buffer is valid, as ReadBuffer() will elog on error. Most of the call
sites of ReadBuffer() got this right, but this patch fixes those call
sites that did not.
in all cases when keep_buf = true. This allows ANALYZE's inner loop to
use heap_release_fetch, which saves multiple buffer lookups for the same
page and avoids overestimation of cost by the vacuum cost mechanism.
http://archives.postgresql.org/pgsql-hackers/2004-10/msg00464.php.
This fix is intended to be permanent: it moves the responsibility for
calling SetBufferCommitInfoNeedsSave() into the tqual.c routines,
eliminating the requirement for callers to test whether t_infomask changed.
Also, tighten validity checking on buffer IDs in bufmgr.c --- several
routines were paranoid about out-of-range shared buffer numbers but not
about out-of-range local ones, which seems a tad pointless.
to unreserved keyword, use ereport not elog, assign a separate error code
for 'could not obtain lock' so that applications will be able to detect
that case cleanly.
creating a new tuple. This is just for debugging sanity, though, since
nothing should be paying any attention to xmax when the HEAP_XMAX_INVALID
bit is set.
as per recent discussions. Invent SubTransactionIds that are managed like
CommandIds (ie, counter is reset at start of each top transaction), and
use these instead of TransactionIds to keep track of subtransaction status
in those modules that need it. This means that a subtransaction does not
need an XID unless it actually inserts/modifies rows in the database.
Accordingly, don't assign it an XID nor take a lock on the XID until it
tries to do that. This saves a lot of overhead for subtransactions that
are only used for error recovery (eg plpgsql exceptions). Also, arrange
to release a subtransaction's XID lock as soon as the subtransaction
exits, in both the commit and abort cases. This avoids holding many
unique locks after a long series of subtransactions. The price is some
additional overhead in XactLockTableWait, but that seems acceptable.
Finally, restructure the state machine in xact.c to have a more orthogonal
set of states for subtransactions.
((Snapshot) NULL) can no longer be confused with a valid snapshot,
as per my recent suggestion. Define a macro InvalidSnapshot for 0.
Use InvalidSnapshot instead of SnapshotAny as the do-nothing special
case for heap_update and heap_delete crosschecks; this seems a little
cleaner even though the behavior is really the same.
recovery more manageable. Also, undo recent change to add FILE_HEADER
and WASTED_SPACE records to XLOG; instead make the XLOG page header
variable-size with extra fields in the first page of an XLOG file.
This should fix the boundary-case bugs observed by Mark Kirkwood.
initdb forced due to change of XLOG representation.
shift support code into heapam.c accordingly. This is in service of
soon-to-be-committed ALTER TABLE SET TABLESPACE code that will want to
use this same record type for both heaps and indexes.
Theoretically I should have forced initdb for this, but in practice there
is no change in xlog contents because CVS tip will never really emit this
record type anyhow...
There are various things left to do: contrib dbsize and oid2name modules
need work, and so does the documentation. Also someone should think about
COMMENT ON TABLESPACE and maybe RENAME TABLESPACE. Also initlocation is
dead, it just doesn't know it yet.
Gavin Sherry and Tom Lane.
of ThisStartUpID and RedoRecPtr into new backends. It's a lot easier just
to make them all grab the values out of shared memory during startup.
This helps to decouple the postmaster from checkpoint execution, which I
need since I'm intending to let the bgwriter do it instead, and it also
fixes a bug in the Win32 port: ThisStartUpID wasn't getting propagated at
all AFAICS. (Doesn't give me a lot of faith in the amount of testing that
port has gotten.)
costing us lots more to maintain than it was worth. On shared tables
it was of exactly zero benefit because we couldn't trust it to be
up to date. On temp tables it sometimes saved an lseek, but not often
enough to be worth getting excited about. And the real problem was that
we forced an lseek on every relcache flush in order to update the field.
