When a standby server follows the master using WAL archive, and it chooses
a new timeline (recovery_target_timeline='latest'), it only fetches the
timeline history file for the chosen target timeline, not any other history
files that might be missing from pg_xlog. For example, if the current
timeline is 2, and we choose 4 as the new recovery target timeline, the
history file for timeline 3 is not fetched, even if it's part of this
server's history. That's enough for the standby itself - the history file
for timeline 4 includes timeline 3 as well - but if a cascading standby
server wants to recover to timeline 3, it needs the history file. To fix,
when a new recovery target timeline is chosen, try to copy any missing
history files from the archive to pg_xlog between the old and new target
timeline.
A second similar issue was with the WAL files. When a standby recovers from
archive, and it reaches a segment that contains a switch to a new timeline,
recovery fetches only the WAL file labelled with the new timeline's ID. The
file from the new timeline contains a copy of the WAL from the old timeline
up to the point where the switch happened, and recovery recovers it from the
new file. But in streaming replication, walsender only tries to read it
from the old timeline's file. To fix, change walsender to read it from the
new file, so that it behaves the same as recovery in that sense, and doesn't
try to open the possibly nonexistent file with the old timeline's ID.
Originally we didn't bother to mark FuncExprs with any indication whether
VARIADIC had been given in the source text, because there didn't seem to be
any need for it at runtime. However, because we cannot fold a VARIADIC ANY
function's arguments into an array (since they're not necessarily all the
same type), we do actually need that information at runtime if VARIADIC ANY
functions are to respond unsurprisingly to use of the VARIADIC keyword.
Add the missing field, and also fix ruleutils.c so that VARIADIC ANY
function calls are dumped properly.
Extracted from a larger patch that also fixes concat() and format() (the
only two extant VARIADIC ANY functions) to behave properly when VARIADIC is
specified. This portion seems appropriate to review and commit separately.
Pavel Stehule
Remove duplicate implementations of catalog munging and miscellaneous
privilege checks. Instead rely on already existing data in
objectaddress.c to do the work.
Author: KaiGai Kohei, changes by me
Reviewed by: Robert Haas, Álvaro Herrera, Dimitri Fontaine
AtEOXact_RelationCache() scanned the entire relation cache at the end of
any transaction that created a new relation or assigned a new relfilenode.
Thus, clients such as pg_restore had an O(N^2) performance problem that
would start to be noticeable after creating 10000 or so tables. Since
typically only a small number of relcache entries need any cleanup, we
can fix this by keeping a small list of their OIDs and doing hash_searches
for them. We fall back to the full-table scan if the list overflows.
Ideally, the maximum list length would be set at the point where N
hash_searches would cost just less than the full-table scan. Some quick
experimentation says that point might be around 50-100; I (tgl)
conservatively set MAX_EOXACT_LIST = 32. For the case that we're worried
about here, which is short single-statement transactions, it's unlikely
there would ever be more than about a dozen list entries anyway; so it's
probably not worth being too tense about the value.
We could avoid the hash_searches by instead keeping the target relcache
entries linked into a list, but that would be noticeably more complicated
and bug-prone because of the need to maintain such a list in the face of
relcache entry drops. Since a relcache entry can only need such cleanup
after a somewhat-heavyweight filesystem operation, trying to save a
hash_search per cleanup doesn't seem very useful anyway --- it's the scan
over all the not-needing-cleanup entries that we wish to avoid here.
Jeff Janes, reviewed and tweaked a bit by Tom Lane
Use of SnapshotNow is known to expose us to race conditions if the tuple(s)
being sought could be updated by concurrently-committing transactions.
CREATE DATABASE and DROP DATABASE are particularly exposed because they do
heavyweight filesystem operations during their scans of pg_tablespace,
so that the scans run for a very long time compared to most. Furthermore,
the potential consequences of a missed or twice-visited row are nastier
than average:
* createdb() could fail with a bogus "file already exists" error, or
silently fail to copy one or more tablespace's worth of files into the
new database.
* remove_dbtablespaces() could miss one or more tablespaces, thus failing
to free filesystem space for the dropped database.
* check_db_file_conflict() could likewise miss a tablespace, leading to an
OID conflict that could result in data loss either immediately or in
future operations. (This seems of very low probability, though, since a
duplicate database OID would be unlikely to start with.)
Hence, it seems worth fixing these three places to use MVCC snapshots, even
though this will someday be superseded by a generic solution to SnapshotNow
race conditions.
Back-patch to all active branches.
Stephen Frost and Tom Lane
This got broken in the original fast-path locking patch, because
I failed to account for the fact that Hot Standby startup process
might take a strong relation lock on a relation in a database to
which it is not bound, and confused MyDatabaseId with the database
ID of the relation being locked.
Report and diagnosis by Andres Freund. Final form of patch by me.
of timeline, take advantage of that in walreceiver.
Startup process is still in control of choosign the target timeline, by
scanning the timeline history files present in pg_xlog, but walreceiver now
uses the next timeline's ID to fetch its history file immediately after it
has finished streaming the old timeline. Before, the standby would first try
to restart streaming on the old timeline, which fetches the missing timeline
history file as a side-effect, and only then restart from the new timeline.
This patch eliminates the extra iteration, which speeds up the timeline
switch and reduces the noise in the log caused by the extra restart on the
old timeline.
The xlogreader refactoring broke the logic to decide which timeline to start
streaming from. XLogPageRead() uses the timeline history to check which
timeline the requested WAL position falls into. However, after the
refactoring, XLogPageRead() is always first called with the first page in
the segment, to verify the segment header, and only then with the actual WAL
position we're interested in. That first read of the segment's header made
XLogPageRead() to always start streaming from the old timeline containing
the segment header, not the timeline containing the actual record, if there
was a timeline switch within the segment.
I thought I fixed this yesterday, but that fix was too narrow and only fixed
this for the corner-case that the timeline switch happened in the first page
of the segment. To fix this more robustly, pass explicitly the position of
the record we're actually interested in to XLogPageRead, and use that to
decide which timeline to read from, rather than deduce it from the page and
offset.
Per report from Fujii Masao.
get a large enough part of the page to include the beginning of the next
record we're interested in. The XLogPageRead callback uses the requested
length to decide which timeline to stream WAL from, and if the first call
is short, and the page contains a timeline switch, we'll repeatedly try
to stream that page from the old timeline, and never get across the
timeline switch.
The patch to allow pg_receivexlog to switch timeline added a result set
after copy has ended in START_STREAMING command, to return the next
timeline's ID to the client. But walreceived didn't get the memo, and threw
an error on the unexpected result set. Fix.
When relations are dropped, at end of transaction we need to remove the
files and clean the buffer pool of buffers containing pages of those
relations. Previously we would scan the buffer pool once per relation
to clean up buffers. When there are many relations to drop, the
repeated scans make this process slow; so we now instead pass a list of
relations to drop and scan the pool once, checking each buffer against
the passed list. When the number of relations is larger than a
threshold (which as of this patch is being set to 20 relations) we sort
the array before starting, and bsearch the array; when it's smaller, we
simply scan the array linearly each time, because that's faster. The
exact optimal threshold value depends on many factors, but the
difference is not likely to be significant enough to justify making it
user-settable.
This has been measured to be a significant win (a 15x win when dropping
100,000 relations; an extreme case, but reportedly a real one).
Author: Tomas Vondra, some tweaks by me
Reviewed by: Robert Haas, Shigeru Hanada, Andres Freund, Álvaro Herrera
This mirrors the changes done earlier to the server in standby mode. When
receivelog reaches the end of a timeline, as reported by the server, it
fetches the timeline history file of the next timeline, and restarts
streaming from the new timeline by issuing a new START_STREAMING command.
When pg_receivexlog crosses a timeline, it leaves the .partial suffix on the
last segment on the old timeline. This helps you to tell apart a partial
segment left in the directory because of a timeline switch, and a completed
segment. If you just follow a single server, it won't make a difference, but
it can be significant in more complicated scenarios where new WAL is still
generated on the old timeline.
This includes two small changes to the streaming replication protocol:
First, when you reach the end of timeline while streaming, the server now
sends the TLI of the next timeline in the server's history to the client.
pg_receivexlog uses that as the next timeline, so that it doesn't need to
parse the timeline history file like a standby server does. Second, when
BASE_BACKUP command sends the begin and end WAL positions, it now also sends
the timeline IDs corresponding the positions.
The code originally just doubled the size of the tuple-pointer array so
long as that would fit in allowedMem. This could result in failing to use
as much as half of allowedMem, if (as is typical) the last doubling attempt
didn't quite fit. Worse, we might double the array size but be unable to
use most of the added slots, because there was no room left within the
allowedMem limit for tuples the slots should point to. To fix, double only
so long as we've used less than half of allowedMem in total. Then do one
more array enlargement, but scale it based on total memory consumption so
far. This will work nicely as long as the average tuple size is reasonably
stable, and in any case should be better than the old method.
This change will result in large sort operations consuming a larger
fraction of work_mem than they typically did in the past. The release
notes should mention that users may want to revisit their work_mem
settings, if they'd tuned those settings based on the old behavior of
sorting.
