Bruce Momjian [Sat, 16 May 2015 04:40:18 +0000 (00:40 -0400)]
pg_upgrade: force timeline 1 in the new cluster
Previously, this prevented promoted standby servers from being upgraded
because of a missing WAL history file. (Timeline 1 doesn't need a
history file, and we don't copy WAL files anyway.)
Bruce Momjian [Sat, 16 May 2015 04:10:03 +0000 (00:10 -0400)]
pg_upgrade: only allow template0 to be non-connectable
This patch causes pg_upgrade to error out during its check phase if:
(1) template0 is marked connectable
or
(2) any other database is marked non-connectable
This is done because, in the first case, pg_upgrade would fail because
the pg_dumpall --globals restore would fail, and in the second case, the
database would not be restored, leading to data loss.
Andres Freund [Sat, 16 May 2015 01:40:59 +0000 (03:40 +0200)]
Support GROUPING SETS, CUBE and ROLLUP.
This SQL standard functionality allows to aggregate data by different
GROUP BY clauses at once. Each grouping set returns rows with columns
grouped by in other sets set to NULL.
This could previously be achieved by doing each grouping as a separate
query, conjoined by UNION ALLs. Besides being considerably more concise,
grouping sets will in many cases be faster, requiring only one scan over
the underlying data.
The current implementation of grouping sets only supports using sorting
for input. Individual sets that share a sort order are computed in one
pass. If there are sets that don't share a sort order, additional sort &
aggregation steps are performed. These additional passes are sourced by
the previous sort step; thus avoiding repeated scans of the source data.
The code is structured in a way that adding support for purely using
hash aggregation or a mix of hashing and sorting is possible. Sorting
was chosen to be supported first, as it is the most generic method of
implementation.
Instead of, as in an earlier versions of the patch, representing the
chain of sort and aggregation steps as full blown planner and executor
nodes, all but the first sort are performed inside the aggregation node
itself. This avoids the need to do some unusual gymnastics to handle
having to return aggregated and non-aggregated tuples from underlying
nodes, as well as having to shut down underlying nodes early to limit
memory usage. The optimizer still builds Sort/Agg node to describe each
phase, but they're not part of the plan tree, but instead additional
data for the aggregation node. They're a convenient and preexisting way
to describe aggregation and sorting. The first (and possibly only) sort
step is still performed as a separate execution step. That retains
similarity with existing group by plans, makes rescans fairly simple,
avoids very deep plans (leading to slow explains) and easily allows to
avoid the sorting step if the underlying data is sorted by other means.
A somewhat ugly side of this patch is having to deal with a grammar
ambiguity between the new CUBE keyword and the cube extension/functions
named cube (and rollup). To avoid breaking existing deployments of the
cube extension it has not been renamed, neither has cube been made a
reserved keyword. Instead precedence hacking is used to make GROUP BY
cube(..) refer to the CUBE grouping sets feature, and not the function
cube(). To actually group by a function cube(), unlikely as that might
be, the function name has to be quoted.
Needs a catversion bump because stored rules may change.
Author: Andrew Gierth and Atri Sharma, with contributions from Andres Freund Reviewed-By: Andres Freund, Noah Misch, Tom Lane, Svenne Krap, Tomas
Vondra, Erik Rijkers, Marti Raudsepp, Pavel Stehule
Discussion: CAOeZVidmVRe2jU6aMk_5qkxnB7dfmPROzM7Ur8JPW5j8Y5X-Lw@mail.gmail.com
Tom Lane [Fri, 15 May 2015 23:35:29 +0000 (19:35 -0400)]
Update time zone data files to tzdata release 2015d.
DST law changes in Egypt, Mongolia, Palestine.
Historical corrections for Canada and Chile.
Revised zone abbreviation for America/Adak (HST/HDT not HAST/HADT).
Alvaro Herrera [Fri, 15 May 2015 21:05:22 +0000 (18:05 -0300)]
Add BRIN infrastructure for "inclusion" opclasses
This lets BRIN be used with R-Tree-like indexing strategies.
Also provided are operator classes for range types, box and inet/cidr.
The infrastructure provided here should be sufficient to create operator
classes for similar datatypes; for instance, opclasses for PostGIS
geometries should be doable, though we didn't try to implement one.