So all in all it seems best to lose the complexity.
the next are handled by ReleaseAndReadBuffer rather than separate
ReleaseBuffer and ReadBuffer calls. This cuts the number of acquisitions
of the BufMgrLock by a factor of 2 (possibly more, if an indexscan happens
to pull successive rows from the same heap page). Unfortunately this
doesn't seem enough to get us out of the recently discussed context-switch
storm problem, but it's surely worth doing anyway.
results with tuples as ordinary varlena Datums. This commit does not
in itself do much for us, except eliminate the horrid memory leak
associated with evaluation of whole-row variables. However, it lays the
groundwork for allowing composite types as table columns, and perhaps
some other useful features as well. Per my proposal of a few days ago.
tuptoaster.c --- fields that are compressed in-line are not a reason
to invoke the toaster. Along the way, add a couple more htup.h macros
to eliminate confusing negated tests, and get rid of the already
vestigial TUPLE_TOASTER_ACTIVE symbol.
pointer type when it is not necessary to do so.
For future reference, casting NULL to a pointer type is only necessary
when (a) invoking a function AND either (b) the function has no prototype
OR (c) the function is a varargs function.
some concurrent changes Jan was making to the bufmgr. Here's an
updated version of the patch -- it should apply cleanly to CVS
HEAD and passes the regression tests.
This patch makes the following changes:
- remove the UnlockAndReleaseBuffer() and UnlockAndWriteBuffer()
macros, and replace uses of them with calls to the appropriate
functions.
- remove a bunch of #ifdef BMTRACE code: it is ugly & broken
(i.e. it doesn't compile)
- make BufferReplace() return a bool, not an int
- cleanup some logic in bufmgr.c; should be functionality
equivalent to the previous code, just cleaner now
- remove the BM_PRIVATE flag as it is unused
- improve a few comments, etc.
discussion on pgsql-hackers: in READ COMMITTED mode we just have to force
a QuerySnapshot update in the trigger, but in SERIALIZABLE mode we have
to run the scan under a current snapshot and then complain if any rows
would be updated/deleted that are not visible in the transaction snapshot.
really general fix might be difficult, I believe the only case where
AtCommit_Notify could see an uncommitted tuple is where the other guy
has just unlistened and not yet committed. The best solution seems to
be to just skip updating that tuple, on the assumption that the other
guy does not want to hear about the notification anyway. This is not
perfect --- if the other guy rolls back his unlisten instead of committing,
then he really should have gotten this notify. But to do that, we'd have
to wait to see if he commits or not, or make UNLISTEN hold exclusive lock
on pg_listener until commit. Either of these answers is deadlock-prone,
not to mention horrible for interactive performance. Do it this way
for now. (What happened to that project to do LISTEN/NOTIFY in memory
with no table, anyway?)
longer works -- IncrHeapAccessStat() didn't actually *do* anything
anymore, so no reason to keep it around AFAICS. I also fixed a
grammatical error in a comment.
Neil Conway
VACUUM FULL tuple moves. Store full-width t_infomask in WAL, rather
than storing low 8 bits and expecting to be able to reconstruct upper
bits. While at it, remove redundant t_oid field from WAL headers
(the OID, if present, is now recorded in the data portion of the tuple).
WAL version number bumped --- this does not force an initdb, you can
instead run pg_resetxlog after a clean shutdown of the old postmaster.
(overlaying low byte of page size) and add HEAP_HASOID bit to t_infomask,
per earlier discussion. Simplify scheme for overlaying fields in tuple
header (no need for cmax to live in more than one place). Don't try to
clear infomask status bits in tqual.c --- not safe to do it there. Don't
try to force output table of a SELECT INTO to have OIDs, either. Get rid
of unnecessarily complex three-state scheme for TupleDesc.tdhasoids, which
has already caused one recent failure. Improve documentation.
heap_mark4update; this avoids situations where a deleted tuple might
look like it is chained to something else. Also, cause all the WAL
redo routines to set t_ctid to equal t_self, rather than leaving it
undefined as before. Make heap_xlog_clean set the page's LSN and SUI
correctly. All per past discussions in pghackers, ranging back to
last December.
The local buffer manager is no longer used for newly-created relations
(unless they are TEMP); a new non-TEMP relation goes through the shared
bufmgr and thus will participate normally in checkpoints. But TEMP relations
use the local buffer manager throughout their lifespan. Also, operations
in TEMP relations are not logged in WAL, thus improving performance.
Since it's no longer necessary to fsync relations as they move out of the
local buffers into shared buffers, quite a lot of smgr.c/md.c/fd.c code
is no longer needed and has been removed: there's no concept of a dirty
relation anymore in md.c/fd.c, and we never fsync anything but WAL.
Still TODO: improve local buffer management algorithms so that it would
be reasonable to increase NLocBuffer.
bitmap, if present).
Per Tom Lane's suggestion the information whether a tuple has an oid
or not is carried in the tuple descriptor. For debugging reasons
tdhasoid is of type char, not bool. There are predefined values for
WITHOID, WITHOUTOID and UNDEFOID.
This patch has been generated against a cvs snapshot from last week
and I don't expect it to apply cleanly to current sources. While I
post it here for public review, I'm working on a new version against a
current snapshot. (There's been heavy activity recently; hope to
catch up some day ...)
This is a long patch; if it is too hard to swallow, I can provide it
in smaller pieces:
Part 1: Accessor macros
Part 2: tdhasoid in TupDesc
Part 3: Regression test
Part 4: Parameter withoid to heap_addheader
Part 5: Eliminate t_oid from HeapTupleHeader
Part 2 is the most hairy part because of changes in the executor and
even in the parser; the other parts are straightforward.
Up to part 4 the patched postmaster stays binary compatible to
databases created with an unpatched version. Part 5 is small (100
lines) and finally breaks compatibility.
Manfred Koizar
all places, where pd_linp is accessed. Also introduce new macros
SizeOfPageHeaderData and BTMaxItemSize. This is just source code
cosmetic, no behaviour changed.
Manfred Koizar
patch from 2002-06-10, is supposed to reduce the heap tuple header size
by four bytes on most architectures. Of course it changes the on-disk
tuple format and therefore requires initdb.
This overlays cmin/cmax/xmax fields into only two fields.
Manfred Koizar
HeapTupleHeaderData in setter and getter macros called
HeapTupleHeaderGetXmin, HeapTupleHeaderSetXmin etc.
It also introduces a "virtual" field xvac by defining
HeapTupleHeaderGetXvac and HeapTupleHeaderSetXvac. Xvac is used by
VACUUM, in fact it is stored in t_cmin.
Manfred Koizar
transaction, so as to avoid returning them out of the index AM. Saves
repeated heap_fetch operations on frequently-updated rows. Also detect
queries on unique keys (equality to all columns of a unique index), and
don't bother continuing scan once we have found first match.
Killing is implemented in the btree and hash AMs, but not yet in rtree
or gist, because there isn't an equally convenient place to do it in
those AMs (the outer amgetnext routine can't do it without re-pinning
the index page).
Did some small cleanup on APIs of HeapTupleSatisfies, heap_fetch, and
index_insert to make this a little easier.
in snapshots, per my proposal of a few days ago. Also, tweak heapam.c
routines (heap_insert, heap_update, heap_delete, heap_mark4update) to
be passed the command ID to use, instead of doing GetCurrentCommandID.