Jeff Janes, reviewed by Peter Geoghegan and Robert Haas
XLogReadRecord should reset its state on every error, to make sure it
re-reads the page on next call. It was inconsistent in that some errors did
that, but some did not.
In ReadRecord(), don't give up on an error if we're in standby mode. The
loop was set up to retry, but the checks within the loop broke out of the
loop on any error.
Andres Freund, with some tweaking by me.
The patch that turned XLogRecPtr into a uint64 inadvertently changed the
on-disk format of GiST indexes, because the NSN field in the GiST page
opaque is an XLogRecPtr. That breaks pg_upgrade. Revert the format of that
field back to the two-field struct that XLogRecPtr was before. This is the
same we did to LSNs in the page header to avoid changing on-disk format.
Bump catversion, as this invalidates any existing GiST indexes built on
9.3devel.
It's better to start from what the OpenSSL people consider a good
default and then remove insecure things (low encryption, exportable
encryption and md5 at this point) from that, instead of starting
from everything that exists and remove from that. We trust the
OpenSSL people to make good choices about what the default is.
This new facility can not only be used by xlog.c to carry out crash
recovery, but also by external programs. By supplying a function to
read XLog pages from somewhere, all the WAL reading can be used for
completely different purposes.
For the standard backend use, the behavior should be pretty much the
same as previously. As for non-backend programs, an hypothetical
pg_xlogdump program is now closer to reality, but some more backend
support is still necessary.
This patch was originally submitted by Andres Freund in a different
form, but Heikki Linnakangas opted for and authored another design of
the concept. Andres has advanced the patch since Heikki's initial
version. Review and some (mostly cosmetics) changes by me.
When attempting to move an object into the schema in which it already
was, for most objects classes we were correctly complaining about
exactly that ("object is already in schema"); but for some other object
classes, such as functions, we were instead complaining of a name
collision ("object already exists in schema"). The latter is wrong and
misleading, per complaint from Robert Haas in
CA+TgmoZ0+gNf7RDKRc3u5rHXffP=QjqPZKGxb4BsPz65k7qnHQ@mail.gmail.com
To fix, refactor the way these checks are done. As a bonus, the
resulting code is smaller and can also share some code with Rename
cases.
While at it, remove use of getObjectDescriptionOids() in error messages.
These are normally disallowed because of translatability considerations,
but this one had slipped through since 9.1. (Not sure that this is
worth backpatching, though, as it would create some untranslated
messages in back branches.)
This is loosely based on a patch by KaiGai Kohei, heavily reworked by
me.
Original coding would corrupt the hashtable if the item being updated was
at the end of its bucket chain and the new hash key hashed to that same
bucket. Diagnosis and fix by Heikki Linnakangas.
Because the return value of lseek() was assigned to an unsigned size_t
variable, we'd fail to notice an error return code -1. Compiler gave a
warning about this.
Andres Freund
Dates outside the supported range could be entered, but would not print
reasonably, and operations such as conversion to timestamp wouldn't behave
sanely either. Since this has the potential to result in undumpable table
data, it seems worth back-patching.
Hitoshi Harada
The code in PostPrepare_Locks supposed that it could reassign locks to
the prepared transaction's dummy PGPROC by deleting the PROCLOCK table
entries and immediately creating new ones. This was safe when that code
was written, but since we invented partitioning of the shared lock table,
it's not safe --- another process could steal away the PROCLOCK entry in
the short interval when it's on the freelist. Then, if we were otherwise
out of shared memory, PostPrepare_Locks would have to PANIC, since it's
too late to back out of the PREPARE at that point.
Fix by inventing a dynahash.c function to atomically update a hashtable
entry's key. (This might possibly have other uses in future.)
This is an ancient bug that in principle we ought to back-patch, but the
odds of someone hitting it in the field seem really tiny, because (a) the
risk window is small, and (b) nobody runs servers with maxed-out lock
tables for long, because they'll be getting non-PANIC out-of-memory errors
anyway. So fixing it in HEAD seems sufficient, at least until the new
code has gotten some testing.
In commit 71450d7fd6, we added code to inform
suitably-intelligent compilers that ereport() doesn't return if the elevel
is ERROR or higher. This patch extends that to elog(), and also fixes a
double-evaluation hazard that the previous commit created in ereport(),
as well as reducing the emitted code size.
The elog() improvement requires the compiler to support __VA_ARGS__, which
should be available in just about anything nowadays since it's required by
C99. But our minimum language baseline is still C89, so add a configure
test for that.
The previous commit assumed that ereport's elevel could be evaluated twice,
which isn't terribly safe --- there are already counterexamples in xlog.c.