(A box/point opclass was also submitted, but we ripped it out before
commit because the handling of floating point comparisons in existing
code is inconsistent and would generate corrupt indexes.)
Author: Emre Hasegeli. Cosmetic changes by me
Review: Andreas Karlsson
Alvaro Herrera [Fri, 15 May 2015 20:03:16 +0000 (17:03 -0300)]
Move strategy numbers to include/access/stratnum.h
For upcoming BRIN opclasses, it's convenient to have strategy numbers
defined in a single place. Since there's nothing appropriate, create
it. The StrategyNumber typedef now lives there, as well as existing
strategy numbers for B-trees (from skey.h) and R-tree-and-friends (from
gist.h). skey.h is forced to include stratnum.h because of the
StrategyNumber typedef, but gist.h is not; extensions that currently
rely on gist.h for rtree strategy numbers might need to add a new
A few .c files can stop including skey.h and/or gist.h, which is a nice
side benefit.
Per discussion:
https://www.postgresql.org/message-id/20150514232132.GZ2523@alvh.no-ip.org
Authored by Emre Hasegeli and Álvaro.
(It's not clear to me why bootscanner.l has any #include lines at all.)
Tom Lane [Fri, 15 May 2015 19:01:59 +0000 (15:01 -0400)]
Extend GB18030 encoding conversion to cover full Unicode range.
Our previous code for GB18030 <-> UTF8 conversion only covered Unicode code
points up to U+FFFF, but the actual spec defines conversions for all code
points up to U+10FFFF. That would be rather impractical as a lookup table,
but fortunately there is a simple algorithmic conversion between the
additional code points and the equivalent GB18030 byte patterns. Make use
of the just-added callback facility in LocalToUtf/UtfToLocal to perform the
additional conversions.
Having created the infrastructure to do that, we can use the same code to
map certain linearly-related subranges of the Unicode space below U+FFFF,
allowing removal of the corresponding lookup table entries. This more
than halves the lookup table size, which is a substantial savings;
utf8_and_gb18030.so drops from nearly a megabyte to about half that.
In support of doing that, replace ISO10646-GB18030.TXT with the data file
gb-18030-2000.xml (retrieved from
http://source.icu-project.org/repos/icu/data/trunk/charset/data/xml/ )
in which these subranges have been deleted from the simple lookup entries.
Per bug #12845 from Arjen Nienhuis. The conversion code added here is
based on his proposed patch, though I whacked it around rather heavily.
Simon Riggs [Fri, 15 May 2015 18:37:10 +0000 (14:37 -0400)]
TABLESAMPLE, SQL Standard and extensible
Add a TABLESAMPLE clause to SELECT statements that allows
user to specify random BERNOULLI sampling or block level
SYSTEM sampling. Implementation allows for extensible
sampling functions to be written, using a standard API.
Basic version follows SQLStandard exactly. Usable
concrete use cases for the sampling API follow in later
commits.
The expected output contained some floating point values which might get
rounded slightly differently on different platforms. The exact output isn't
very interesting in this test, so just round it.
Fix datatype confusion with the new lossy GiST distance functions.
We can only support a lossy distance function when the distance function's
datatype is comparable with the original ordering operator's datatype.
The distance function always returns a float8, so we are limited to float8,
and float4 (by a hard-coded cast of the float8 to float4).
In light of this limitation, it seems like a good idea to have a separate
'recheck' flag for the ORDER BY expressions, so that if you have a non-lossy
distance function, it still works with lossy quals. There are cases like
that with the build-in or contrib opclasses, but it's plausible.
There was a hidden assumption that the ORDER BY values returned by GiST
match the original ordering operator's return type, but there are plenty
of examples where that's not true, e.g. in btree_gist and pg_trgm. As long
as the distance function is not lossy, we can tolerate that and just not
return the distance to the executor (or rather, always return NULL). The
executor doesn't need the distances if there are no lossy results.
There was another little bug: the recheck variable was not initialized
before calling the distance function. That revealed the bigger issue,
as the executor tried to reorder tuples that didn't need reordering, and
that failed because of the datatype mismatch.
The previous coding effectively only verified that the second byte of a
multibyte character was in the expected range; moreover, it wasn't careful
to make sure that the second byte even exists in the buffer before touching
it. The latter seems unlikely to cause any real problems in the field
(in particular, it could never be a problem with null-terminated input),
but it's still a bug.