For catalog updates they'll still get passed current command ID, but
for updates generated from the main executor they'll get passed the
command ID saved in the snapshot the query is using. This should fix
some corner cases associated with functions and triggers that advance
current command ID while an outer query is still in progress.
yesterday's proposal to pghackers. Also remove unnecessary parameters
to heap_beginscan, heap_rescan. I modified pg_proc.h to reflect the
new numbers of parameters for the AM interface routines, but did not
force an initdb because nothing actually looks at those fields.
in schemas other than the system namespace; however, there's no search
path yet, and not all operations work yet on tables outside the system
namespace.
speed up repetitive failed searches; per pghackers discussion in late
January. inval.c logic substantially simplified, since we can now treat
inserts and deletes alike as far as inval events are concerned. Some
repair work needed in heap_create_with_catalog, which turns out to have
been doing CommandCounterIncrement at a point where the new relation has
non-self-consistent catalog entries. With the new inval code, that
resulted in assert failures during a relcache entry rebuild.
now just below FATAL in server_min_messages. Added more text to
highlight ordering difference between it and client_min_messages.
---------------------------------------------------------------------------
REALLYFATAL => PANIC
STOP => PANIC
New INFO level the prints to client by default
New LOG level the prints to server log by default
Cause VACUUM information to print only to the client
NOTICE => INFO where purely information messages are sent
DEBUG => LOG for purely server status messages
DEBUG removed, kept as backward compatible
DEBUG5, DEBUG4, DEBUG3, DEBUG2, DEBUG1 added
DebugLvl removed in favor of new DEBUG[1-5] symbols
New server_min_messages GUC parameter with values:
DEBUG[5-1], INFO, NOTICE, ERROR, LOG, FATAL, PANIC
New client_min_messages GUC parameter with values:
DEBUG[5-1], LOG, INFO, NOTICE, ERROR, FATAL, PANIC
Server startup now logged with LOG instead of DEBUG
Remove debug_level GUC parameter
elog() numbers now start at 10
Add test to print error message if older elog() values are passed to elog()
Bootstrap mode now has a -d that requires an argument, like postmaster
to prevent spreading of corruption when page header pointers are bad.
Merge PageZero into PageInit, since it was never used separately, and
remove separate memset calls used at most other PageInit call points.
Remove IndexPageCleanup, which wasn't used at all.
recreated since the start of our transaction, our first reference to it
errored out because we'd try to reuse our old relcache entry for it.
Do this by accepting SI inval messages just before relcache search in
heap_openr, so that dead relcache entries will be flushed before we
search. Also, break heap_open/openr into two pairs of routines,
relation_open(r) and heap_open(r). The relation_open routines make
no tests on relkind and so can be used to open anything that has a
pg_class entry. The heap_open routines are wrappers that add a relkind
test to preserve their established behavior. Use the relation_open
routines in several places that had various kluge solutions for opening
rels that might be either heap or index rels.
Also, remove the old 'heap stats' code that's been superseded by Jan's
stats collector, and clean up some inconsistencies in error reporting
between the different types of ALTER TABLE.
default, but OIDS are removed from many system catalogs that don't need them.
Some interesting side effects: TOAST pointers are 20 bytes not 32 now;
pg_description has a three-column key instead of one.
Bugs fixed in passing: BINARY cursors work again; pg_class.relhaspkey
has some usefulness; pg_dump dumps comments on indexes, rules, and
triggers in a valid order.
initdb forced.
validity checking rules for VACUUM. Make some other rearrangements of the
VACUUM code to allow more code to be shared between full and lazy VACUUM.
Minor code cleanups and added comments for TransactionId manipulations.
stub) into the rest of the system. Adopt a cleaner approach to preventing
deadlock in concurrent heap_updates: allow RelationGetBufferForTuple to
select any page of the rel, and put the onus on it to lock both buffers
in a consistent order. Remove no-longer-needed isExtend hack from
API of ReleaseAndReadBuffer.
do anything yet, but it has the necessary connections to initialization
and so forth. Make some gestures towards allowing number of blocks in
a relation to be BlockNumber, ie, unsigned int, rather than signed int.
