be a
re-implementation of MIT Kerberos, the API never made it to a released
version (even though it was added to the Heimdal git repo over 5 years
ago..).
This post-feature-freeze change was approved by the RMT.
Discussion: https://postgr.es/m/ZDDO6jaESKaBgej0%40tamriel.snowman.net
We can't assert that we're the only process attached to a barrier after
BarrierArriveAndDetachExceptLast(). Although that'll be true almost
always, a late-starting parallel worker can attach very briefly (that
is, immediately detach after checking the phase) right at that moment.
BarrierArriveAndDetachExceptLast() already contains an assertion like
that, but it holds a spinlock preventing the race. This thinko caused a
one-off failure on build farm animal chimaera.
Diagnosed-by: Melanie Plageman <melanieplageman@gmail.com>
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/3590249.1680971629@sss.pgh.pa.us
This reverts commit e056c557ae and minor later fixes thereof.
There's a few problems in this new feature -- most notably regarding
pg_upgrade behavior, but others as well. This new feature is not in any
way critical on its own, so instead of scrambling to fix it we revert it
and try again in early 17 with these issues in mind.
Discussion: https://postgr.es/m/3801207.1681057430@sss.pgh.pa.us
Our policy since commit ab77a5a45 has been that a plan node having
any initplans is automatically not parallel-safe. (This could be
relaxed, but not today.) clean_up_removed_plan_level neglected
this, and could attach initplans to a parallel-safe child plan
node without clearing the plan's parallel-safe flag. That could
lead to "subplan was not initialized" errors at runtime, in case
an initplan referenced another one and only the referencing one
got transmitted to parallel workers.
The fix in clean_up_removed_plan_level is trivial enough.
materialize_finished_plan also moves initplans from one node
to another, but it's okay because it already copies the source
node's parallel_safe flag. The other place that does this kind
of thing is standard_planner's hack to inject a top-level Gather
when debug_parallel_query is active. But that's actually dead
code given that we're correctly enforcing the "initplans aren't
parallel safe" rule, so just replace it with an Assert that
there are no initplans.
Also improve some related comments.
Normally we'd add a regression test case for this sort of bug.
The mistake itself is already reached by existing tests, but there
is accidentally no visible problem. The only known test case that
creates an actual failure seems too indirect and fragile to justify
keeping it as a regression test (not least because it fails to fail
in v11, though the bug is clearly present there too).
Per report from Justin Pryzby. Back-patch to all supported branches.
Discussion: https://postgr.es/m/ZDVt6MaNWkRDO1LQ@telsasoft.com
Clarify the goals of the recently added guidelines for rmgrdesc authors:
to avoid gratuitous inconsistencies across resource managers, and to
make it reasonably easy to write a reusable custom parser.
Beyond that, the guidelines leave rmgrdesc authors with a significant
amount of leeway. This even includes the leeway to invent custom
conventions (in cases where it's warranted).
Follow-up to commit 7d8219a4.
Author: Peter Geoghegan <pg@bowt.ie>
Reviewed-By: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/CAH2-WzkbYuvwYKm-Y-72QEh6SPMQcAo9uONv+mR3bMGcu9E_Cg@mail.gmail.com
Make heap desc routines that output status bit as arrays of constants
avoid outputting array literals that contain superfluous punctuation
characters that complicate parsing the output. Also make sure that no
heap desc routine repeats the same key name (at the same nesting level),
for the same reason. Arguably, these were both oversights in commit
7d8219a4.
In passing, make the desc output code (which covers Heap's DELETE,
UPDATE, HOT_UPDATE, LOCK, and LOCK_UPDATED record types) consistent in
terms of the output order of each field. This order also matches WAL
record struct order. Heap's DELETE desc output now shows the record's
xmax field for the first time (just like UPDATE/HOT_UPDATE records).
Author: Peter Geoghegan <pg@bowt.ie>
Reviewed-By: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/CAH2-Wz=pNYtxiJ2Jx5Lj=fKo1OEZ4GE0p_kct+ugAUTqBwU46g@mail.gmail.com
Make xl_heap_lock's infobits_set field of type uint8, not int8. Using
int8 isn't appropriate given that the field just holds status bits.
This fixes an oversight in commit 0ac5ad5134.
In passing rename the nearby TransactionId field to "xmax" to make
things consistency with related records, such as xl_heap_lock_updated.
Deliberately avoid a bump in XLOG_PAGE_MAGIC. No backpatch, either.
Author: Peter Geoghegan <pg@bowt.ie>
Discussion: https://postgr.es/m/CAH2-WzkCd3kOS8b7Rfxw7Mh1_6jvX=Nzo-CWR1VBTiOtVZkWHA@mail.gmail.com
There's no need to check if opt->arg is NULL since defGetString() already
does that and raises an ERROR if it is. Let's just remove that check.
Also, combine the two remaining ERRORs into a single check. It seems
better to give an indication about what sort of values we're looking for
rather than just to state that the value given isn't valid. Make
BUFFER_USAGE_LIMIT uppercase in this ERROR message too. It's already
upper case in one other error message, so make that consistent.
Reported-by: Kyotaro Horiguchi
Discussion: https://postgr.es/m/20230411.102335.1643720544536884844.horikyota.ntt@gmail.com
We cannot use the generic array_desc approach with per-tuple nbtree
posting list update metadata because array_desc can only deal with fixed
width elements (e.g., page offset numbers). Using array_desc led to
incorrect rmgr descriptions for updates from nbtree DELETE/VACUUM WAL
records.
To fix, add specialized code to describe the update metadata as array
elements in desc output. We now iterate over the update metadata using
an approach that matches related REDO routines.
Also stop showing the updates offset number array separately in nbtree
DELETE/VACUUM desc output. It's redundant information, since the same
page offset numbers appear in the description of each individual update
element. Also make some small tweaks to the way that we format arrays
in all desc routines (not just nbtree desc routines) to make arrays a
little less verbose.
Oversight in commit 1c453cfd, which enhanced the nbtree rmgr desc
routines.
Author: Peter Geoghegan <pg@bowt.ie>
Discussion: https://postgr.es/m/CAH2-WzkbYuvwYKm-Y-72QEh6SPMQcAo9uONv+mR3bMGcu9E_Cg@mail.gmail.com
This reverts commit 3d4fa227bc.
Per discussion and buildfarm, this depends on APIs that seem to not
be available on at least one platform (NetBSD). Should be certainly
possible to rework to be optional on that platform if necessary but bit
late for that at this point.
Discussion: https://postgr.es/m/3286097.1680922218@sss.pgh.pa.us
Previously, a PostgreSQL-specific callback checked by the regex engine
had a way to trigger a special error code REG_CANCEL if it detected that
the next call to CHECK_FOR_INTERRUPTS() would certainly throw via
ereport().
A later proposed bugfix aims to move some complex logic out of signal
handlers, so that it won't run until the next CHECK_FOR_INTERRUPTS(),
which makes the above design impossible unless we split
CHECK_FOR_INTERRUPTS() into two phases, one to run logic and another to
ereport(). We may develop such a system in the future, but for the
regex code it is no longer necessary.
An earlier commit moved regex memory management over to our
MemoryContext system. Given that the purpose of the two-phase interrupt
checking was to free memory before throwing, something we don't need to
worry about anymore, it seems simpler to inject CHECK_FOR_INTERRUPTS()
directly into cancelation points, and just let it throw.
Since the plan is to keep PostgreSQL-specific concerns separate from the
main regex engine code (with a view to bein able to stay in sync with
other projects), do this with a new macro INTERRUPT(), customizable in
regcustom.h and defaulting to nothing.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CA%2BhUKGK3PGKwcKqzoosamn36YW-fsuTdOPPF1i_rtEO%3DnEYKSg%40mail.gmail.com
Now that our regex engine uses palloc(), it's not necessary to set up a
special memory context callback to free compiled regexes. The regex has
no resources other than the memory that is already going to be freed in
bulk.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CA%2BhUKGK3PGKwcKqzoosamn36YW-fsuTdOPPF1i_rtEO%3DnEYKSg%40mail.gmail.com
Previously, regex_t objects' memory was managed with malloc() and free()
directly. Switch to palloc()-based memory management instead.
Advantages:
* memory used by cached regexes is now visible with MemoryContext
observability tools
* cleanup can be done automatically in certain failure modes
(something that later commits will take advantage of)
* cleanup can be done in bulk
On the downside, there may be more fragmentation (wasted memory) due to
per-regex MemoryContext objects. This is a problem shared with other
cached objects in PostgreSQL and can probably be improved with later
tuning.
Thanks to Noah Misch for suggesting this general approach, which
unblocks later work on interrupts.
Suggested-by: Noah Misch <noah@leadboat.com>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CA%2BhUKGK3PGKwcKqzoosamn36YW-fsuTdOPPF1i_rtEO%3DnEYKSg%40mail.gmail.com
Unsurprisingly, this requires wal_level = logical to be set on the primary and
standby. The infrastructure added in 26669757b6 ensures that slots are
invalidated if the primary's wal_level is lowered.
Creating a slot on a standby waits for a xl_running_xact record to be
processed. If the primary is idle (and thus not emitting xl_running_xact
records), that can take a while. To make that faster, this commit also
introduces the pg_log_standby_snapshot() function. By executing it on the
primary, completion of slot creation on the standby can be accelerated.
Note that logical decoding on a standby does not itself enforce that required
catalog rows are not removed. The user has to use physical replication slots +
hot_standby_feedback or other measures to prevent that. If catalog rows
required for a slot are removed, the slot is invalidated.
See 6af1793954 for an overall design of logical decoding on a standby.
Bumps catversion, for the addition of the pg_log_standby_snapshot() function.
Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Author: Andres Freund <andres@anarazel.de> (in an older version)
Author: Amit Khandekar <amitdkhan.pg@gmail.com> (in an older version)
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: FabrÌzio de Royes Mello <fabriziomello@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Reviewed-By: Robert Haas <robertmhaas@gmail.com>
Physical walsenders can't send data until it's been flushed; logical
walsenders can't decode and send data until it's been applied. On the
standby, the WAL is flushed first, which will only wake up physical
walsenders; and then applied, which will only wake up logical
walsenders.
Previously, all walsenders were awakened when the WAL was flushed. That
was fine for logical walsenders on the primary; but on the standby the
flushed WAL would have been not applied yet, so logical walsenders were
awakened too early.
Per idea from Jeff Davis and Amit Kapila.
Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Reviewed-By: Jeff Davis <pgsql@j-davis.com>
Reviewed-By: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Discussion: https://postgr.es/m/CAA4eK1+zO5LUeisabX10c81LU-fWMKO4M9Wyg1cdkbW7Hqh6vQ@mail.gmail.com
During WAL replay on the standby, when a conflict with a logical slot is
identified, invalidate such slots. There are two sources of conflicts:
1) Using the information added in 6af1793954, logical slots are invalidated if
required rows are removed
2) wal_level on the primary server is reduced to below logical
Uses the infrastructure introduced in the prior commit. FIXME: add commit
reference.
Change InvalidatePossiblyObsoleteSlot() to use a recovery conflict to
interrupt use of a slot, if called in the startup process. The new recovery
conflict is added to pg_stat_database_conflicts, as confl_active_logicalslot.
See 6af1793954 for an overall design of logical decoding on a standby.
Bumps catversion for the addition of the pg_stat_database_conflicts column.
Bumps PGSTAT_FILE_FORMAT_ID for the same reason.
Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Author: Andres Freund <andres@anarazel.de>
Author: Amit Khandekar <amitdkhan.pg@gmail.com> (in an older version)
Reviewed-by: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Fabrízio de Royes Mello <fabriziomello@gmail.com>
Reviewed-by: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
Discussion: https://postgr.es/m/20230407075009.igg7be27ha2htkbt@awork3.anarazel.de
Needed for logical decoding on a standby. Slots need to be invalidated because
of the horizon if rows required for logical decoding are removed. If the
primary's wal_level is lowered from 'logical', logical slots on the standby
need to be invalidated.
The new invalidation methods will be used in a subsequent commit.
Logical slots that have been invalidated can be identified via the new
pg_replication_slots.conflicting column.
See 6af1793954 for an overall design of logical decoding on a standby.
Bumps catversion for the addition of the new pg_replication_slots column.
Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Author: Andres Freund <andres@anarazel.de>
Author: Amit Khandekar <amitdkhan.pg@gmail.com> (in an older version)
Reviewed-by: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Fabrízio de Royes Mello <fabriziomello@gmail.com>
Reviewed-by: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>
Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
Discussion: https://postgr.es/m/20230407075009.igg7be27ha2htkbt@awork3.anarazel.de
Previously we had checks for this in multiple places. Support for logical
decoding on standbys will add other forms of invalidation, making it worth
while to centralize the checks.
This slightly changes the error message for both the walsender and SQL
interface. Particularly the SQL interface error was inaccurate, as the "This
slot has never previously reserved WAL" portion was unreachable.
Reviewed-by: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/20230407075009.igg7be27ha2htkbt@awork3.anarazel.de
This is mainly useful because the upcoming logical-decoding-on-standby feature
adds further reasons for invalidating slots, and we don't want to end up with
multiple invalidated_* fields, or check different attributes.
Eventually we should consider not resetting restart_lsn when invalidating a
slot due to max_slot_wal_keep_size. But that's a user visible change, so left
for later.
Increases SLOT_VERSION, due to the changed field (with a different alignment,
no less).
Reviewed-by: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/20230407075009.igg7be27ha2htkbt@awork3.anarazel.de
Provide a way to ask the kernel to use O_DIRECT (or local equivalent)
where available for data and WAL files, to avoid or minimize kernel
caching. This hurts performance currently and is not intended for end
users yet. Later proposed work would introduce our own I/O clustering,
read-ahead, etc to replace the facilities the kernel disables with this
option.
The only user-visible change, if the developer-only GUC is not used, is
that this commit also removes the obscure logic that would activate
O_DIRECT for the WAL when wal_sync_method=open_[data]sync and
wal_level=minimal (which also requires max_wal_senders=0). Those are
non-default and unlikely settings, and this behavior wasn't (correctly)
documented. The same effect can be achieved with io_direct=wal.
Author: Thomas Munro <thomas.munro@gmail.com>
Author: Andres Freund <andres@anarazel.de>
Author: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>
Reviewed-by: Justin Pryzby <pryzby@telsasoft.com>
Reviewed-by: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>
Discussion: https://postgr.es/m/CA%2BhUKGK1X532hYqJ_MzFWt0n1zt8trz980D79WbjwnT-yYLZpg%40mail.gmail.com
In order to have the option to use O_DIRECT/FILE_FLAG_NO_BUFFERING in a
later commit, we need the addresses of user space buffers to be well
aligned. The exact requirements vary by OS and file system (typically
sectors and/or memory pages). The address alignment size is set to
4096, which is enough for currently known systems: it matches modern
sectors and common memory page size. There is no standard governing
O_DIRECT's requirements so we might eventually have to reconsider this
with more information from the field or future systems.
Aligning I/O buffers on memory pages is also known to improve regular
buffered I/O performance.
Three classes of I/O buffers for regular data pages are adjusted:
(1) Heap buffers are now allocated with the new palloc_aligned() or
MemoryContextAllocAligned() functions introduced by commit 439f6175.
(2) Stack buffers now use a new struct PGIOAlignedBlock to respect
PG_IO_ALIGN_SIZE, if possible with this compiler. (3) The buffer
pool is also aligned in shared memory.
WAL buffers were already aligned on XLOG_BLCKSZ. It's possible for
XLOG_BLCKSZ to be configured smaller than PG_IO_ALIGNED_SIZE and thus
for O_DIRECT WAL writes to fail to be well aligned, but that's a
pre-existing condition and will be addressed by a later commit.
BufFiles are not yet addressed (there's no current plan to use O_DIRECT
for those, but they could potentially get some incidental speedup even
in plain buffered I/O operations through better alignment).
If we can't align stack objects suitably using the compiler extensions
we know about, we disable the use of O_DIRECT by setting PG_O_DIRECT to
0. This avoids the need to consider systems that have O_DIRECT but
can't align stack objects the way we want; such systems could in theory
be supported with more work but we don't currently know of any such
machines, so it's easier to pretend there is no O_DIRECT support
instead. That's an existing and tested class of system.
Add assertions that all buffers passed into smgrread(), smgrwrite() and
smgrextend() are correctly aligned, unless PG_O_DIRECT is 0 (= stack
alignment tricks may be unavailable) or the block size has been set too
small to allow arrays of buffers to be all aligned.
Author: Thomas Munro <thomas.munro@gmail.com>
Author: Andres Freund <andres@anarazel.de>
Reviewed-by: Justin Pryzby <pryzby@telsasoft.com>
Discussion: https://postgr.es/m/CA+hUKGK1X532hYqJ_MzFWt0n1zt8trz980D79WbjwnT-yYLZpg@mail.gmail.com
Support GSSAPI/Kerberos credentials being delegated to the server by a
client. With this, a user authenticating to PostgreSQL using Kerberos
(GSSAPI) credentials can choose to delegate their credentials to the
PostgreSQL server (which can choose to accept them, or not), allowing
the server to then use those delegated credentials to connect to
another service, such as with postgres_fdw or dblink or theoretically
any other service which is able to be authenticated using Kerberos.
Both postgres_fdw and dblink are changed to allow non-superuser
password-less connections but only when GSSAPI credentials have been
delegated to the server by the client and GSSAPI is used to
authenticate to the remote system.
Authors: Stephen Frost, Peifeng Qiu
Reviewed-By: David Christensen
Discussion: https://postgr.es/m/CO1PR05MB8023CC2CB575E0FAAD7DF4F8A8E29@CO1PR05MB8023.namprd05.prod.outlook.com
a9c70b46db and 8aaa04b32S added counting of IO operations to a new view,
pg_stat_io. Now, add IO timing for reads, writes, extends, and fsyncs to
pg_stat_io as well.
This combines the tracking for pgBufferUsage with the tracking for pg_stat_io
into a new function pgstat_count_io_op_time(). This should make it a bit
easier to avoid the somewhat costly instr_time conversion done for
pgBufferUsage.
Author: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Discussion: https://postgr.es/m/flat/CAAKRu_ay5iKmnbXZ3DsauViF3eMxu4m1oNnJXqV_HyqYeg55Ww%40mail.gmail.com
Show a detailed description of the page offset number arrays that appear
in certain nbtree WAL records.
Also brings nbtree desc routines in line with the guidelines established
by recent commit 7d8219a4.
Author: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-By: Peter Geoghegan <pg@bowt.ie>
Discussion: https://postgr.es/m/flat/20230109215842.fktuhesvayno6o4g%40awork3.anarazel.de
Add helper functions that output arrays in a standard format, and use
the functions inside heapdesc routines. This allows tools like
pg_walinspect to show a detailed description of the page offset number
arrays for records like PRUNE and VACUUM (unless there was an FPI).
Also document the conventions that desc routines should follow. Only
the heapdesc routines follow the conventions for now, so they're just
guidelines for the time being.
Based on a suggestion from Andres Freund.
Author: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-By: Peter Geoghegan <pg@bowt.ie>
Discussion: https://postgr.es/m/flat/20230109215842.fktuhesvayno6o4g%40awork3.anarazel.de
Both pgstat_database and pgBufferUsage count IO timing for reads of temporary
relation blocks into local buffers. However, both failed to count write IO
timing for flushes of dirty local buffers. Fix.
Additionally, FlushRelationBuffers() seems to have omitted counting write
IO (both count and timing) stats for both pgstat_database and
pgBufferUsage. Fix.
Author: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/20230321023451.7rzy4kjj2iktrg2r%40awork3.anarazel.de
We now create pg_constaint rows for NOT NULL constraints with
contype='n'.
We propagate these constraints during operations such as adding
inheritance relationships, creating and attaching partitions, creating
tables LIKE other tables. We mostly follow the well-known rules of
conislocal and coninhcount that we have for CHECK constraints, with some
adaptations; for example, as opposed to CHECK constraints, we don't
match NOT NULL ones by name when descending a hierarchy to alter it;
instead we match by column number. This means we don't require the
constraint names to be identical across a hierarchy.
For now, we omit them from system catalogs. Maybe this is worth
reconsidering. We don't support NOT VALID nor DEFERRABLE clauses
either; these can be added as separate features later (this patch is
already large and complicated enough.)
This has been very long in the making. The first patch was written by
Bernd Helmle in 2010 to add a new pg_constraint.contype value ('n'),
which I (Álvaro) then hijacked in 2011 and 2012, until that one was
killed by the realization that we ought to use contype='c' instead:
manufactured CHECK constraints. However, later SQL standard
development, as well as nonobvious emergent properties of that design
(mostly, failure to distinguish them from "normal" CHECK constraints as
well as the performance implication of having to test the CHECK
expression) led us to reconsider this choice, so now the current
implementation uses contype='n' again.
In 2016 Vitaly Burovoy also worked on this feature[1] but found no
consensus for his proposed approach, which was claimed to be closer to
the letter of the standard, requiring additional pg_attribute columns to
track the OID of the NOT NULL constraint for that column.
[1] https://postgr.es/m/CAKOSWNkN6HSyatuys8xZxzRCR-KL1OkHS5-b9qd9bf1Rad3PLA@mail.gmail.com
Author: Álvaro Herrera <alvherre@alvh.no-ip.org>
Author: Bernd Helmle <mailings@oopsware.de>
Reviewed-by: Justin Pryzby <pryzby@telsasoft.com>
Reviewed-by: Peter Eisentraut <peter.eisentraut@enterprisedb.com>
Discussion: https://postgr.es/m/CACA0E642A0267EDA387AF2B%40%5B172.26.14.62%5D
Discussion: https://postgr.es/m/AANLkTinLXMOEMz+0J29tf1POokKi4XDkWJ6-DDR9BKgU@mail.gmail.com
Discussion: https://postgr.es/m/20110707213401.GA27098@alvh.no-ip.org
Discussion: https://postgr.es/m/1343682669-sup-2532@alvh.no-ip.org
Discussion: https://postgr.es/m/CAKOSWNkN6HSyatuys8xZxzRCR-KL1OkHS5-b9qd9bf1Rad3PLA@mail.gmail.com
Discussion: https://postgr.es/m/20220817181249.q7qvj3okywctra3c@alvherre.pgsql
The old wording described these as being multiplied by max_connections
plus max_prepared_transactions, which hasn't been exactly right for
some time thanks to the addition of various auxiliary processes.
Moreover, exactness here is a bit pointless given that the lock tables
can expand into the initially-unallocated "slop" space in shared
memory. Rather than trying to track exactly what the code is doing,
let's just use the term "server processes".
Likewise adjust these GUCs' description strings in guc_tables.c.
Wang Wei, reviewed by Nathan Bossart and myself
Discussion: https://postgr.es/m/OS3PR01MB6275BDD09C9B875C65FCC5AB9EA39@OS3PR01MB6275.jpnprd01.prod.outlook.com
The same condition accidentally was copied to both branches. Manual testing
confirms that otherwise the error recovery path works fine.
Found while reviewing the logical-decoding-on-standby patch.
This commit adds a limit to the size of an XLogRecord at 1020MB, based
on a suggestion by Heikki Linnakangas. This counts for the overhead
needed by the XLogReader when allocating the memory it needs to read a
record in DecodeXLogRecordRequiredSpace(), based on the record size. An
assertion based on that is added to detect that any additions in the
XLogReader facilities would not cause any overflows. If that's ever the
case, the upper bound allowed would need to be adjusted.
Before this, it was possible for an external module to create WAL
records large enough to be assembled but not replayable, causing
failures when replaying such WAL records on standbys. One case
mentioned where this is possible is the in-core function
pg_logical_emit_message() (wrapper for LogLogicalMessage), that allows
to emit WAL records with an arbitrary amount of data potentially higher
than the replay limit of approximately 1GB (limit of a palloc, minus the
overhead needed by a XLogReader).
This commit is a follow-up of ffd1b6b that has added similar protections
for the block-level data. Here, the checks are extended to the whole
record length, mainrdata_len being extended from uint32 to uint64 with
the routines registering buffer and record data still limited to uint32
to minimize the checks when assembling a record. All the error messages
related to overflow checks are improved to provide more context about
the error happening.
Author: Matthias van de Meent
Reviewed-by: Andres Freund, Heikki Linnakangas, Michael Paquier
Discussion: https://postgr.es/m/CAEze2WgGiw+LZt+vHf8tWqB_6VxeLsMeoAuod0N=ij1q17n5pw@mail.gmail.com
1cbbee033 added BUFFER_USAGE_LIMIT to the VACUUM and ANALYZE commands, so
here we permit that option to be specified in vacuumdb.
In passing, adjust the documents for vacuum_buffer_usage_limit and the
BUFFER_USAGE_LIMIT VACUUM option to mention "kB" rather than "KB". Do the
same for the ERROR message in ExecVacuum() and
check_vacuum_buffer_usage_limit(). Without that we might tell a user that
the valid minimum value is 128 KB only to reject that because we accept
only "kB" and not "KB".
Also, add a small reminder comment in vacuum.h to try to trigger the
memory of anyone adding new fields to VacuumParams that they might want to
consider if vacuumdb needs to grow a new option too.
Author: Melanie Plageman
Reviewed-by: Justin Pryzby
Reviewed-by: David Rowley
Discussion: https://postgr.es/m/ZAzTg3iEnubscvbf@telsasoft.com
While we already had some form of bulk extension for relations, it was fairly
limited. It only amortized the cost of acquiring the extension lock, the
relation itself was still extended one-by-one. Bulk extension was also solely
triggered by contention, not by the amount of data inserted.
To address this, use ExtendBufferedRelBy(), introduced in 31966b151e, to
extend the relation. We try to extend the relation by multiple blocks in two
situations:
1) The caller tells RelationGetBufferForTuple() that it will need multiple
pages. For now that's only used by heap_multi_insert(), see commit FIXME.
2) If there is contention on the extension lock, use the number of waiters for
the lock as a multiplier for the number of blocks to extend by. This is
similar to what we already did. Previously we additionally multiplied the
numbers of waiters by 20, but with the new relation extension
infrastructure I could not see a benefit in doing so.
Using the freespacemap to provide empty pages can cause significant
contention, and adds measurable overhead, even if there is no contention. To
reduce that, remember the blocks the relation was extended by in the
BulkInsertState, in the extending backend. In case 1) from above, the blocks
the extending backend needs are not entered into the FSM, as we know that we
will need those blocks.
One complication with using the FSM to record empty pages, is that we need to
insert blocks into the FSM, when we already hold a buffer content lock. To
avoid doing IO while holding a content lock, release the content lock before
recording free space. Currently that opens a small window in which another
backend could fill the block, if a concurrent VACUUM records the free
space. If that happens, we retry, similar to the already existing case when
otherBuffer is provided. In the future it might be worth closing the race by
preventing VACUUM from recording the space in newly extended pages.
This change provides very significant wins (3x at 16 clients, on my
workstation) for concurrent COPY into a single relation. Even single threaded
COPY is measurably faster, primarily due to not dirtying pages while
extending, if supported by the operating system (see commit 4d330a61bb). Even
single-row INSERTs benefit, although to a much smaller degree, as the relation
extension lock rarely is the primary bottleneck.
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/20221029025420.eplyow6k7tgu6he3@awork3.anarazel.de
Add new options to the VACUUM and ANALYZE commands called
BUFFER_USAGE_LIMIT to allow users more control over how large to make the
buffer access strategy that is used to limit the usage of buffers in
shared buffers. Larger rings can allow VACUUM to run more quickly but
have the drawback of VACUUM possibly evicting more buffers from shared
buffers that might be useful for other queries running on the database.
Here we also add a new GUC named vacuum_buffer_usage_limit which controls
how large to make the access strategy when it's not specified in the
VACUUM/ANALYZE command. This defaults to 256KB, which is the same size as
the access strategy was prior to this change. This setting also
controls how large to make the buffer access strategy for autovacuum.
Per idea by Andres Freund.
Author: Melanie Plageman
Reviewed-by: David Rowley
Reviewed-by: Andres Freund
Reviewed-by: Justin Pryzby
Reviewed-by: Bharath Rupireddy
Discussion: https://postgr.es/m/20230111182720.ejifsclfwymw2reb@awork3.anarazel.de
A future commit will use this information to determine how aggressively to
extend the relation by. In heap_multi_insert() we know accurately how many
pages we need once we need to extend the relation, providing an accurate lower
bound for how much to extend.
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/20221029025420.eplyow6k7tgu6he3@awork3.anarazel.de
Allow autovacuum to reload the config file more often so that cost-based
delay parameters can take effect while VACUUMing a relation. Previously,
autovacuum workers only reloaded the config file once per relation
vacuumed, so config changes could not take effect until beginning to
vacuum the next table.
Now, check if a reload is pending roughly once per block, when checking
if we need to delay.
In order for autovacuum workers to safely update their own cost delay
and cost limit parameters without impacting performance, we had to
rethink when and how these values were accessed.
Previously, an autovacuum worker's wi_cost_limit was set only at the
beginning of vacuuming a table, after reloading the config file.
Therefore, at the time that autovac_balance_cost() was called, workers
vacuuming tables with no cost-related storage parameters could still
have different values for their wi_cost_limit_base and wi_cost_delay.
Now that the cost parameters can be updated while vacuuming a table,
workers will (within some margin of error) have no reason to have
different values for cost limit and cost delay (in the absence of
cost-related storage parameters). This removes the rationale for keeping
cost limit and cost delay in shared memory. Balancing the cost limit
requires only the number of active autovacuum workers vacuuming a table
with no cost-based storage parameters.
Author: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Reviewed-by: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/CAAKRu_ZngzqnEODc7LmS1NH04Kt6Y9huSjz5pp7%2BDXhrjDA0gw%40mail.gmail.com
Vacuum code run both by autovacuum workers and a backend doing
VACUUM/ANALYZE previously inspected VacuumCostLimit and VacuumCostDelay,
which are the global variables backing the GUCs vacuum_cost_limit and
vacuum_cost_delay.
Autovacuum workers needed to override these variables with their
own values, derived from autovacuum_vacuum_cost_limit and
autovacuum_vacuum_cost_delay and worker cost limit balancing logic.
This led to confusing code which, in some cases, both derived and
set a new value of VacuumCostLimit from VacuumCostLimit.
In preparation for refreshing these GUC values more often, introduce
new, independent global variables and add a function to update them
using the GUCs and existing logic.
Per suggestion by Kyotaro Horiguchi
Author: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Reviewed-by: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/CAAKRu_ZngzqnEODc7LmS1NH04Kt6Y9huSjz5pp7%2BDXhrjDA0gw%40mail.gmail.com
While vacuuming a table in failsafe mode, VacuumCostActive should
not be re-enabled. This currently isn't a problem because vacuum
cost parameters are only refreshed in between vacuuming tables and
failsafe status is reset for every table.
In preparation for allowing vacuum cost parameters to be updated
more frequently, elevate LVRelState->failsafe_active to a global,
VacuumFailsafeActive, which will be checked when determining whether
or not to re-enable vacuum cost-related delays.
Author: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Reviewed-by: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/CAAKRu_ZngzqnEODc7LmS1NH04Kt6Y9huSjz5pp7%2BDXhrjDA0gw%40mail.gmail.com
tsquery's GETQUERY() macro is only safe to apply to a tsquery
that is known non-empty; otherwise it gives a pointer to garbage.
Before commit 5a617d75d, ts_headline() avoided this pitfall, but
only in a very indirect, nonobvious way. (hlCover could not reach
its TS_execute call, because if the query contains no lexemes
then hlFirstIndex would surely return -1.) After that commit,
it fell into the trap, resulting in weird errors such as
"unrecognized operator" and/or valgrind complaints. In HEAD,
fix this by not calling TS_execute_locations() at all for an
empty query. In the back branches, add a defensive check to
hlCover() --- that's not fixing any live bug, but I judge the
code a bit too fragile as-is.
Also, both mark_hl_fragments() and mark_hl_words() were careless
about the possibility of empty search text: in the cases where
no match has been found, they'd end up telling mark_fragment() to
mark from word indexes 0 to 0 inclusive, even when there is no
word 0. This is harmless since we over-allocated the prs->words
array, but it does annoy valgrind. Fix so that the end index is -1
and thus mark_fragment() will do nothing in such cases.
Bottom line is that this fixes a live bug in HEAD, but in the
back branches it's only getting rid of a valgrind nitpick.
Back-patch anyway.
Per report from Alexander Lakhin.
Discussion: https://postgr.es/m/c27f642d-020b-01ff-ae61-086af287c4fd@gmail.com
Starting with commit 7db0cd2145, RelationGetBufferForTuple() did a
visibilitymap_pin() while holding an exclusive buffer content lock on a newly
extended page, when using COPY FREEZE. We elsewhere try hard to avoid to doing
IO while holding a content lock. And until 14f98e0af9, that happened while
holding the relation extension lock.
Practically, this isn't a huge issue, because COPY FREEZE is restricted to
relations created or truncated in the current session, so it's unlikely
there's a lot of contention.
We can't avoid doing IO while holding the content lock by pinning the VM
earlier, because we don't know which page it will be on.
While we could just ignore the issue in this case, a future commit will add
bulk relation extension, which needs to enter pages into the FSM while also
trying to hold onto a buffer lock.
To address this issue, use visibilitymap_pin_ok() to see if the relevant
buffer is already pinned. If not, release the buffer, pin the VM buffer, and
acquire the lock again. This opens up a small window for other backends to
insert data onto the page - as the page is not entered into the freespacemap,
other backends won't see it normally, but a concurrent vacuum could enter the
page, if it started just after the relation is extended. In case the page is
used by another backend, retry. This is very similar to how locking
"otherBuffer" is already dealt with.
Reviewed-by: Tomas Vondra <tomas.vondra@enterprisedb.com>
Discussion: http://postgr.es/m/20230325025740.wzvchp2kromw4zqz@awork3.anarazel.de
GetVisibilityMapPins() insisted on the buffer1/buffer2 being in a specific
order. This required checks at the callsite. As a subsequent patch will add
another callsite, move related logic into GetVisibilityMapPins().
Discussion: https://postgr.es/m/20230403190030.fk2frxv6faklrseb@awork3.anarazel.de
zstd compression supports a special mode for finding matched in distant
past, which may result in better compression ratio, at the expense of
using more memory (the window size is 128MB).
To enable this optional mode, use the "long" keyword when specifying the
compression method (--compress=zstd:long).
Author: Justin Pryzby
Reviewed-by: Tomas Vondra, Jacob Champion
Discussion: https://postgr.es/m/20230224191840.GD1653@telsasoft.com
Discussion: https://postgr.es/m/20220327205020.GM28503@telsasoft.com
vacuum() is used for both the VACUUM command and for autovacuum. There
were many prechecks being done inside vacuum() that were just not relevant
to autovacuum. Let's move the bulk of these into ExecVacuum() so that
they're only executed when running the VACUUM command. This removes a
small amount of overhead when autovacuum vacuums a table.
While we are at it, allocate VACUUM's BufferAccessStrategy in ExecVacuum()
and pass it into vacuum() instead of expecting vacuum() to make it if it's
not already made by the calling function. To make this work, we need to
create the vacuum memory context slightly earlier, so we now need to pass
that down to vacuum() so that it's available for use in other memory
allocations.
Author: Melanie Plageman
Reviewed-by: David Rowley
Discussion: https://postgr.es/m/20230405211534.4skgskbilnxqrmxg@awork3.anarazel.de
A few places are not converted. Some because they are tackled in later
commits (e.g. hio.c, xlogutils.c), some because they are more
complicated (e.g. brin_pageops.c). Having a few users of ReadBuffer(P_NEW) is
good anyway, to ensure the backward compat path stays working.
Discussion: https://postgr.es/m/20221029025420.eplyow6k7tgu6he3@awork3.anarazel.de
32fbe0239 changed things so we didn't bother allocating the
BufferAccessStrategy during VACUUM (ONLY_DATABASE_STATS); and VACUUM
(FULL), however, it forgot to consider that VACUUM (FULL, ANALYZE) is a
possible combination. That change would have resulted in such a command
allowing ANALYZE to make full use of shared buffers, which wasn't
intended, so fix that.
Reported-by: Melanie Plageman
Discussion: https://postgr.es/m/CAAKRu_bJRKe+v_=OqwC+5sA3j5qv8rqdAwy3+yHaO3wmtfrCRg@mail.gmail.com
pg_stat_statements relies on EState->es_processed to count the number of
rows processed by ExecutorRun(). This proves to be a problem under the
extended query protocol when the result of a query is fetched through
more than one call of ExecutorRun(), as es_processed is reset each time
ExecutorRun() is called. This causes pg_stat_statements to report the
number of rows calculated in the last execute fetch, rather than the
global sum of all the rows processed.
As pquery.c tells, this is a problem when a portal does not use
holdStore. For example, DMLs with RETURNING would report a correct
tuple count as these do one execution cycle when the query is first
executed to fill in the portal's store with one ExecutorRun(), feeding
on the portal's store for each follow-up execute fetch depending on the
fetch size requested by the client.
The fix proposed for this issue is simple with the addition of an extra
counter in EState that's preserved across multiple ExecutorRun() calls,
incremented with the value calculated in es_processed. This approach is
not back-patchable, unfortunately.
Note that libpq does not currently give any way to control the fetch
size when using the extended v3 protocol, meaning that in-core testing
is not possible yet. This issue can be easily verified with the JDBC
driver, though, with *autocommit disabled*. Hence, having in-core tests
requires more features, left for future discussion:
- At least two new libpq routines splitting PQsendQueryGuts(), one for
the bind/describe and a second for a series of execute fetches with a
custom fetch size, likely in a fashion similar to what JDBC does.
- A psql meta-command for the execute phase. This part is not strictly
mandatory, still it could be handy.
Reported-by: Andrew Dunstan (original discovery by Simon Siggs)
Author: Sami Imseih
Reviewed-by: Tom Lane, Michael Paquier
Discussion: https://postgr.es/m/EBE6C507-9EB6-4142-9E4D-38B1673363A7@amazon.com
Discussion: https://postgr.es/m/c90890e7-9c89-c34f-d3c5-d5c763a34bd8@dunslane.net
The primary bottlenecks for relation extension are:
1) The extension lock is held while acquiring a victim buffer for the new
page. Acquiring a victim buffer can require writing out the old page
contents including possibly needing to flush WAL.
2) When extending via ReadBuffer() et al, we write a zero page during the
extension, and then later write out the actual page contents. This can
nearly double the write rate.
3) The existing bulk relation extension infrastructure in hio.c just amortized
the cost of acquiring the relation extension lock, but none of the other
costs.
Unfortunately 1) cannot currently be addressed in a central manner as the
callers to ReadBuffer() need to acquire the extension lock. To address that,
this this commit moves the responsibility for acquiring the extension lock
into bufmgr.c functions. That allows to acquire the relation extension lock
for just the required time. This will also allow us to improve relation
extension further, without changing callers.
The reason we write all-zeroes pages during relation extension is that we hope
to get ENOSPC errors earlier that way (largely works, except for CoW
filesystems). It is easier to handle out-of-space errors gracefully if the
page doesn't yet contain actual tuples. This commit addresses 2), by using the
recently introduced smgrzeroextend(), which extends the relation, without
dirtying the kernel page cache for all the extended pages.
To address 3), this commit introduces a function to extend a relation by
multiple blocks at a time.
There are three new exposed functions: ExtendBufferedRel() for extending the
relation by a single block, ExtendBufferedRelBy() to extend a relation by
multiple blocks at once, and ExtendBufferedRelTo() for extending a relation up
to a certain size.
To avoid duplicating code between ReadBuffer(P_NEW) and the new functions,
ReadBuffer(P_NEW) now implements relation extension with
ExtendBufferedRel(), using a flag to tell ExtendBufferedRel() that the
relation lock is already held.
Note that this commit does not yet lead to a meaningful performance or
scalability improvement - for that uses of ReadBuffer(P_NEW) will need to be
converted to ExtendBuffered*(), which will be done in subsequent commits.
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/20221029025420.eplyow6k7tgu6he3@awork3.anarazel.de
A future patch will add support for extending relations by multiple blocks at
once. To be concurrency safe, the buffers for those blocks need to be marked
as BM_IO_IN_PROGRESS. Until now we only had infrastructure for recovering from
an IO error for a single buffer. This commit extends that infrastructure to
multiple buffers by using the resource owner infrastructure.
This commit increases the size of the ResourceOwnerData struct, which appears
to have a just about measurable overhead in very extreme workloads. Medium
term we are planning to substantially shrink the size of
ResourceOwnerData. Short term the increase is small enough to not worry about
it for now.
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/20221029025420.eplyow6k7tgu6he3@awork3.anarazel.de
Discussion: https://postgr.es/m/20221029200025.w7bvlgvamjfo6z44@awork3.anarazel.de
Merge and hash joins can support antijoin with the non-nullable input
on the right, using very simple combinations of their existing logic
for right join and anti join. This gives the planner more freedom
about how to order the join. It's particularly useful for hash join,
since we may now have the option to hash the smaller table instead
of the larger.
Richard Guo, reviewed by Ronan Dunklau and myself
Discussion: https://postgr.es/m/CAMbWs48xh9hMzXzSy3VaPzGAz+fkxXXTUbCLohX1_L8THFRm2Q@mail.gmail.com
Previously we held buffer locks for two buffer mapping partitions at the same
time to change the identity of buffers. Particularly for extending relations
needing to hold the extension lock while acquiring a victim buffer is
painful.But it also creates a bottleneck for workloads that just involve
reads.
Now we instead first acquire a victim buffer and write it out, if
necessary. Then we remove that buffer from the old partition with just the old
partition's partition lock held and insert it into the new partition with just
that partition's lock held.
By separating out the victim buffer acquisition, future commits will be able
to change relation extensions to scale better.
On my workstation, a micro-benchmark exercising buffered reads strenuously and
under a lot of concurrency, sees a >2x improvement.
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/20221029025420.eplyow6k7tgu6he3@awork3.anarazel.de
Commit 47bb9db75 taught AcquireExecutorLocks to re-acquire locks
on views using data from their RTE_SUBQUERY replacements, but
it now seems like we should make AcquirePlannerLocks do the same.
In this way, if a view has been redefined, we will notice that
a bit earlier while checking validity of a cached plan and thereby
avoid some wasted work.
Report and patch by Amit Langote.
Discussion: https://postgr.es/m/CA+HiwqH0xZOQ+GQAdKeckY1R4NOeHdzhtfxkAMJLSchpapNk5w@mail.gmail.com
This adds a few more assertions against a buffer being local in places we
don't expect, and extracts the check for a buffer being pinned exactly once
from LockBufferForCleanup() into its own function. Later commits will use this
function.
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: http://postgr.es/m/419312fd-9255-078c-c3e3-f0525f911d7f@iki.fi
smgrzeroextend() uses FileFallocate() to efficiently extend files by multiple
blocks. When extending by a small number of blocks, use FileZero() instead, as
using posix_fallocate() for small numbers of blocks is inefficient for some
file systems / operating systems. FileZero() is also used as the fallback for
FileFallocate() on platforms / filesystems that don't support fallocate.
A big advantage of using posix_fallocate() is that it typically won't cause
dirty buffers in the kernel pagecache. So far the most common pattern in our
code is that we smgrextend() a page full of zeroes and put the corresponding
page into shared buffers, from where we later write out the actual contents of
the page. If the kernel, e.g. due to memory pressure or elapsed time, already
wrote back the all-zeroes page, this can lead to doubling the amount of writes
reaching storage.
There are no users of smgrzeroextend() as of this commit. That will follow in
future commits.
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Reviewed-by: David Rowley <dgrowleyml@gmail.com>
Reviewed-by: John Naylor <john.naylor@enterprisedb.com>
Discussion: https://postgr.es/m/20221029025420.eplyow6k7tgu6he3@awork3.anarazel.de
In v13 and v14, the ENABLE/DISABLE TRIGGER USER variant malfunctioned
on cloned triggers, failing to find the clones because it thought they
were system triggers. Other variants of ENABLE/DISABLE TRIGGER would
improperly apply a superuserness check. Fix by adjusting the is-it-
a-system-trigger check to match reality in those branches. (As far
as I can find, this is the only place that got it wrong.)
There's no such bug in v15/HEAD, because we revised the catalog
representation of system triggers to be what this code was expecting.
However, add the test case to these branches anyway, because this area
is visibly pretty fragile. Also remove an obsoleted comment.
The recent v15/HEAD commit 6949b921d fixed a nearby bug. I now see
that my commit message for that was inaccurate: the behavior of
recursing to clone triggers is older than v15, but it didn't apply
to the case in v13/v14 because in those branches parent partitioned
tables have no pg_trigger entries for foreign-key triggers. But add
the test case from that commit to v13/v14, just to show what is
happening there.
Per bug #17886 from DzmitryH.
Discussion: https://postgr.es/m/17886-5406d5d828aa4aa3@postgresql.org
The read path needs to be able to initialize pages anyway, as relation
extensions are not durable. By avoiding initializing pages, we can, in a
future patch, extend the relation by multiple blocks at once.
Using smgrextend() for {vm,fsm}_extend() is not a good idea in general, as at
least one page of the VM/FSM will be read immediately after, always causing a
cache miss, requiring us to read content we just wrote.
Discussion: https://postgr.es/m/20230301223515.pucbj7nb54n4i4nv@awork3.anarazel.de
This is mainly a light renumbering to match the sections in the
standard.
The comments for SQL_IMPLEMENTATION_INFO and SQL_SIZING are no longer
applicable because the required information has been moved into part
9.
fill_seq_fork_with_data(), used to initialize a new sequence relation, only
locked the buffer after calling PageInit(), even though PageInit() modifies
page contents.
