Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit40708ea

Browse files
Fix undercounting in VACUUM VERBOSE output.
The logic for determining how many nbtree pages in an index are deletedpages sometimes undercounted pages. Pages that were deleted by thecurrent VACUUM operation (as opposed to some previous VACUUM operationwhose deleted pages have yet to be reused) were sometimes overlooked.The final count is exposed to users through VACUUM VERBOSE's "%u indexpages have been deleted" output.btvacuumpage() avoided double-counting when _bt_pagedel() deleted morethan one page by assuming that only one page was deleted, and that theadditional deleted pages would get picked up during a future call tobtvacuumpage() by the same VACUUM operation. _bt_pagedel() canlegitimately delete pages that the btvacuumscan() scan will not visitagain, though, so that assumption was slightly faulty.Fix the accounting by teaching _bt_pagedel() about its caller'srequirements. It now only reports on pages that it knows btvacuumscan()won't visit again (including the current btvacuumpage() page), soeverything works out in the end.This bug has been around forever. Only backpatch to v11, though, tokeep _bt_pagedel() is sync on the branches that have today's bugfixcommitb0229f2. Note that this commit changes the signature of_bt_pagedel(), just like commitb0229f2.Author: Peter GeogheganReviewed-By: Masahiko SawadaDiscussion:https://postgr.es/m/CAH2-WzkrXBcMQWAYUJMFTTvzx_r4q=pYSjDe07JnUXhe+OZnJA@mail.gmail.comBackpatch: 11-
1 parent05b7326 commit40708ea

File tree

3 files changed

+38
-18
lines changed

3 files changed

+38
-18
lines changed

‎src/backend/access/nbtree/nbtpage.c

Lines changed: 30 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -37,8 +37,10 @@ static BTMetaPageData *_bt_getmeta(Relation rel, Buffer metabuf);
3737
staticbool_bt_mark_page_halfdead(Relationrel,Bufferleafbuf,
3838
BTStackstack);
3939
staticbool_bt_unlink_halfdead_page(Relationrel,Bufferleafbuf,
40+
BlockNumberscanblkno,
4041
bool*rightsib_empty,
41-
TransactionId*oldestBtpoXact);
42+
TransactionId*oldestBtpoXact,
43+
uint32*ndeleted);
4244
staticbool_bt_lock_branch_parent(Relationrel,BlockNumberchild,
4345
BTStackstack,Buffer*topparent,OffsetNumber*topoff,
4446
BlockNumber*target,BlockNumber*rightsib);
@@ -1301,7 +1303,9 @@ _bt_lock_branch_parent(Relation rel, BlockNumber child, BTStack stack,
13011303
*
13021304
* Returns the number of pages successfully deleted (zero if page cannot
13031305
* be deleted now; could be more than one if parent or right sibling pages
1304-
* were deleted too).
1306+
* were deleted too). Note that this does not include pages that we delete
1307+
* that the btvacuumscan scan has yet to reach; they'll get counted later
1308+
* instead.
13051309
*
13061310
* Maintains *oldestBtpoXact for any pages that get deleted. Caller is
13071311
* responsible for maintaining *oldestBtpoXact in the case of pages that were
@@ -1311,15 +1315,21 @@ _bt_lock_branch_parent(Relation rel, BlockNumber child, BTStack stack,
13111315
* carefully, it's better to run it in a temp context that can be reset
13121316
* frequently.
13131317
*/
1314-
int
1318+
uint32
13151319
_bt_pagedel(Relationrel,Bufferleafbuf,TransactionId*oldestBtpoXact)
13161320
{
1317-
intndeleted=0;
1321+
uint32ndeleted=0;
13181322
BlockNumberrightsib;
13191323
boolrightsib_empty;
13201324
Pagepage;
13211325
BTPageOpaqueopaque;
13221326

1327+
/*
1328+
* Save original leafbuf block number from caller. Only deleted blocks
1329+
* that are <= scanblkno get counted in ndeleted return value.
1330+
*/
1331+
BlockNumberscanblkno=BufferGetBlockNumber(leafbuf);
1332+
13231333
/*
13241334
* "stack" is a search stack leading (approximately) to the target page.
13251335
* It is initially NULL, but when iterating, we keep it to avoid
@@ -1370,8 +1380,9 @@ _bt_pagedel(Relation rel, Buffer leafbuf, TransactionId *oldestBtpoXact)
13701380
if (P_ISDELETED(opaque))
13711381
ereport(LOG,
13721382
(errcode(ERRCODE_INDEX_CORRUPTED),
1373-
errmsg_internal("found deleted block %u while following right link in index \"%s\"",
1383+
errmsg_internal("found deleted block %u while following right linkfrom block %uin index \"%s\"",
13741384
BufferGetBlockNumber(leafbuf),
1385+
scanblkno,
13751386
RelationGetRelationName(rel))));
13761387