(I doubt I got all the places that are sloppy about it, yet.) On the
way, replace the hardwired NLOCKS_PER_XACT fudge factor with a GUC
variable.
appropriate pin-count manipulation, and instead use ReleaseAndReadBuffer.
Make use of the fact that the passed-in buffer (if there is one) must
be pinned to avoid grabbing the bufmgr spinlock when we are able to
return this same buffer. Eliminate unnecessary 'previous tuple' and
'next tuple' fields of HeapScanDesc and IndexScanDesc, thereby removing
a whole lot of bookkeeping from heap_getnext() and related routines.
PageGetFreeSpace() was being called while not holding the buffer lock, which
not only could yield a garbage answer, but even if it's the right answer there
might be less space available after we reacquire the buffer lock.
Also repair potential deadlock introduced by my recent performance improvement
in RelationGetBufferForTuple(): it was possible for two heap_updates to try to
lock two buffers in opposite orders. The fix creates a global rule that
buffers of a single heap relation should be locked in decreasing block number
order. Currently, this only applies to heap_update; VACUUM can get away with
ignoring the rule since it holds exclusive lock on the whole relation anyway.
However, if we try to implement a VACUUM that can run in parallel with other
transactions, VACUUM will also have to obey the lock order rule.
when we need to move to a new page; as long as we can insert the new
tuple on the same page as before, we only need LockBuffer and not the
expensive stuff. Also, twiddle bufmgr interfaces to avoid redundant
lseeks in RelationGetBufferForTuple and BufferAlloc. Successive inserts
now require one lseek per page added, rather than one per tuple with
several additional ones at each page boundary as happened before.
Lock contention when multiple backends are inserting in same table
is also greatly reduced.
bothering to check the return value --- which meant that in case the
update or delete failed because of a concurrent update, you'd not find
out about it, except by observing later that the transaction produced
the wrong outcome. There are now subroutines simple_heap_update and
simple_heap_delete that should be used anyplace that you're not prepared
to do the full nine yards of coping with concurrent updates. In
practice, that seems to mean absolutely everywhere but the executor,
because *noplace* else was checking.
are now critical sections, so as to ensure die() won't interrupt us while
we are munging shared-memory data structures. Avoid insecure intermediate
states in some code that proc_exit will call, like palloc/pfree. Rename
START/END_CRIT_CODE to START/END_CRIT_SECTION, since that seems to be
what people tend to call them anyway, and make them be called with () like
a function call, in hopes of not confusing pg_indent.
I doubt that this is sufficient to make SIGTERM safe anywhere; there's
just too much code that could get invoked during proc_exit().
before calling RelationInvalidateHeapTuple(), which is bad because the
latter needs to look at the tuple data, which is in the shared disk
buffer. If another backend manages to recycle the buffer while this
is going on, we will compute the wrong hashindex for the tuple or
maybe even crash outright. Must hold buffer refcount until afterwards.
(This bug is not in 7.0.*; seems to be have introduced during WAL changes.)
1. Distinguish cases where a Datum representing a tuple datatype is an OID
from cases where it is a pointer to TupleTableSlot, and make sure we use
the right typlen in each case.
2. Make fetchatt() and related code support 8-byte by-value datatypes on
machines where Datum is 8 bytes. Centralize knowledge of the available
by-value datatype sizes in two macros in tupmacs.h, so that this will be
easier if we ever have to do it again.
both MULTIBYTE and TOAST prevent char(n) from being truly fixed-size.
Simplify and speed up fastgetattr() and index_getattr() macros by
eliminating special cases for attnum=1. It's just as fast to handle
the first attribute by presetting its attcacheoff to zero; so do that
instead when loading the tupledesc in relcache.c.
(WAL logging for this is not done yet, however.) Clean up a number of really
crufty things that are no longer needed now that DROP behaves nicely. Make
temp table mapper do the right things when drop or rename affecting a temp
table is rolled back. Also, remove "relation modified while in use" error
check, in favor of locking tables at first reference and holding that lock
throughout the statement.
trying to toast tuples inserted into toast tables! Fix is two-pronged:
first, ensure all columns of a toast table are marked attstorage='p',
and second, alter the target chunk size so that it's less than the
threshold for trying to toast a tuple. (Code tried to do that but the
expression was wrong.) A few cosmetic cleanups in tuptoaster too.
NOTE: initdb forced due to change in toaster chunk-size.
These two routines will now ALWAYS elog() on failure, whether you ask for
a lock or not. If you really want to get a NULL return on failure, call
the new routines heap_open_nofail()/heap_openr_nofail(). By my count there
are only about three places that actually want that behavior. There were
rather more than three places that were missing the check they needed to
make under the old convention :-(.
Don't use DISABLE_COMPLEX_MACRO on Solaris. Don't define the
replacement function in the header file. Use -KPIC, not -K PIC.
Use CC to link C++ libraries, not ld/ar.
Eliminate file not found warnings in tcl build code.
for details). It doesn't really do that much yet, since there are no
short-term memory contexts in the executor, but the infrastructure is
in place and long-term contexts are handled reasonably. A few long-
standing bugs have been fixed, such as 'VACUUM; anything' in a single
query string crashing. Also, out-of-memory is now considered a
recoverable ERROR, not FATAL.
Eliminate a large amount of crufty, now-dead code in and around
memory management.
Fix problem with holding off SIGTRAP, SIGSEGV, etc in postmaster and
backend startup.
really ought to fix relcache entry construction so that it does not
do so much with CurrentMemoryContext = CacheCxt. As is, relatively
harmless leaks in either sequential or index scanning translate to
permanent leaks if they occur when called from relcache build.
For the moment, however, the path of least resistance is to repair
all such leaks...
* Buffer refcount cleanup (per my "progress report" to pghackers, 9/22).
* Add links to backend PROC structs to sinval's array of per-backend info,
and use these links for routines that need to check the state of all
backends (rather than the slow, complicated search of the ShmemIndex
hashtable that was used before). Add databaseOID to PROC structs.
* Use this to implement an interlock that prevents DESTROY DATABASE of
a database containing running backends. (It's a little tricky to prevent
a concurrently-starting backend from getting in there, since the new
backend is not able to lock anything at the time it tries to look up
its database in pg_database. My solution is to recheck that the DB is
OK at the end of InitPostgres. It may not be a 100% solution, but it's
a lot better than no interlock at all...)
* In ALTER TABLE RENAME, flush buffers for the relation before doing the
rename of the physical files, to ensure we don't get failures later from
mdblindwrt().
* Update TRUNCATE patch so that it actually compiles against current
sources :-(.
You should do "make clean all" after pulling these changes.
additional argument specifying the kind of lock to acquire/release (or
'NoLock' to do no lock processing). Ensure that all relations are locked
with some appropriate lock level before being examined --- this ensures
that relevant shared-inval messages have been processed and should prevent
problems caused by concurrent VACUUM. Fix several bugs having to do with
mismatched increment/decrement of relation ref count and mismatched
heap_open/close (which amounts to the same thing). A bogus ref count on
a relation doesn't matter much *unless* a SI Inval message happens to
arrive at the wrong time, which is probably why we got away with this
sloppiness for so long. Repair missing grab of AccessExclusiveLock in
DROP TABLE, ALTER/RENAME TABLE, etc, as noted by Hiroshi.
Recommend 'make clean all' after pulling this update; I modified the
Relation struct layout slightly.
Will post further discussion to pghackers list shortly.
and possibly for other cases too:
DO NOT cache status of transaction in unknown state
(i.e. non-committed and non-aborted ones)
Example:
T1 reads row updated/inserted by running T2 and cache T2 status.
T2 commits.
Now T1 reads a row updated by T2 and with HEAP_XMAX_COMMITTED
in t_infomask (so cached T2 status is not changed).