This is unlikely to cause real-world issues, as the relation is exclusively
locked at that point, and the buffer not yet marked dirty, so other processes
should not access the buffer.
This issue looks to have been present since the introduction of sequences in
e8647c45d6.
Given the low risk, it does not seem worth backpatching the fix.
Discussion: https://postgr.es/m/20230404185501.wdkmo3k7bedlx6qk@awork3.anarazel.de
Convert to BCP47 language tags before storing in the catalog, except
during binary upgrade or when the locale comes from an existing
collation or template database.
The resulting language tags can vary slightly between ICU
versions. For instance, "@colBackwards=yes" is converted to
"und-u-kb-true" in older versions of ICU, and to the simpler (but
equivalent) "und-u-kb" in newer versions.
The process of canonicalizing to a language tag also understands more
input locale string formats than ucol_open(). For instance,
"fr_CA.UTF-8" is misinterpreted by ucol_open() and the region is
ignored; effectively treating it the same as the locale "fr" and
opening the wrong collator. Canonicalization properly interprets the
language and region, resulting in the language tag "fr-CA", which can
then be understood by ucol_open().
This commit fixes a problem in prior versions due to ucol_open()
misinterpreting locale strings as described above. For instance,
creating an ICU collation with locale "fr_CA.UTF-8" would store that
string directly in the catalog, which would later be passed to (and
misinterpreted by) ucol_open(). After this commit, the locale string
will be canonicalized to language tag "fr-CA" in the catalog, which
will be properly understood by ucol_open(). Because this fix affects
the resulting collator, we cannot change the locale string stored in
the catalog for existing databases or collations; otherwise we'd risk
corrupting indexes. Therefore, only canonicalize locales for
newly-created (not upgraded) collations/databases. For similar
reasons, do not backport.
Discussion: https://postgr.es/m/8c7af6820aed94dc7bc259d2aa7f9663518e6137.camel@j-davis.com
Reviewed-by: Peter Eisentraut
This option is normally false, but can be set to true to obtain
the legacy behavior where the subscription runs with the permissions
of the subscription owner rather than the permissions of the
table owner. The advantages of this mode are (1) it doesn't require
that the subscription owner have permission to SET ROLE to each
table owner and (2) since no role switching occurs, the
SECURITY_RESTRICTED_OPERATION restrictions do not apply.
On the downside, it allows any table owner to easily usurp
the privileges of the subscription owner - basically, to take
over their account. Because that's generally quite undesirable,
we don't make this mode the default, but we do make it available,
just in case the new behavior causes too many problems for someone.
Discussion: http://postgr.es/m/CA+TgmoZ-WEeG6Z14AfH7KhmpX2eFh+tZ0z+vf0=eMDdbda269g@mail.gmail.com
Up until now, logical replication actions have been performed as the
subscription owner, who will generally be a superuser. Commit
cec57b1a0f documented hazards
associated with that situation, namely, that any user who owns a
table on the subscriber side could assume the privileges of the
subscription owner by attaching a trigger, expression index, or
some other kind of executable code to it. As a remedy, it suggested
not creating configurations where users who are not fully trusted
own tables on the subscriber.
Although that will work, it basically precludes using logical
replication in the way that people typically want to use it,
namely, to replicate a database from one node to another
without necessarily having any restrictions on which database
users can own tables. So, instead, change logical replication to
execute INSERT, UPDATE, DELETE, and TRUNCATE operations as the
table owner when they are replicated.
Since this involves switching the active user frequently within
a session that is authenticated as the subscription user, also
impose SECURITY_RESTRICTED_OPERATION restrictions on logical
replication code. As an exception, if the table owner can SET
ROLE to the subscription owner, these restrictions have no
security value, so don't impose them in that case.
Subscription owners are now required to have the ability to
SET ROLE to every role that owns a table that the subscription
is replicating. If they don't, replication will fail. Superusers,
who normally own subscriptions, satisfy this property by default.
Non-superusers users who own subscriptions will need to be
granted the roles that own relevant tables.
Patch by me, reviewed (but not necessarily in its entirety) by
Jelte Fennema, Jeff Davis, and Noah Misch.
Discussion: http://postgr.es/m/CA+TgmoaSCkg9ww9oppPqqs+9RVqCexYCE6Aq=UsYPfnOoDeFkw@mail.gmail.com
- At the last minute and for no particularly good reason, I changed the
WITHOUT token to be marked especially for lookahead, from the one in
WITHOUT TIME to the one in WITHOUT UNIQUE. Study of upcoming patches
(where a new WITHOUT ARRAY WRAPPER clause is added) showed me that the
former was better, so put it back the way the original patch had it.
- update exprTypmod() for JsonConstructorExpr to return the typmod of
the RETURNING clause, as a comment there suggested. Perhaps it's
possible for this to make a difference with datetime types, but I
didn't try to build a test case.
- The nodeFuncs.c support code for new nodes was calling walker()
directly instead of the WALK() macro as introduced by commit 1c27d16e6e.
Modernize that. Also add exprLocation() support for a couple of nodes
that missed it. Lastly, reorder the code more sensibly.
The WITHOUT_LA -> WITHOUT change means that stored rules containing
either WITHOUT TIME ZONE or WITHOUT UNIQUE KEYS would change
representation. Therefore, bump catversion.
Discussion: https://postgr.es/m/20230329181708.e64g2tpy7jyufqkr@alvherre.pgsql
The trace point was using the relfileno / fork / block for the to-be-read-in
buffer. Some upcoming work would make that more expensive to provide. We still
have buffer-flush-start/done, which can serve most tracing needs that
buffer-write-dirty could serve.
Discussion: https://postgr.es/m/f5164e7a-eef6-8972-75a3-8ac622ed0c6e@iki.fi
Commit 61b313e4 made VACUUM pass down a heaprel to index AM bulkdelete
and vacuumcleanup routines. Although this was primarily intended as
preparation for logical decoding on standbys, it also made it easy to
correct an old deficiency in how we determine how to cleanup SP-GiST
redirect and placeholder tuples.
Pass the heaprel to GlobalVisTestFor() during cleanup of redirect and
placeholder tuples, rather than pessimistically passing NULL.
Author: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Discussion: https://postgr.es/m/02392033-f030-a3c8-c7d0-5c27eb529fec@gmail.com
Commit 61b313e4 made nbtree consistently pass down a heaprel to low
level routines like _bt_getbuf(). Although this was primarily intended
as preparation for logical decoding on standbys, it also made it easy to
correct an old deficiency in how nbtree VACUUM determines whether or not
it's now safe to recycle deleted pages.
Pass the heaprel to GlobalVisTestFor() in nbtree routines that deal with
recycle safety. nbtree now makes less pessimistic assumptions about
recycle safety within non-catalog relations. This enhancement
complements the recycling enhancement added by commit 9dd963ae25.
nbtree remains just as pessimistic as ever when it comes to recycle
safety within indexes on catalog relations. There is no fundamental
reason why we need to treat catalog relations differently, though. The
behavioral inconsistency is a consequence of the way that nbtree uses
nextXID values to implement what Lanin and Shasha call "the drain
technique". Note in particular that it has nothing to do with whether
or not index tuples might still be required for an older MVCC snapshot.
Author: Bertrand Drouvot <bertranddrouvot.pg@gmail.com>
Discussion: https://postgr.es/m/CAH2-WzkaiDxCje0yPuH=3Uh2p1V_2pFGY==xfbZoZu7Ax_NB8g@mail.gmail.com
In USE_VALGRIND builds, add code to print the text of the current query
to the valgrind log whenever the valgrind error count has increased
during the query. Valgrind will already have printed its report,
if the error is distinct from ones already seen, so that this works
out fairly well as a way of annotating the log.