On compilers that have __builtin_constant_p, we can use that to protect the
second test, since there's no possible optimization gain if the compiler
doesn't know the value of elevel. Otherwise, use a local variable inside
the macros to prevent double evaluation. The local-variable solution is
inferior because (a) it leads to useless code being emitted when elevel
isn't constant, and (b) it increases the optimization level needed for the
compiler to recognize that subsequent code is unreachable. But it seems
better than not teaching non-gcc compilers about unreachability at all.
Lastly, if the compiler has __builtin_unreachable(), we can use that
instead of abort(), resulting in a noticeable code savings since no
function call is actually emitted. However, it seems wise to do this only
in non-assert builds. In an assert build, continue to use abort(), so that
the behavior will be predictable and debuggable if the "impossible"
happens.
These changes involve making the ereport and elog macros emit do-while
statement blocks not just expressions, which forces small changes in
a few call sites.
Andres Freund, Tom Lane, Heikki Linnakangas
Historically we've used a couple of very ad-hoc fudge factors to try to
get the right results when indexes of different sizes would satisfy a
query with the same number of index leaf tuples being visited. In
commit 21a39de580 I tweaked one of these
fudge factors, with results that proved disastrous for larger indexes.
Commit bf01e34b55 fudged it some more,
but still with not a lot of principle behind it.
What seems like a better way to address these issues is to explicitly model
index-descent costs, since that's what's really at stake when considering
diferent indexes with similar leaf-page-level costs. We tried that once
long ago, and found that charging random_page_cost per page descended
through was way too much, because upper btree levels tend to stay in cache
in real-world workloads. However, there's still CPU costs to think about,
and the previous fudge factors can be seen as a crude attempt to account
for those costs. So this patch replaces those fudge factors with explicit
charges for the number of tuple comparisons needed to descend the index
tree, plus a small charge per page touched in the descent. The cost
multipliers are chosen so that the resulting charges are in the vicinity of
the historical (pre-9.2) fudge factors for indexes of up to about a million
tuples, while not ballooning unreasonably beyond that, as the old fudge
factor did (even more so in 9.2).
To make this work accurately for btree indexes, add some code that allows
extraction of the known root-page height from a btree. There's no
equivalent number readily available for other index types, but we can use
the log of the number of index pages as an approximate substitute.
This seems like too much of a behavioral change to risk back-patching,
but it should improve matters going forward. In 9.2 I'll just revert
the fudge-factor change.
If VirtualXactLock() has to wait for a transaction that holds its VXID lock
as a fast-path lock, it must first convert the fast-path lock to a regular
lock. It failed to take the required "partition" lock on the main
shared-memory lock table while doing so. This is the direct cause of the
assert failure in GetLockStatusData() recently observed in the buildfarm,
but more worryingly it could result in arbitrary corruption of the shared
lock table if some other process were concurrently engaged in modifying the
same partition of the lock table. Fortunately, VirtualXactLock() is only
used by CREATE INDEX CONCURRENTLY and DROP INDEX CONCURRENTLY, so the
opportunities for failure are fewer than they might have been.
In passing, improve some comments and be a bit more consistent about
order of operations.
SPI_execute() and related functions create a CachedPlan, execute it once,
and immediately discard it, so that the functionality offered by
plancache.c is of no value in this code path. And performance measurements
show that the extra data copying and invalidation checking done by
plancache.c slows down simple queries by 10% or more compared to 9.1.
However, enough of the SPI code is shared with functions that do need plan
caching that it seems impractical to bypass plancache.c altogether.
Instead, let's invent a variant version of cached plans that preserves
99% of the API but doesn't offer any of the actual functionality, nor the
overhead. This puts SPI_execute() performance back on par, or maybe even
slightly better, than it was before. This change should resolve recent
complaints of performance degradation from Dong Ye, Pavel Stehule, and
others.
By avoiding data copying, this change also reduces the amount of memory
needed to execute many-statement SPI_execute() strings, as for instance in
a recent complaint from Tomas Vondra.
An additional benefit of this change is that multi-statement SPI_execute()
query strings are now processed fully serially, that is we complete
execution of earlier statements before running parse analysis and planning
on following ones. This eliminates a long-standing POLA violation, in that
DDL that affects the behavior of a later statement will now behave as
expected.
Back-patch to 9.2, since this was a performance regression compared to 9.1.
(In 9.2, place the added struct fields so as to avoid changing the offsets
of existing fields.)
Heikki Linnakangas and Tom Lane
If you take a base backup from a standby server with "pg_basebackup -X
fetch", and the timeline switches while the backup is being taken, the
backup used to fail with an error "requested WAL segment %s has already
been removed". This is because the server-side code that sends over the
required WAL files would not construct the WAL filename with the correct
timeline after a switch.