13771388
_bt_relbuf(rel,leafbuf);
@@ -1521,13 +1532,13 @@ _bt_pagedel(Relation rel, Buffer leafbuf, TransactionId *oldestBtpoXact)
15211532
while (P_ISHALFDEAD(opaque))
15221533
{
15231534
/* Check for interrupts in _bt_unlink_halfdead_page */
1524-
if (!_bt_unlink_halfdead_page(rel,leafbuf,&rightsib_empty,
1525-
oldestBtpoXact))
1535+
if (!_bt_unlink_halfdead_page(rel,leafbuf,scanblkno,
1536+
&rightsib_empty,oldestBtpoXact,
1537+
&ndeleted))
15261538
{
15271539
/* _bt_unlink_halfdead_page failed, released buffer */
15281540
returnndeleted;
15291541
}
1530-
ndeleted++;
15311542
}
15321543

15331544
Assert(P_ISLEAF(opaque)&&P_ISDELETED(opaque));
@@ -1779,8 +1790,9 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
17791790
* to avoid having to reacquire a lock we already released).
17801791
*/
17811792
staticbool
1782-
_bt_unlink_halfdead_page(Relationrel,Bufferleafbuf,bool*rightsib_empty,
1783-
TransactionId*oldestBtpoXact)
1793+
_bt_unlink_halfdead_page(Relationrel,Bufferleafbuf,BlockNumberscanblkno,
1794+
bool*rightsib_empty,TransactionId*oldestBtpoXact,
1795+
uint32*ndeleted)
17841796
{
17851797
BlockNumberleafblkno=BufferGetBlockNumber(leafbuf);
17861798
BlockNumberleafleftsib;
@@ -2166,6 +2178,14 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, bool *rightsib_empty,
21662178
TransactionIdPrecedes(opaque->btpo.xact,*oldestBtpoXact))
21672179
*oldestBtpoXact=opaque->btpo.xact;
21682180

2181+
/*
2182+
* If btvacuumscan won't revisit this page in a future btvacuumpage call
2183+
* and count it as deleted then, we count it as deleted by current
2184+
* btvacuumpage call
2185+
*/
2186+
if (target <=scanblkno)
2187+
(*ndeleted)++;
2188+
21692189
/*
21702190
* Release the target, if it was not the leaf block. The leaf is always
21712191
* kept locked.

‎src/backend/access/nbtree/nbtree.c

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1349,17 +1349,17 @@ btvacuumpage(BTVacState *vstate, BlockNumber blkno, BlockNumber orig_blkno)
13491349
if (delete_now)
13501350
{
13511351
MemoryContextoldcontext;
1352-
intndel;
13531352

13541353
/* Run pagedel in a temp context to avoid memory leakage */
13551354
MemoryContextReset(vstate->pagedelcontext);
13561355
oldcontext=MemoryContextSwitchTo(vstate->pagedelcontext);
13571356

1358-
ndel=_bt_pagedel(rel,buf,&vstate->oldestBtpoXact);
1359-
1360-
/* count only this page, else may double-count parent */
1361-
if (ndel)
1362-
stats->pages_deleted++;
1357+
/*
1358+
* We trust the _bt_pagedel return value because it does not include
1359+
* any page that a future call here from btvacuumscan is expected to
1360+
* count. There will be no double-counting.
1361+
*/
1362+
stats->pages_deleted+=_bt_pagedel(rel,buf,&vstate->oldestBtpoXact);
13631363

13641364
MemoryContextSwitchTo(oldcontext);
13651365
/* pagedel released buffer, so we shouldn't */

‎src/include/access/nbtree.h

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -764,8 +764,8 @@ extern void _bt_delitems_delete(Relation rel, Buffer buf,
764764
externvoid_bt_delitems_vacuum(Relationrel,Bufferbuf,
765765
OffsetNumber*itemnos,intnitems,
766766
BlockNumberlastBlockVacuumed);
767-
externint_bt_pagedel(Relationrel,Bufferleafbuf,
768-
TransactionId*oldestBtpoXact);
767+
externuint32_bt_pagedel(Relationrel,Bufferleafbuf,
768+
TransactionId*oldestBtpoXact);
769769

770770
/*
771771
* prototypes for functions in nbtsearch.c

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp