Teach contrib/amcheck's bt_index_parent_check() function to take
advantage of the uniqueness property of heapkeyspace indexes in support
of a new verification option: non-pivot tuples (non-highkey tuples on
the leaf level) can optionally be re-found using a new search for each,
that starts from the root page. If a tuple cannot be re-found, report
that the index is corrupt.
The new "rootdescend" verification option is exhaustive, and can
therefore make a call to bt_index_parent_check() take a lot longer.
Re-finding tuples during verification is mostly intended as an option
for backend developers, since the corruption scenarios that it alone is
uniquely capable of detecting seem fairly far-fetched.
For example, "rootdescend" verification is much more likely to detect
corruption of the least significant byte of a key from a pivot tuple in
the root page of a B-Tree that already has at least three levels.
Typically, only a few tuples on a cousin leaf page are at risk of
"getting overlooked" by index scans in this scenario. The corrupt key
in the root page is only slightly corrupt: corrupt enough to give wrong
answers to some queries, and yet not corrupt enough to allow the problem
to be detected without verifying agreement between the leaf page and the
root page, skipping at least one internal page level. The existing
bt_index_parent_check() checks never cross more than a single level.
Author: Peter Geoghegan
Reviewed-By: Heikki Linnakangas
Discussion: https://postgr.es/m/CAH2-Wz=yTWnVu+HeHGKb2AGiADL9eprn-cKYAto4MkKOuiGtRQ@mail.gmail.com
Teach nbtree to give some consideration to how "distinguishing"
candidate leaf page split points are. This should not noticeably affect
the balance of free space within each half of the split, while still
making suffix truncation truncate away significantly more attributes on
average.
The logic for choosing a leaf split point now uses a fallback mode in
the case where the page is full of duplicates and it isn't possible to
find even a minimally distinguishing split point. When the page is full
of duplicates, the split should pack the left half very tightly, while
leaving the right half mostly empty. Our assumption is that logical
duplicates will almost always be inserted in ascending heap TID order
with v4 indexes. This strategy leaves most of the free space on the
half of the split that will likely be where future logical duplicates of
the same value need to be placed.
The number of cycles added is not very noticeable. This is important
because deciding on a split point takes place while at least one
exclusive buffer lock is held. We avoid using authoritative insertion
scankey comparisons to save cycles, unlike suffix truncation proper. We
use a faster binary comparison instead.
Note that even pg_upgrade'd v3 indexes make use of these optimizations.
Benchmarking has shown that even v3 indexes benefit, despite the fact
that suffix truncation will only truncate non-key attributes in INCLUDE
indexes. Grouping relatively similar tuples together is beneficial in
and of itself, since it reduces the number of leaf pages that must be
accessed by subsequent index scans.
Author: Peter Geoghegan
Reviewed-By: Heikki Linnakangas
Discussion: https://postgr.es/m/CAH2-WzmmoLNQOj9mAD78iQHfWLJDszHEDrAzGTUMG3mVh5xWPw@mail.gmail.com
Make nbtree treat all index tuples as having a heap TID attribute.
Index searches can distinguish duplicates by heap TID, since heap TID is
always guaranteed to be unique. This general approach has numerous
benefits for performance, and is prerequisite to teaching VACUUM to
perform "retail index tuple deletion".
Naively adding a new attribute to every pivot tuple has unacceptable
overhead (it bloats internal pages), so suffix truncation of pivot
tuples is added. This will usually truncate away the "extra" heap TID
attribute from pivot tuples during a leaf page split, and may also
truncate away additional user attributes. This can increase fan-out,
especially in a multi-column index. Truncation can only occur at the
attribute granularity, which isn't particularly effective, but works
well enough for now. A future patch may add support for truncating
"within" text attributes by generating truncated key values using new
opclass infrastructure.