Now T1 EvalPlanQual gets updated row version without HEAP_XMIN_COMMITTED
-> TransactionIdDidCommit(t_xmin) and TransactionIdDidAbort(t_xmin)
return FALSE and T2 decides that t_xmin is not committed and gets
ERROR above.
It's too late to find more smart way to handle such cases and so
I just changed xact status caching and got rid TransactionIdFlushCache()
from code.
Changed: transam.c, xact.c, lmgr.c and transam.h - last three
just because of TransactionIdFlushCache() is removed.
2. heapam.c:
T1 marked a row for update. T2 waits for T1 commit/abort.
T1 commits. T3 updates the row before T2 locks row page.
Now T2 sees that new row t_xmax is different from xact id (T1)
T2 was waiting for. Old code did Assert here. New one goes to
HeapTupleSatisfiesUpdate. Obvious changes too.
3. Added Assert to vacuum.c
4. bufmgr.c: break
Assert(buf->r_locks == 0 && !buf->ri_lock)
into two Asserts.
2. Much faster btree tuples deletion in the case when first on page
index tuple is deleted (no movement to the left page(s)).
3. Remember blkno of new root page in BTPageOpaque of
left/right siblings when root page is splitted.
no longer returns buffer pointer, can be gotten from scan;
descriptor; bootstrap can create multi-key indexes;
pg_procname index now is multi-key index; oidint2, oidint4, oidname
are gone (must be removed from regression tests); use System Cache
rather than sequential scan in many places; heap_modifytuple no
longer takes buffer parameter; remove unused buffer parameter in
a few other functions; oid8 is not index-able; remove some use of
single-character variable names; cleanup Buffer variables usage
and scan descriptor looping; cleaned up allocation and freeing of
tuples; 18k lines of diff;
Reply-To: hackers@hub.org, Dan McGuirk <mcguirk@indirect.com>
To: hackers@hub.org
Subject: [HACKERS] tmin writeback optimization
I was doing some profiling of the backend, and noticed that during a certain
benchmark I was running somewhere between 30% and 75% of the backend's CPU
time was being spent in calls to TransactionIdDidCommit() from
HeapTupleSatisfiesNow() or HeapTupleSatisfiesItself() to determine that
changed rows' transactions had in fact been committed even though the rows'
tmin values had not yet been set.
When a query looks at a given row, it needs to figure out whether the
transaction that changed the row has been committed and hence it should pay
attention to the row, or whether on the other hand the transaction is still
in progress or has been aborted and hence the row should be ignored. If
a tmin value is set, it is known definitively that the row's transaction
has been committed. However, if tmin is not set, the transaction
referred to in xmin must be looked up in pg_log, and this is what the
backend was spending a lot of time doing during my benchmark.
So, implementing a method suggested by Vadim, I created the following
patch that, the first time a query finds a committed row whose tmin value
is not set, sets it, and marks the buffer where the row is stored as
dirty. (It works for tmax, too.) This doesn't result in the boost in
real time performance I was hoping for, however it does decrease backend
CPU usage by up to two-thirds in certain situations, so it could be
rather beneficial in high-concurrency settings.
It adds a WITH OIDS option to the copy command, which allows
dumping and loading of oids.
If a copy command tried to load in an oid that is greater than
its current system max oid, the system max oid is incremented. No
checking is done to see if other backends are running and have cached
oids.
pg_dump as its first step when using the -o (oid) option, will
copy in a dummy row to set the system max oid value so as rows are
loaded in, they are certain to be lower than the system oid.
pg_dump now creates indexes at the end to speed loading
Submitted by: Bruce Momjian <maillist@candle.pha.pa.us>
Neil Conway
Tom Lane
Peter Eisentraut
Bruce Momjian
Tatsuo Ishii
PostgreSQL Daemon
Hiroshi Inoue
Jan Wieck
Vadim B. Mikheev
Marc G. Fournier