Since GB18030 is not a supported backend encoding, the only thing we'd
really be doing with GB18030 text is converting it to UTF8 in LocalToUtf,
which would fail anyway on any invalid character for lack of a match in
its lookup table. So the only user-visible consequence of this change
should be that you'll get "invalid byte sequence for encoding" rather than
"character has no equivalent" for malformed GB18030 input. However,
impending changes to the GB18030 conversion code will require these tighter
up-front checks to avoid producing bogus results.
Stephen Frost [Fri, 15 May 2015 14:41:53 +0000 (10:41 -0400)]
Remove useless pg_audit.conf
No need to have pg_audit.conf any longer since the regression tests are
just loading the module at the start of each session (to simulate being
in shared_preload_libraries, which isn't something we can actually make
happen on the buildfarm itself, it seems).
Allow GiST distance function to return merely a lower-bound.
The distance function can now set *recheck = false, like index quals. The
executor will then re-check the ORDER BY expressions, and use a queue to
reorder the results on the fly.
This makes it possible to do kNN-searches on polygons and circles, which
don't store the exact value in the index, but just a bounding box.
Tom Lane [Fri, 15 May 2015 02:27:07 +0000 (22:27 -0400)]
Teach UtfToLocal/LocalToUtf to support algorithmic encoding conversions.
Until now, these functions have only supported encoding conversions using
lookup tables, which is fine as long as there's not too many code points
to convert. However, GB18030 expects all 1.1 million Unicode code points
to be convertible, which would require a ridiculously-sized lookup table.
Fortunately, a large fraction of those conversions can be expressed through
arithmetic, ie the conversions are one-to-one in certain defined ranges.
To support that, provide a callback function that is used after consulting
the lookup tables. (This patch doesn't actually change anything about the
GB18030 conversion behavior, just provide infrastructure for fixing it.)
Since this requires changing the APIs of UtfToLocal/LocalToUtf anyway,
take the opportunity to rearrange their argument lists into what seems
to me a saner order. And beautify the call sites by using lengthof()
instead of error-prone sizeof() arithmetic.
In passing, also mark all the lookup tables used by these calls "const".
This moves an impressive amount of stuff into the text segment, at least
on my machine, and is safer anyhow.
Stephen Frost [Thu, 14 May 2015 19:41:39 +0000 (15:41 -0400)]
Make repeated 'make installcheck' runs work
In pg_audit, set client_min_messages up to warning, then reset the role
attributes, to completely reset the session while not making the
regression tests depend on being run by any particular user.
Stephen Frost [Thu, 14 May 2015 19:16:27 +0000 (15:16 -0400)]
Improve pg_audit regression tests
Instead of creating a new superuser role, extract out what the current
user is and use that user instead. Further, clean up and drop all
objects created by the regression test.
Tom Lane [Thu, 14 May 2015 16:08:40 +0000 (12:08 -0400)]
Support "expanded" objects, particularly arrays, for better performance.
This patch introduces the ability for complex datatypes to have an
in-memory representation that is different from their on-disk format.
On-disk formats are typically optimized for minimal size, and in any case
they can't contain pointers, so they are often not well-suited for
computation. Now a datatype can invent an "expanded" in-memory format
that is better suited for its operations, and then pass that around among
the C functions that operate on the datatype. There are also provisions
(rudimentary as yet) to allow an expanded object to be modified in-place
under suitable conditions, so that operations like assignment to an element
of an array need not involve copying the entire array.
The initial application for this feature is arrays, but it is not hard
to foresee using it for other container types like JSON, XML and hstore.
I have hopes that it will be useful to PostGIS as well.
In this initial implementation, a few heuristics have been hard-wired
into plpgsql to improve performance for arrays that are stored in
plpgsql variables. We would like to generalize those hacks so that
other datatypes can obtain similar improvements, but figuring out some
appropriate APIs is left as a task for future work. (The heuristics
themselves are probably not optimal yet, either, as they sometimes
force expansion of arrays that would be better left alone.)
Preliminary performance testing shows impressive speed gains for plpgsql
functions that do element-by-element access or update of large arrays.
There are other cases that get a little slower, as a result of added array
format conversions; but we can hope to improve anything that's annoyingly
bad. In any case most applications should see a net win.