Only new indexes (BTREE_VERSION 4 indexes) will have insertions that
treat heap TID as a tiebreaker attribute, or will have pivot tuples
undergo suffix truncation during a leaf page split (on-disk
compatibility with versions 2 and 3 is preserved). Upgrades to version
4 cannot be performed on-the-fly, unlike upgrades from version 2 to
version 3. contrib/amcheck continues to work with version 2 and 3
indexes, while also enforcing stricter invariants when verifying version
4 indexes. These stricter invariants are the same invariants described
by "3.1.12 Sequencing" from the Lehman and Yao paper.
A later patch will enhance the logic used by nbtree to pick a split
point. This patch is likely to negatively impact performance without
smarter choices around the precise point to split leaf pages at. Making
these two mostly-distinct sets of enhancements into distinct commits
seems like it might clarify their design, even though neither commit is
particularly useful on its own.
The maximum allowed size of new tuples is reduced by an amount equal to
the space required to store an extra MAXALIGN()'d TID in a new high key
during leaf page splits. The user-facing definition of the "1/3 of a
page" restriction is already imprecise, and so does not need to be
revised. However, there should be a compatibility note in the v12
release notes.
Author: Peter Geoghegan
Reviewed-By: Heikki Linnakangas, Alexander Korotkov
Discussion: https://postgr.es/m/CAH2-WzkVb0Kom=R+88fDFb=JSxZMFvbHVC6Mn9LJ2n=X=kS-Uw@mail.gmail.com
Use dedicated struct to represent nbtree insertion scan keys. Having a
dedicated struct makes the difference between search type scankeys and
insertion scankeys a lot clearer, and simplifies the signature of
several related functions. This is based on a suggestion by Andrey
Lepikhov.
Streamline how unique index insertions cache binary search progress.
Cache the state of in-progress binary searches within _bt_check_unique()
for later instead of having callers avoid repeating the binary search in
an ad-hoc manner. This makes it easy to add a new optimization:
_bt_check_unique() now falls out of its loop immediately in the common
case where it's already clear that there couldn't possibly be a
duplicate.
The new _bt_check_unique() scheme makes it a lot easier to manage cached
binary search effort afterwards, from within _bt_findinsertloc(). This
is needed for the upcoming patch to make nbtree tuples unique by
treating heap TID as a final tiebreaker column. Unique key binary
searches need to restore lower and upper bounds. They cannot simply
continue to use the >= lower bound as the offset to insert at, because
the heap TID tiebreaker column must be used in comparisons for the
restored binary search (unlike the original _bt_check_unique() binary
search, where scankey's heap TID column must be omitted).
Author: Peter Geoghegan, Heikki Linnakangas
Reviewed-By: Heikki Linnakangas, Andrey Lepikhov
Discussion: https://postgr.es/m/CAH2-WzmE6AhUdk9NdWBf4K3HjWXZBX3+umC7mH7+WDrKcRtsOw@mail.gmail.com
Jsonpath grammar and scanner are both quite small. It doesn't worth complexity
to compile them separately. This commit makes grammar and scanner be compiled
at once. Therefore, jsonpath_gram.h and jsonpath_gram.h are no longer needed.
This commit also does some reorganization of code in jsonpath_gram.y.
Discussion: https://postgr.es/m/d47b2023-3ecb-5f04-d253-d557547cf74f%402ndQuadrant.com
There are 2-arguments and 4-arguments versions of jsonb_path_match() and
jsonb_path_exists(). But 4-arguments versions have optional 3rd and 4th
arguments, that leads to ambiguity. In the same time 2-arguments versions are
needed only for @@ and @? operators. So, rename 2-arguments versions to
remove the ambiguity.
Catversion is bumped.
Originally, if libpq got a failure (e.g., ECONNRESET) while trying to
send data to the server, it would just report that and wash its hands
of the matter. It was soon found that that wasn't a very pleasant way
of coping with server-initiated disconnections, so we introduced a hack
(pqHandleSendFailure) in the code that sends queries to make it peek
ahead for server error reports before reporting the send failure.