Onur Tirtir and Tom Lane
Discussion: https://postgr.es/m/AM9PR83MB0498531E804DC8DF8CFF0E8FE9D09@AM9PR83MB0498.EURPRD83.prod.outlook.com
The new name better reflects what the field is - the size of the buffers[]
array. ring_size sounded more like it is in units of bytes.
An upcoming commit allows a BufferAccessStrategy of custom sizes, so it
seems relevant to improve this beforehand.
Author: Melanie Plageman
Reviewed-by: David Rowley
Discussion: https://postgr.es/m/CAAKRu_YefVbhg4rAxU2frYFdTap61UftH-kUNQZBvAs%2BYK81Zg%40mail.gmail.com
Traditionally, vacuum always makes use of a buffer access strategy 32
buffers in size. This means that running vacuums tend not to cause too
many shared buffers to become dirty, however, this can cause vacuums to
run much more slowly than they otherwise could as WAL flushes will occur
more frequently due to having to flush WAL out to the LSN of the dirty
page before that page can be written to disk.
When we are performing failsafe VACUUMs (as added in 1e55e7d17), we really
want to make the vacuum work go as quickly as possible, so here we disable
the buffer access strategy when entering failsafe mode while vacuuming a
relation.
Per idea and analyis from Andres Freund.
In passing, also include some changes I had intended for 32fbe0239.
Author: Melanie Plageman
Reviewed-by: Justin Pryzby, David Rowley
Discussion: https://postgr.es/m/20230111182720.ejifsclfwymw2reb%40awork3.anarazel.de
VACUUM FULL and VACUUM ONLY_DATABASE_STATS will not use the vacuum
strategy ring created in vacuum(), so don't waste effort making it in
those cases.
There are other conceivable cases where the buffer strategy also won't be
used, but those are probably less common and not worth troubling over too
much. For example VACUUM (PROCESS_MAIN false, PROCESS_TOAST false).
There are other cases too, but many of these are only discovered once
inside vacuum_rel().
Author: Melanie Plageman
Reviewed-by: David Rowley
Discussion: https://postgr.es/m/CAAKRu_ZLRuzkM3zKogiZAz2hUony37yLY4aaLb8fPf8fgqs5Og@mail.gmail.com
Using global variables because we don't want to pass these values around
as parameters does not really seem like a great idea, so let's remove
these two global variables and adjust a few functions to accept these
values as parameters instead.
This is part of a wider patch which intends to allow the size of the
buffer access strategy that vacuum uses to be adjusted.
Author: Melanie Plageman
Reviewed-by: Bharath Rupireddy
Discussion: https://postgr.es/m/CAAKRu_b1q_07uquUtAvLqTM%3DW9nzee7QbtzHwA4XdUo7KX_Cnw%40mail.gmail.com
This commit only implements one prerequisite part for allowing logical
decoding. The commit message contains an explanation of the overall design,
which later commits will refer back to.
Overall design:
1. We want to enable logical decoding on standbys, but replay of WAL
from the primary might remove data that is needed by logical decoding,
causing error(s) on the standby. To prevent those errors, a new replication
conflict scenario needs to be addressed (as much as hot standby does).
2. Our chosen strategy for dealing with this type of replication slot
is to invalidate logical slots for which needed data has been removed.
3. To do this we need the latestRemovedXid for each change, just as we
do for physical replication conflicts, but we also need to know
whether any particular change was to data that logical replication
might access. That way, during WAL replay, we know when there is a risk of
conflict and, if so, if there is a conflict.
4. We can't rely on the standby's relcache entries for this purpose in
any way, because the startup process can't access catalog contents.
5. Therefore every WAL record that potentially removes data from the
index or heap must carry a flag indicating whether or not it is one
that might be accessed during logical decoding.
Why do we need this for logical decoding on standby?
First, let's forget about logical decoding on standby and recall that
on a primary database, any catalog rows that may be needed by a logical
decoding replication slot are not removed.
This is done thanks to the catalog_xmin associated with the logical
replication slot.
But, with logical decoding on standby, in the following cases:
- hot_standby_feedback is off
- hot_standby_feedback is on but there is no a physical slot between
the primary and the standby. Then, hot_standby_feedback will work,
but only while the connection is alive (for example a node restart
would break it)
Then, the primary may delete system catalog rows that could be needed
by the logical decoding on the standby (as it does not know about the
catalog_xmin on the standby).
So, it’s mandatory to identify those rows and invalidate the slots
that may need them if any. Identifying those rows is the purpose of
this commit.
Implementation:
When a WAL replay on standby indicates that a catalog table tuple is
to be deleted by an xid that is greater than a logical slot's
catalog_xmin, then that means the slot's catalog_xmin conflicts with
the xid, and we need to handle the conflict. While subsequent commits
will do the actual conflict handling, this commit adds a new field
isCatalogRel in such WAL records (and a new bit set in the
xl_heap_visible flags field), that is true for catalog tables, so as to
arrange for conflict handling.
The affected WAL records are the ones that already contain the
snapshotConflictHorizon field, namely:
- gistxlogDelete
- gistxlogPageReuse
- xl_hash_vacuum_one_page
- xl_heap_prune
- xl_heap_freeze_page
- xl_heap_visible
- xl_btree_reuse_page
- xl_btree_delete
- spgxlogVacuumRedirect
Due to this new field being added, xl_hash_vacuum_one_page and
gistxlogDelete do now contain the offsets to be deleted as a
FLEXIBLE_ARRAY_MEMBER. This is needed to ensure correct alignment.
It's not needed on the others struct where isCatalogRel has
been added.
This commit just introduces the WAL format changes mentioned above. Handling
the actual conflicts will follow in future commits.
Bumps XLOG_PAGE_MAGIC as the several WAL records are changed.
Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Author: Andres Freund <andres@anarazel.de> (in an older version)
Author: Amit Khandekar <amitdkhan.pg@gmail.com> (in an older version)
Reviewed-by: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Reviewed-by: Fabrízio de Royes Mello <fabriziomello@gmail.com>
Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
This is done in preparation for logical decoding on standby, which needs to
include whether visibility affecting WAL records are about a (user) catalog
table. Which is only known for the table, not the indexes.
It's also nice to be able to pass the heap relation to GlobalVisTestFor() in
vacuumRedirectAndPlaceholder().
Author: "Drouvot, Bertrand" <bertranddrouvot.pg@gmail.com>
Discussion: https://postgr.es/m/21b700c3-eecf-2e05-a699-f8c78dd31ec7@gmail.com
If visibilitymap_set() is called with flags containing a higher bit than
VISIBILITYMAP_ALL_FROZEN, the state of neighboring pages is affected. While
there was an assertion that *some* valid bits were set, it did not check
that *only* valid bits were. Change that.
Discussion: https://postgr.es/m/20230331043300.gux3s5wzrapqi4oe@awork3.anarazel.de
PageInit() while holding the extension lock is unnecessary after 0d1fe9f74e
started to use RBM_ZERO_AND_LOCK - nobody can look at the new page before we
release the page lock. PageInit() zeroes the page, which isn't that cheap, so
deferring it until after the extension lock is released seems like a good idea.
Doing visibilitymap_pin() while holding the extension lock, introduced in
7db0cd2145, looks like an accident. Due to the restrictions on
HEAP_INSERT_FROZEN it's unlikely to be a performance issue, but it still seems
better to move it out. We also are doing the visibilitymap_pin() while
holding the buffer lock, which will be fixed in a separate commit.
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: http://postgr.es/m/419312fd-9255-078c-c3e3-f0525f911d7f@iki.fi
Tom Lane
David Rowley
Andres Freund
Thomas Munro
Michael Paquier
Peter Geoghegan
Stephen Frost
Alvaro Herrera
Peter Eisentraut
Daniel Gustafsson
Tomas Vondra
Robert Haas
Jeff Davis
Alexander Korotkov