Stephen Frost [Thu, 14 May 2015 15:55:36 +0000 (11:55 -0400)]
Further fixes for the buildfarm for pg_audit
Also, use a function to load the extension ahead of all other calls,
simulating load from shared_libraries_preload, to make sure the
hooks are in place before logging start.
Stephen Frost [Thu, 14 May 2015 14:57:12 +0000 (10:57 -0400)]
Fix buildfarm with regard to pg_audit
Remove the check that pg_audit be installed by
shared_preload_libraries as that's not going to work when running the
regressions tests in the buildfarm. That check was primairly a nice to
have and isn't required anyway.
Stephen Frost [Thu, 14 May 2015 14:36:16 +0000 (10:36 -0400)]
Add pg_audit, an auditing extension
This extension provides detailed logging classes, ability to control
logging at a per-object level, and includes fully-qualified object
names for logged statements (DML and DDL) in independent fields of the
log output.
Authors: Ian Barwick, Abhijit Menon-Sen, David Steele
Reviews by: Robert Haas, Tatsuo Ishii, Sawada Masahiko, Fujii Masao,
Simon Riggs
Discussion with: Josh Berkus, Jaime Casanova, Peter Eisentraut,
David Fetter, Yeb Havinga, Alvaro Herrera, Petr Jelinek, Tom Lane,
MauMau, Bruce Momjian, Jim Nasby, Michael Paquier,
Fabrízio de Royes Mello, Neil Tiffin
Tom Lane [Wed, 13 May 2015 22:48:05 +0000 (18:48 -0400)]
Fix distclean/maintainer-clean targets to remove top-level tmp_install dir.
The top-level makefile removes tmp_install in its "clean" target, but the
distclean and maintainer-clean targets overlooked that (and they don't
simply invoke clean, because that would result in an extra tree traversal).
While at it, let's just make sure that removing GNUmakefile itself is the
very last step of the recipe.
Tom Lane [Wed, 13 May 2015 18:05:17 +0000 (14:05 -0400)]
Fix postgres_fdw to return the right ctid value in EvalPlanQual cases.
If a postgres_fdw foreign table is a non-locked source relation in an
UPDATE, DELETE, or SELECT FOR UPDATE/SHARE, and the query selects its
ctid column, the wrong value would be returned if an EvalPlanQual
recheck occurred. This happened because the foreign table's result row
was copied via the ROW_MARK_COPY code path, and EvalPlanQualFetchRowMarks
just unconditionally set the reconstructed tuple's t_self to "invalid".
To fix that, we can have EvalPlanQualFetchRowMarks copy the composite
datum's t_ctid field, and be sure to initialize that along with t_self
when postgres_fdw constructs a tuple to return.
If we just did that much then EvalPlanQualFetchRowMarks would start
returning "(0,0)" as ctid for all other ROW_MARK_COPY cases, which perhaps
does not matter much, but then again maybe it might. The cause of that is
that heap_form_tuple, which is the ultimate source of all composite datums,
simply leaves t_ctid as zeroes in newly constructed tuples. That seems
like a bad idea on general principles: a field that's really not been
initialized shouldn't appear to have a valid value. So let's eat the
trivial additional overhead of doing "ItemPointerSetInvalid(&(td->t_ctid))"
in heap_form_tuple.
This closes out our handling of Etsuro Fujita's report that tableoid and
ctid weren't correctly set in postgres_fdw EvalPlanQual cases. Along the
way we did a great deal of work to improve FDWs' ability to control row
locking behavior; which was not wasted effort by any means, but it didn't
end up being a fix for this problem because that feature would be too
expensive for postgres_fdw to use all the time.
Although the fix for the tableoid misbehavior was back-patched, I'm
hesitant to do so here; it seems far less likely that people would care
about remote ctid than tableoid, and even such a minor behavioral change
as this in heap_form_tuple is perhaps best not back-patched. So commit
to HEAD only, at least for the moment.
Andrew Dunstan [Wed, 13 May 2015 17:52:08 +0000 (13:52 -0400)]
Fix jsonb replace and delete on scalars and empty structures
These operations now error out if attempted on scalars, and simply
return the input if attempted on empty arrays or objects. Along the way
we remove the unnecessary cloning of the input when it's known to be
unchanged. Regression tests covering these cases are added.
Andres Freund [Wed, 13 May 2015 05:31:04 +0000 (07:31 +0200)]
Add pgstattuple_approx() to the pgstattuple extension.
The new function allows to estimate bloat and other table level statics
in a faster, but approximate, way. It does so by using information from
the free space map for pages marked as all visible in the visibility
map. The rest of the table is actually read and free space/bloat is
measured accurately. In many cases that allows to get bloat information
much quicker, causing less IO.
Author: Abhijit Menon-Sen Reviewed-By: Andres Freund, Amit Kapila and Tomas Vondra
Discussion: 20140402214144.GA28681@kea.toroid.org
Peter Eisentraut [Wed, 13 May 2015 02:52:18 +0000 (22:52 -0400)]
PL/Python: Remove procedure cache invalidation
This was added to react to changes in the pg_transform catalog, but
building with CLOBBER_CACHE_ALWAYS showed that PL/Python was not
prepared for having its procedure cache cleared. Since this is a
marginal use case, and we don't do this for other catalogs anyway, we
can postpone this to another day.
Andres Freund [Tue, 12 May 2015 22:13:22 +0000 (00:13 +0200)]
Fix ON CONFLICT bugs that manifest when used in rules.
Specifically the tlist and rti of the pseudo "excluded" relation weren't
properly treated by expression_tree_walker, which lead to errors when
excluded was referenced inside a rule because the varnos where not
properly adjusted. Similar omissions in OffsetVarNodes and
expression_tree_mutator had less impact, but should obviously be fixed
nonetheless.
A couple tests of for ON CONFLICT UPDATE into INSERT rule bearing
relations have been added.
Andrew Dunstan [Tue, 12 May 2015 19:52:45 +0000 (15:52 -0400)]
Additional functions and operators for jsonb
jsonb_pretty(jsonb) produces nicely indented json output.
jsonb || jsonb concatenates two jsonb values.
jsonb - text removes a key and its associated value from the json
jsonb - int removes the designated array element
jsonb - text[] removes a key and associated value or array element at
the designated path
jsonb_replace(jsonb,text[],jsonb) replaces the array element designated
by the path or the value associated with the key designated by the path
with the given value.
Original work by Dmitry Dolgov, adapted and reworked for PostgreSQL core
by Andrew Dunstan, reviewed and tidied up by Petr Jelinek.
Tom Lane [Tue, 12 May 2015 18:10:10 +0000 (14:10 -0400)]
Add support for doing late row locking in FDWs.
Previously, FDWs could only do "early row locking", that is lock a row as
soon as it's fetched, even though local restriction/join conditions might
discard the row later. This patch adds callbacks that allow FDWs to do
late locking in the same way that it's done for regular tables.
To make use of this feature, an FDW must support the "ctid" column as a
unique row identifier. Currently, since ctid has to be of type TID,
the feature is of limited use, though in principle it could be used by
postgres_fdw. We may eventually allow FDWs to specify another data type
for ctid, which would make it possible for more FDWs to use this feature.
This commit does not modify postgres_fdw to use late locking. We've
tested some prototype code for that, but it's not in committable shape,
and besides it's quite unclear whether it actually makes sense to do late
locking against a remote server. The extra round trips required are likely
to outweigh any benefit from improved concurrency.
Etsuro Fujita, reviewed by Ashutosh Bapat, and hacked up a lot by me
Stephen Frost [Tue, 12 May 2015 17:13:12 +0000 (13:13 -0400)]
pgbench: Don't fail during startup
In pgbench, report, but ignore, any errors returned when attempting to
vacuum/truncate the default tables during startup. If the tables are
needed, we'll error out soon enough anyway.
Per discussion with Tatsuo, David Rowley, Jim Nasby, Robert, Andres,
Fujii, Fabrízio de Royes Mello, Tomas Vondra, Michael Paquier, Peter,
based on a suggestion from Jeff Janes, patch from Robert, additional
message wording from Tom.
Andrew Dunstan [Tue, 12 May 2015 13:29:10 +0000 (09:29 -0400)]
Map basebackup tablespaces using a tablespace_map file
Windows can't reliably restore symbolic links from a tar format, so
instead during backup start we create a tablespace_map file, which is
used by the restoring postgres to create the correct links in pg_tblspc.
The backup protocol also now has an option to request this file to be
included in the backup stream, and this is used by pg_basebackup when
operating in tar mode.
This is done on all platforms, not just Windows.
This means that pg_basebackup will not not work in tar mode against 9.4
and older servers, as this protocol option isn't implemented there.
Amit Kapila, reviewed by Dilip Kumar, with a little editing from me.
Alvaro Herrera [Mon, 11 May 2015 22:14:31 +0000 (19:14 -0300)]
Allow on-the-fly capture of DDL event details
This feature lets user code inspect and take action on DDL events.
Whenever a ddl_command_end event trigger is installed, DDL actions
executed are saved to a list which can be inspected during execution of
a function attached to ddl_command_end.
The set-returning function pg_event_trigger_ddl_commands can be used to
list actions so captured; it returns data about the type of command
executed, as well as the affected object. This is sufficient for many
uses of this feature. For the cases where it is not, we also provide a
"command" column of a new pseudo-type pg_ddl_command, which is a
pointer to a C structure that can be accessed by C code. The struct
contains all the info necessary to completely inspect and even
reconstruct the executed command.
There is no actual deparse code here; that's expected to come later.
What we have is enough infrastructure that the deparsing can be done in
an external extension. The intention is that we will add some deparsing
code in a later release, as an in-core extension.
A new test module is included. It's probably insufficient as is, but it
should be sufficient as a starting point for a more complete and
future-proof approach.
Authors: Álvaro Herrera, with some help from Andres Freund, Ian Barwick,
Abhijit Menon-Sen.
Reviews by Andres Freund, Robert Haas, Amit Kapila, Michael Paquier,
Craig Ringer, David Steele.
Additional input from Chris Browne, Dimitri Fontaine, Stephen Frost,
Petr Jelínek, Tom Lane, Jim Nasby, Steven Singer, Pavel Stěhule.
Based on original work by Dimitri Fontaine, though I didn't use his
code.
Stephen Frost [Mon, 11 May 2015 19:44:12 +0000 (15:44 -0400)]
Allow LOCK TABLE .. ROW EXCLUSIVE MODE with INSERT
INSERT acquires RowExclusiveLock during normal operation and therefore
it makes sense to allow LOCK TABLE .. ROW EXCLUSIVE MODE to be executed
by users who have INSERT rights on a table (even if they don't have
UPDATE or DELETE).
Not back-patching this as it's a behavior change which, strictly
speaking, loosens security restrictions.
Tom Lane [Mon, 11 May 2015 16:25:28 +0000 (12:25 -0400)]
Fix incorrect checking of deferred exclusion constraint after a HOT update.
If a row that potentially violates a deferred exclusion constraint is
HOT-updated later in the same transaction, the exclusion constraint would
be reported as violated when the check finally occurs, even if the row(s)
the new row originally conflicted with have since been removed. This
happened because the wrong TID was passed to check_exclusion_constraint(),
causing the live HOT-updated row to be seen as a conflicting row rather
than recognized as the row-under-test.
Per bug #13148 from Evan Martin. It's been broken since exclusion
constraints were invented, so back-patch to all supported branches.
Robert Haas [Mon, 11 May 2015 16:07:13 +0000 (12:07 -0400)]
Increase threshold for multixact member emergency autovac to 50%.
Analysis by Noah Misch shows that the 25% threshold set by commit 53bb309d2d5a9432d2602c93ed18e58bd2924e15 is lower than any other,
similar autovac threshold. While we don't know exactly what value
will be optimal for all users, it is better to err a little on the
high side than on the low side. A higher value increases the risk
that users might exhaust the available space and start seeing errors
before autovacuum can clean things up sufficiently, but a user who
hits that problem can compensate for it by reducing
autovacuum_multixact_freeze_max_age to a value dependent on their
average multixact size. On the flip side, if the emergency cap
imposed by that patch kicks in too early, the user will experience
excessive wraparound scanning and will be unable to mitigate that
problem by configuration. The new value will hopefully reduce the
risk of such bad experiences while still providing enough headroom
to avoid multixact member exhaustion for most users.
Along the way, adjust the documentation to reflect the effects of
commit 04e6d3b877e060d8445eb653b7ea26b1ee5cec6b, which taught
autovacuum to run for multixact wraparound even when autovacuum
is configured off.
Robert Haas [Mon, 11 May 2015 02:21:20 +0000 (22:21 -0400)]
Advance the stop point for multixact offset creation only at checkpoint.
Commit b69bf30b9bfacafc733a9ba77c9587cf54d06c0c advanced the stop point
at vacuum time, but this has subsequently been shown to be unsafe as a
result of analysis by myself and Thomas Munro and testing by Thomas
Munro. The crux of the problem is that the SLRU deletion logic may
get confused about what to remove if, at exactly the right time during
the checkpoint process, the head of the SLRU crosses what used to be
the tail.
This patch, by me, fixes the problem by advancing the stop point only
following a checkpoint. This has the additional advantage of making
the removal logic work during recovery more like the way it works during
normal running, which is probably good.
At least one of the calls to DetermineSafeOldestOffset which this patch
removes was already dead, because MultiXactAdvanceOldest is called only
during recovery and DetermineSafeOldestOffset was set up to do nothing
during recovery. That, however, is inconsistent with the principle that
recovery and normal running should work similarly, and was confusing to
boot.
Along the way, fix some comments that previous patches in this area
neglected to update. It's not clear to me whether there's any
concrete basis for the decision to use only half of the multixact ID
space, but it's neither necessary nor sufficient to prevent multixact
member wraparound, so the comments should not say otherwise.
Tom Lane [Sun, 10 May 2015 18:36:30 +0000 (14:36 -0400)]
Code review for foreign/custom join pushdown patch.
Commit e7cb7ee14555cc9c5773e2c102efd6371f6f2005 included some design
decisions that seem pretty questionable to me, and there was quite a lot
of stuff not to like about the documentation and comments. Clean up
as follows:
* Consider foreign joins only between foreign tables on the same server,
rather than between any two foreign tables with the same underlying FDW
handler function. In most if not all cases, the FDW would simply have had
to apply the same-server restriction itself (far more expensively, both for
lack of caching and because it would be repeated for each combination of
input sub-joins), or else risk nasty bugs. Anyone who's really intent on
doing something outside this restriction can always use the
set_join_pathlist_hook.
* Rename fdw_ps_tlist/custom_ps_tlist to fdw_scan_tlist/custom_scan_tlist
to better reflect what they're for, and allow these custom scan tlists
to be used even for base relations.
* Change make_foreignscan() API to include passing the fdw_scan_tlist
value, since the FDW is required to set that. Backwards compatibility
doesn't seem like an adequate reason to expect FDWs to set it in some
ad-hoc extra step, and anyway existing FDWs can just pass NIL.
* Change the API of path-generating subroutines of add_paths_to_joinrel,
and in particular that of GetForeignJoinPaths and set_join_pathlist_hook,
so that various less-used parameters are passed in a struct rather than
as separate parameter-list entries. The objective here is to reduce the
probability that future additions to those parameter lists will result in
source-level API breaks for users of these hooks. It's possible that this
is even a small win for the core code, since most CPU architectures can't
pass more than half a dozen parameters efficiently anyway. I kept root,
joinrel, outerrel, innerrel, and jointype as separate parameters to reduce
code churn in joinpath.c --- in particular, putting jointype into the
struct would have been problematic because of the subroutines' habit of
changing their local copies of that variable.
* Avoid ad-hocery in ExecAssignScanProjectionInfo. It was probably all
right for it to know about IndexOnlyScan, but if the list is to grow
we should refactor the knowledge out to the callers.
* Restore nodeForeignscan.c's previous use of the relcache to avoid
extra GetFdwRoutine lookups for base-relation scans.
* Lots of cleanup of documentation and missed comments. Re-order some
code additions into more logical places.
Andrew Dunstan [Sat, 9 May 2015 17:06:49 +0000 (13:06 -0400)]
Add new OID alias type regrole
The new type has the scope of whole the database cluster so it doesn't
behave the same as the existing OID alias types which have database
scope,
concerning object dependency. To avoid confusion constants of the new
type are prohibited from appearing where dependencies are made involving
it.
Also, add a note to the docs about possible MVCC violation and
optimization issues, which are general over the all reg* types.
Stephen Frost [Sat, 9 May 2015 15:13:37 +0000 (11:13 -0400)]
Improve ParseConfigFp comment wrt head/tail
The head_p and tail_p pointers passed to ParseConfigFp() are actually
input/output parameters, not strictly output paramaters. This updates
the function comment to reflect that.