It now emerges that related cases can occur during connection setup;
in particular, as of TLS 1.3 it's unsafe to assume that SSL connection
failures will be reported by SSL_connect rather than during our first
send attempt. We could have fixed that in a hacky way by applying
pqHandleSendFailure after a startup packet send failure, but
(a) pqHandleSendFailure explicitly disclaims suitability for use in any
state except query startup, and (b) the problem still potentially exists
for other send attempts in libpq.
Instead, let's fix this in a more general fashion by eliminating
pqHandleSendFailure altogether, and instead arranging to postpone
all reports of send failures in libpq until after we've made an
attempt to read and process server messages. The send failure won't
be reported at all if we find a server message or detect input EOF.
(Note: this removes one of the reasons why libpq typically overwrites,
rather than appending to, conn->errorMessage: pqHandleSendFailure needed
that behavior so that the send failure report would be replaced if we
got a server message or read failure report. Eventually I'd like to get
rid of that overwrite behavior altogether, but today is not that day.
For the moment, pqSendSome is assuming that its callees will overwrite
not append to conn->errorMessage.)
Possibly this change should get back-patched someday; but it needs
testing first, so let's not consider that till after v12 beta.
Discussion: https://postgr.es/m/CAEepm=2n6Nv+5tFfe8YnkUm1fXgvxR0Mm1FoD+QKG-vLNGLyKg@mail.gmail.com
Commit 6f6a6d8b1 introduced a delay of up to 2 seconds if we're trying
to request a checkpoint but the checkpointer hasn't started yet (or,
much less likely, our kill() call fails). However buildfarm experience
shows that that's not quite enough for slow or heavily-loaded machines.
There's no good reason to assume that the checkpointer won't start
eventually, so we may as well make the timeout much longer, say 60 sec.
However, if the caller didn't say CHECKPOINT_WAIT, it seems like a bad
idea to be waiting at all, much less for as long as 60 sec. We can
remove the need for that, and make this whole thing more robust, by
adjusting the code so that the existence of a pending checkpoint
request is clear from the contents of shared memory, and making sure
that the checkpointer process will notice it at startup even if it did
not get a signal. In this way there's no need for a non-CHECKPOINT_WAIT
call to wait at all; if it can't send the signal, it can nonetheless
assume that the checkpointer will eventually service the request.
A potential downside of this change is that "kill -INT" on the checkpointer
process is no longer enough to trigger a checkpoint, should anyone be
relying on something so hacky. But there's no obvious reason to do it
like that rather than issuing a plain old CHECKPOINT command, so we'll
assume that nobody is. There doesn't seem to be a way to preserve this
undocumented quasi-feature without introducing race conditions.
Since a principal reason for messing with this is to prevent intermittent
buildfarm failures, back-patch to all supported branches.
Discussion: https://postgr.es/m/27830.1552752475@sss.pgh.pa.us
Typedef name should be both unique and non-intersect with variable names
across all the sources. That makes both pg_indent and debuggers happy.
Discussion: https://postgr.es/m/23865.1552936099%40sss.pgh.pa.us
Running ALTER TABLE on any table will check if a TOAST table needs to be
added. On shared tables, this would previously fail, thus effectively
disabling ALTER TABLE for those tables. On (non-shared) system
catalogs, on the other hand, it would add a TOAST table, even though we
don't really want TOAST tables on some system catalogs. In some cases,
it would also fail with an error "AccessExclusiveLock required to add
toast table.", depending on what locks the ALTER TABLE actions had
already taken.
So instead, just ignore attempts to add TOAST tables to such tables,
outside of bootstrap mode, pretending they don't need one.
This allows running ALTER TABLE on such tables without messing up the
TOAST situation. Legitimate uses for ALTER TABLE on system catalogs
include setting reloptions (say, fillfactor or autovacuum settings).
(All this still requires allow_system_table_mods, which is independent
of this.)
Discussion: https://www.postgresql.org/message-id/flat/e49f825b-fb25-0bc8-8afc-d5ad895c7975@2ndquadrant.com
Unrecognized attribute names are supposed to be ignored. But the code
would error out on an unrecognized attribute value even if it did not
recognize the attribute name. So unrecognized attributes wouldn't
really be ignored unless the value happened to be one that matched a
recognized value. This would break some important cases where the
attribute would be processed by ucol_open() directly. Fix that and
add a test case.
The restructured code should also avoid compiler warnings about
initializing a UColAttribute value to -1, because the type might be an
unsigned enum. (reported by Andres Freund)
Aggregates have acquired a dozen or so optional attributes in recent
years for things like parallel query and moving-aggregate mode; the
lack of an OR REPLACE option to add or change these for an existing
agg makes extension upgrades gratuitously hard. Rectify.
Commit f2dec34e1 changed things so that printtup's output stringinfo
buffer was allocated outside the per-row temporary context, not inside
it. This creates a need to free that buffer explicitly when the temp
context is freed, but that was overlooked. In most cases, this is all
happening inside a portal or executor context that will go away shortly
anyhow, but that's not always true. Notably, the stringinfo ends up
getting leaked when JDBC uses row-at-a-time fetches. For a query
that returns wide rows, that adds up after awhile.
Per bug #15700 from Matthias Otterbach. Back-patch to v11 where the
faulty code was added.
Discussion: https://postgr.es/m/15700-8c408321a87d56bb@postgresql.org
We should try to prewarm each database only once. Otherwise, if
prewarming fails for some reason, it will just keep retrying in an
infnite loop. This can happen if, for example, the database has been
dropped. The existing code was intended to implement the try-once
behavior, but failed to do so because it neglected to set
worker.bgw_restart_time to BGW_NEVER_RESTART.
Mithun Cy, per a report from Hans Buschmann
Discussion: http://postgr.es/m/CA+hUKGKpQJCWcgyy3QTC9vdn6uKAR_8r__A-MMm2GYfj45caag@mail.gmail.com
Like commit f41551f61f, this aims
to make it easier to add non-Boolean options to VACUUM (or, in
this case, to ANALYZE). Instead of building up a bitmap of
options directly in the parser, build up a list of DefElem
objects and let ExecVacuum() sort it out; right now, we make
no use of the fact that a DefElem can carry an associated value,
but it will be easy to make that change in the future.
Masahiko Sawada
Discussion: http://postgr.es/m/CAD21AoATE4sn0jFFH3NcfUZXkU2BMbjBWB_kDj-XWYA-LXDcQA@mail.gmail.com
Many places need both, so this allows a few functions to take one
fewer parameter. More importantly, as soon as we add a VACUUM
option that takes a non-Boolean parameter, we need to replace
'int options' with a struct, and it seems better to think
of adding more fields to VacuumParams rather than passing around
both VacuumParams and a separate struct as well.
Patch by me, reviewed by Masahiko Sawada
Discussion: http://postgr.es/m/CA+Tgmob6g6-s50fyv8E8he7APfwCYYJ4z0wbZC2yZeSz=26CYQ@mail.gmail.com
In RI_FKey_pk_upd_check_required(), we check among other things
whether the old and new key are equal, so that we don't need to run
cascade actions when nothing has actually changed. This was using the
equality operator. But the effect of this is that if a value in the
primary key is changed to one that "looks" different but compares as
equal, the update is not propagated. (Examples are float -0 and 0 and
case-insensitive text.) This appears to violate the SQL standard, and
it also behaves inconsistently if in a multicolumn key another key is
also updated that would cause the row to compare as not equal.
To fix, if we are looking at the PK table in ri_KeysEqual(), then do a
bytewise comparison similar to record_image_eq() instead of using the
equality operators. This only makes a difference for ON UPDATE
CASCADE, but for consistency we treat all changes to the PK the same. For
the FK table, we continue to use the equality operators.
Discussion: https://www.postgresql.org/message-id/flat/3326fc2e-bc02-d4c5-e3e5-e54da466e89a@2ndquadrant.com
ce6afc6 has begun the refactoring work by plugging pg_rewind into a
central routine to update the control file, and left around two extra
copies, with one in xlog.c for the backend and one in pg_resetwal.c. By
adding an extra option to the central routine in controldata_utils.c to
control if a flush of the control file needs to be done, it is proving
to be straight-forward to make xlog.c and pg_resetwal.c use the central
code path at the condition of moving the wait event tracking there.
Hence, this allows to have only one central code path to update the
control file, shaving the code from the duplicates.
This refactoring actually fixes a problem in pg_resetwal. Previously,
the control file was first removed before being recreated. So if a
crash happened between the moment the file was removed and the moment
the file was created, then it would have been possible to not have a
control file anymore in the database folder.
Author: Fabien Coelho
Reviewed-by: Michael Paquier
Discussion: https://postgr.es/m/alpine.DEB.2.21.1903170935210.2506@lancre
This fixes an issue introduced by 266b6ac, which has added filters to
exclude file patterns on the target and source data directories to
reduce the number of files transferred. Filters get applied to both
the target and source data files, and include pg_internal.init which is
present for each database once relations are created on it. However, if
the target differed from the source with at least one new database with
relations, the rewind would fail due to the exclusion filters applied on
the target files, causing pg_internal.init to still be present on the
target database folder, while its contents should have been completely
removed so as there is nothing remaining inside at the time of the
folder deletion.
Applying exclusion filters on the source files is fine, because this way
the amount of data copied from the source to the target is reduced. And
actually, not applying the filters on the target is what pg_rewind
should do, because this causes such files to be automatically removed
during the rewind on the target. Exclusion filters apply to paths which
are removed or recreated automatically at startup, so removing all those
files on the target during the rewind is a win.
The existing set of TAP tests already stresses the rewind of databases,
but it did not include any tables on those newly-created databases.
Creating extra tables in this case is enough to reproduce the failure,
so the existing tests are extended to close the gap.
Reported-by: Mithun Cy
Author: Michael Paquier
Discussion: https://postgr.es/m/CADq3xVYt6_pO7ZzmjOqPgY9HWsL=kLd-_tNyMtdfjKqEALDyTA@mail.gmail.com
Backpatch-through: 11
pg_checksums is compiled with a given block size and has a hard
dependency to it per the way checksums are calculated via
checksum_impl.h, and trying to use the tool on a data folder which has
not the same block size would result in incorrect checksum calculations
and/or block read errors, meaning that the data folder is corrupted.
This is harmless as checksums are only checked now, but very confusing
for the user so issue an error properly if the block size used at
compilation and the block size used in the data folder do not match.
Reported-by: Sergei Kornilov
Author: Michael Banck, Michael Paquier
Reviewed-by: Fabien Coelho, Magnus Hagander
Discussion: https://postgr.es/m/20190317054657.GA3357@paquier.xyz
ackpatch-through: 11
Starting in ICU 54, collation customization attributes can be
specified in the locale string, for example
"@colStrength=primary;colCaseLevel=yes". Add support for this for
older ICU versions as well, by adding some minimal parsing of the
attributes in the locale string and calling ucol_setAttribute() on
them. This is essentially what never ICU versions do internally in
ucol_open(). This was we can offer this functionality in a consistent
way in all ICU versions supported by PostgreSQL.
Also add some tests for ICU collation customization.
Reported-by: Daniel Verite <daniel@manitou-mail.org>
Discussion: https://www.postgresql.org/message-id/0270ebd4-f67c-8774-1a5a-91adfb9bb41f@2ndquadrant.com
It looks like we can leave in most of the test cases for Infinity/NaN
inputs, but buildfarm member jacana gets the wrong answer for acosh(Inf).
It's not worth carrying a variant expected file for that, so just disable
that one test.
Discussion: https://postgr.es/m/E1h3nUY-0000sM-Vf@gemulon.postgresql.org
Add support of numeric error suppression to jsonpath as it's required by
standard. This commit doesn't use PG_TRY()/PG_CATCH() in order to implement
that. Instead, it provides internal versions of numeric functions used, which
support error suppression.
Discussion: https://postgr.es/m/fcc6fc6a-b497-f39a-923d-aa34d0c588e8%402ndQuadrant.com
Author: Alexander Korotkov, Nikita Glukhov
Reviewed-by: Tomas Vondra
SQL 2016 standards among other things contains set of SQL/JSON features for
JSON processing inside of relational database. The core of SQL/JSON is JSON
path language, allowing access parts of JSON documents and make computations
over them. This commit implements partial support JSON path language as
separate datatype called "jsonpath". The implementation is partial because
it's lacking datetime support and suppression of numeric errors. Missing
features will be added later by separate commits.
Support of SQL/JSON features requires implementation of separate nodes, and it
will be considered in subsequent patches. This commit includes following
set of plain functions, allowing to execute jsonpath over jsonb values:
* jsonb_path_exists(jsonb, jsonpath[, jsonb, bool]),
* jsonb_path_match(jsonb, jsonpath[, jsonb, bool]),
* jsonb_path_query(jsonb, jsonpath[, jsonb, bool]),
* jsonb_path_query_array(jsonb, jsonpath[, jsonb, bool]).
* jsonb_path_query_first(jsonb, jsonpath[, jsonb, bool]).
This commit also implements "jsonb @? jsonpath" and "jsonb @@ jsonpath", which
are wrappers over jsonpath_exists(jsonb, jsonpath) and jsonpath_predicate(jsonb,
jsonpath) correspondingly. These operators will have an index support
(implemented in subsequent patches).
Catversion bumped, to add new functions and operators.
Code was written by Nikita Glukhov and Teodor Sigaev, revised by me.
Documentation was written by Oleg Bartunov and Liudmila Mantrova. The work
was inspired by Oleg Bartunov.
Discussion: https://postgr.es/m/fcc6fc6a-b497-f39a-923d-aa34d0c588e8%402ndQuadrant.com
Author: Nikita Glukhov, Teodor Sigaev, Alexander Korotkov, Oleg Bartunov, Liudmila Mantrova
Reviewed-by: Tomas Vondra, Andrew Dunstan, Pavel Stehule, Alexander Korotkov
When libpq is loaded in the server (for instance, by
libpqwalreceiver), it may use libpq environment variables set in the
postmaster environment for connection parameter defaults. This has
some confusing effects in our test suites. For example, the TAP test
infrastructure sets PGAPPNAME to allow identifying clients in the
server log. But this environment variable is also inherited by
temporary servers started with pg_ctl and is then in turn used by
libpqwalreceiver as the application_name for connecting to remote
servers where it then shows up in pg_stat_replication and is relevant
for things like synchronous_standby_names. Replication already has a
suitable default for application_name, and overriding that
accidentally then requires the individual test cases to re-override
that, which is all very confusing and unnecessary.
To fix, unset PGAPPNAME temporarily before running pg_ctl start or
restart in the tests.
More comprehensive approaches like unsetting all environment variables
in pg_ctl were considered but might be too complicated to achieve
portably.
The now unnecessary re-overriding of application_name by test cases is
also removed.
Reviewed-by: Noah Misch <noah@leadboat.com>
Discussion: https://www.postgresql.org/message-id/flat/33383613-690e-6f1b-d5ba-4957ff40f6ce@2ndquadrant.com
Some buildfarm members using CLOBBER_CACHE_ALWAYS have been having OOM
problems of late. Commit 2455ab488 addressed this problem by recovering
space transiently used within RelationBuildPartitionDesc, but it turns
out that leaves quite a lot on the table, because other subroutines of
RelationBuildDesc also leak memory like mad. Let's move the temp-context
management into RelationBuildDesc so that leakage from the other
subroutines is also recovered.
I examined this issue by arranging for postgres.c to dump the size of
MessageContext just before resetting it in each command cycle, and
then running the update.sql regression test (which is one of the two
that are seeing buildfarm OOMs) with and without CLOBBER_CACHE_ALWAYS.
Before 2455ab488, the peak space usage with CCA was as much as 250MB.
That patch got it down to ~80MB, but with this patch it's about 0.5MB,
and indeed the space usage now seems nearly indistinguishable from a
non-CCA build.
RelationBuildDesc's traditional behavior of not worrying about leaking
transient data is of many years' standing, so I'm pretty hesitant to
change that without more evidence that it'd be useful in a normal build.
(So far as I can see, non-CCA memory consumption is about the same with
or without this change, whuch if anything suggests that it isn't useful.)
Hence, configure the patch so that we recover space only when
CLOBBER_CACHE_ALWAYS or CLOBBER_CACHE_RECURSIVELY is defined. However,
that choice can be overridden at compile time, in case somebody would
like to do some performance testing and try to develop evidence for
changing that decision.
It's possible that we ought to back-patch this change, but in the
absence of back-branch OOM problems in the buildfarm, I'm not in
a hurry to do that.
Discussion: https://postgr.es/m/CA+TgmoY3bRmGB6-DUnoVy5fJoreiBJ43rwMrQRCdPXuKt4Ykaw@mail.gmail.com
The trigger tests for PL/Tcl were spread aroud pltcl_setup.sql and
pltcl_queries.sql, mixed with other tests, which makes them hard to
follow and edit. Move all the trigger-related pieces to a new file
pltcl_trigger.sql. This also makes the test setup more similar to
plperl and plpython.
Add a separate walreceiver API function walrcv_server_version() to get
the version of the remote server, instead of doing it as part of
walrcv_identify_system(). This allows the server version to be
available even for uses that don't call IDENTIFY_SYSTEM, and it seems
cleaner anyway.
This is for an upcoming patch, not currently used.
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://www.postgresql.org/message-id/20190115071359.GF1433@paquier.xyz
Previously, the SERIALIZABLE isolation level prevented parallel query
from being used. Allow the two features to be used together by
sharing the leader's SERIALIZABLEXACT with parallel workers.
An extra per-SERIALIZABLEXACT LWLock is introduced to make it safe to
share, and new logic is introduced to coordinate the early release
of the SERIALIZABLEXACT required for the SXACT_FLAG_RO_SAFE
optimization, as follows:
The first backend to observe the SXACT_FLAG_RO_SAFE flag (set by
some other transaction) will 'partially release' the SERIALIZABLEXACT,
meaning that the conflicts and locks it holds are released, but the
SERIALIZABLEXACT itself will remain active because other backends
might still have a pointer to it.
Whenever any backend notices the SXACT_FLAG_RO_SAFE flag, it clears
its own MySerializableXact variable and frees local resources so that
it can skip SSI checks for the rest of the transaction. In the
special case of the leader process, it transfers the SERIALIZABLEXACT
to a new variable SavedSerializableXact, so that it can be completely
released at the end of the transaction after all workers have exited.
Remove the serializable_okay flag added to CreateParallelContext() by
commit 9da0cc35, because it's now redundant.
Author: Thomas Munro
Reviewed-by: Haribabu Kommi, Robert Haas, Masahiko Sawada, Kevin Grittner
Discussion: https://postgr.es/m/CAEepm=0gXGYhtrVDWOTHS8SQQy_=S9xo+8oCxGLWZAOoeJ=yzQ@mail.gmail.com
Peter Geoghegan