Fix that by using readdir() to scan pg_xlog for all the WAL segments in the
range, regardless of timeline.
Also, include all timeline history files in the backup, if taken with
"-X fetch". That fixes another related bug: If a timeline switch happened
just before the backup was initiated in a standby, the WAL segment
containing the initial checkpoint record contains WAL from the older
timeline too. Recovery will not accept that without a timeline history file
that lists the older timeline.
Backpatch to 9.2. Versions prior to that were not affected as you could not
take a base backup from a standby before 9.2.
Streaming replication can fetch any missing timeline history files from the
master, but recovery would read the timeline history file for the target
timeline before reading the checkpoint record, and before walreceiver has
had a chance to fetch it from the master. Delay reading it, and the sanity
checks involving timeline history, until after reading the checkpoint
record.
There is at least one scenario where this makes a difference: if you take
a base backup from a standby server right after a timeline switch, the
WAL segment containing the initial checkpoint record will begin with an
older timeline ID. Without the timeline history file, recovering that file
will fail as the older timeline ID is not recognized to be an ancestor of
the target timeline. If you try to recover from such a backup, using only
streaming replication to fetch the WAL, this patch is required for that to
work.
... not on auxiliary processes. I managed to overlook the fact that I
had disabled assertions on my HEAD checkout long ago.
Hopefully this will turn the buildfarm green again, and put an end to
today's silliness.
This makes it possible to include them only where they are used, so
we can avoid the conflict of the uid_t and gid_t datatypes that happened
in plperl (since plperl doesn't need the tar functions)
Commit da07a1e8 was broken for EXEC_BACKEND because I failed to realize
that the MaxBackends recomputation needed to be duplicated by
subprocesses in SubPostmasterMain. However, instead of having the value
be recomputed at all, it's better to assign the correct value at
postmaster initialization time, and have it be propagated to exec'ed
backends via BackendParameters.
MaxBackends stays as zero until after modules in
shared_preload_libraries have had a chance to register bgworkers, since
the value is going to be untrustworthy till that's finished.
Heikki Linnakangas and Álvaro Herrera
After receiving some WAL over streaming replication, try to open the file
from the timeline we're currently recieving, not recoveryTargetTLI. They
are usually the same, which is why wasn't noticed before, but you'd get
an error if there have been more than one timeline switch between the
current point in WAL and the recovery target.
Move some of the tar functionality that existed mostly duplicated
in both pg_dump and the walsender basebackup functionality into
port/tar.c instead, so it can be used from both. It will also be
used by pg_basebackup in the future, which would've caused a third
copy of it around.
Zoltan Boszormenyi and Magnus Hagander
In commit 11e131854f, we improved the
rule/view dumping code so that it would produce valid query representations
even if some of the tables involved in a query had been renamed since the
query was parsed. This patch extends that idea to fix problems that occur
when individual columns are renamed, or added or dropped. As before, the
core of the fix is to assign unique new aliases when a name conflict has
been created. This is complicated by the JOIN USING feature, which
requires the same column alias to be used in both input relations, but we
can handle that with a sufficiently complex approach to assigning aliases.
A fortiori, this patch takes care of situations where the query didn't have
unique column names to begin with, such as in a recent complaint from Bryan
Nuse. (Because of expansion of "SELECT *", re-parsing a dumped query can
require column name uniqueness even though the original text did not.)
The cascading standby patch in 9.2 changed the way WAL files are treated
when restored from the archive. Before, they were restored under a temporary
filename, and not kept in pg_xlog, but after the patch, they were copied
under pg_xlog. This is necessary for a cascading standby to find them, but
it also means that if the archive goes offline and a standby is restarted,
it can recover back to where it was using the files in pg_xlog. It also
means that if you take an offline backup from a standby server, it includes
all the required WAL files in pg_xlog.
However, the same change was not made to timeline history files, so if the
WAL segment containing the checkpoint record contains a timeline switch, you
will still get an error if you try to restart recovery without the archive,
or recover from an offline backup taken from the standby.
With this patch, timeline history files restored from archive are copied
into pg_xlog like WAL files are, so that pg_xlog contains all the files
required to recover. This is a corner-case pre-existing issue in 9.2, but
even more important in master where it's possible for a standby to follow a
timeline switch through streaming replication. To make that possible, the
timeline history files must be present in pg_xlog.
Heikki Linnakangas
Robert Haas
Tom Lane
Alvaro Herrera
Magnus Hagander
Peter Eisentraut
Bruce Momjian