Stephen Frost [Fri, 8 May 2015 23:39:42 +0000 (19:39 -0400)]
Change default for include_realm to 1
The default behavior for GSS and SSPI authentication methods has long
been to strip the realm off of the principal, however, this is not a
secure approach in multi-realm environments and the use-case for the
parameter at all has been superseded by the regex-based mapping support
available in pg_ident.conf.
Change the default for include_realm to be '1', meaning that we do
NOT remove the realm from the principal by default. Any installations
which depend on the existing behavior will need to update their
configurations (ideally by leaving include_realm set to 1 and adding a
mapping in pg_ident.conf, but alternatively by explicitly setting
include_realm=0 prior to upgrading). Note that the mapping capability
exists in all currently supported versions of PostgreSQL and so this
change can be done today. Barring that, existing users can update their
configurations today to explicitly set include_realm=0 to ensure that
the prior behavior is maintained when they upgrade.
Stephen Frost [Fri, 8 May 2015 23:25:30 +0000 (19:25 -0400)]
Modify pg_stat_get_activity to build a tuplestore
This updates pg_stat_get_activity() to build a tuplestore for its
results instead of using the old-style multiple-call method. This
simplifies the function, though that wasn't the primary motivation for
the change, which is that we may turn it into a helper function which
can filter the results (or not) much more easily.
Stephen Frost [Fri, 8 May 2015 23:09:26 +0000 (19:09 -0400)]
Add pg_file_settings view and function
The function and view added here provide a way to look at all settings
in postgresql.conf, any #include'd files, and postgresql.auto.conf
(which is what backs the ALTER SYSTEM command).
The information returned includes the configuration file name, line
number in that file, sequence number indicating when the parameter is
loaded (useful to see if it is later masked by another definition of the
same parameter), parameter name, and what it is set to at that point.
This information is updated on reload of the server.
This is unfiltered, privileged, information and therefore access is
restricted to superusers through the GRANT system.
Author: Sawada Masahiko, various improvements by me.
Reviewers: David Steele
Andres Freund [Fri, 8 May 2015 20:22:05 +0000 (22:22 +0200)]
Fix two problems in infer_arbiter_indexes().
The first is a pretty simple bug where a relcache entry is used after
the relation is closed. In this particular situation it does not appear
to have bad consequences unless compiled with RELCACHE_FORCE_RELEASE.
The second is that infer_arbiter_indexes() skipped indexes that aren't
yet valid according to indcheckxmin. That's not required here, because
uniqueness checks don't care about visibility according to an older
snapshot. While thats not really a bug, it makes things undesirably
non-deterministic. There is some hope that this explains a test failure
on buildfarm member jaguarundi.
At promotion, archive last segment from old timeline with .partial suffix.
Previously, we would archive the possible-incomplete WAL segment with its
normal filename, but that causes trouble if the server owning that timeline
is still running, and tries to archive the same segment later. It's not nice
for the standby to trip up the master's archival like that. And it's pretty
confusing, anyway, to have an incomplete segment in the archive that's
indistinguishable from a normal, complete segment.
To avoid such confusion, add a .partial suffix to the file. Or to be more
precise, make a copy of the old segment under the .partial suffix, and
archive that instead of the original file. pg_receivexlog also uses the
.partial suffix for the same purpose, to tell apart incompletely streamed
files from complete ones.
There is no automatic mechanism to use the .partial files at recovery, so
they will go unused, unless the administrator manually copies to them to
the pg_xlog directory (and removes the .partial suffix). Recovery won't
normally need the WAL - when recovering to the new timeline, it will find
the same WAL on the first segment on the new timeline instead - but it
nevertheless feels better to archive the file with the .partial suffix, for
debugging purposes if nothing else.
To fix this, progressively ramp down the effective value (but not the
actual contents) of autovacuum_multixact_freeze_max_age as member space
utilization increases. This makes autovacuum more aggressive and also
reduces the threshold for a manual VACUUM to perform a full-table scan.
This patch leaves unsolved the problem of ensuring that emergency
autovacuums are triggered even when autovacuum=off. We'll need to fix
that via a separate patch.
projects / postgresql / log