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

Commit8bba10f

Browse files
committed
Do not select new object OIDs that match recently-dead entries.
When selecting a new OID, we take care to avoid picking one that's alreadyin use in the target table, so as not to create duplicates after the OIDcounter has wrapped around. However, up to now we used SnapshotDirty whenscanning for pre-existing entries. That ignores committed-dead rows, sothat we could select an OID matching a deleted-but-not-yet-vacuumed row.While that mostly worked, it has two problems:* If recently deleted, the dead row might still be visible to MVCCsnapshots, creating a risk for duplicate OIDs when examining the catalogswithin our own transaction. Such duplication couldn't be visible outsidethe object-creating transaction, though, and we've heard few if any fieldreports corresponding to such a symptom.* When selecting a TOAST OID, deleted toast rows definitely *are* visibleto SnapshotToast, and will remain so until vacuumed away. This leads toa conflict that will manifest in errors like "unexpected chunk number 0(expected 1) for toast value nnnnn". We've been seeing reports of sucherrors from the field for years, but the cause was unclear before.The fix is simple: just use SnapshotAny to search for conflicting rows.This results in a slightly longer window before object OIDs can berecycled, but that seems unlikely to create any large problems.Pavan DeolaseeDiscussion:https://postgr.es/m/CABOikdOgWT2hHkYG3Wwo2cyZJq2zfs1FH0FgX-=h4OLosXHf9w@mail.gmail.com
1 parent74dc05e commit8bba10f

File tree

2 files changed

+12
-11
lines changed

2 files changed

+12
-11
lines changed

‎src/backend/access/heap/tuptoaster.c

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1725,7 +1725,9 @@ toast_delete_datum(Relation rel, Datum value, bool is_speculative)
17251725
/* ----------
17261726
* toastrel_valueid_exists -
17271727
*
1728-
*Test whether a toast value with the given ID exists in the toast relation
1728+
*Test whether a toast value with the given ID exists in the toast relation.
1729+
*For safety, we consider a value to exist if there are either live or dead
1730+
*toast rows with that ID; see notes for GetNewOid().
17291731
* ----------
17301732
*/
17311733
staticbool
@@ -1737,7 +1739,6 @@ toastrel_valueid_exists(Relation toastrel, Oid valueid)
17371739
intnum_indexes;
17381740
intvalidIndex;
17391741
Relation*toastidxs;
1740-
SnapshotDataSnapshotToast;
17411742

17421743
/* Fetch a valid index relation */
17431744
validIndex=toast_open_indexes(toastrel,
@@ -1756,10 +1757,9 @@ toastrel_valueid_exists(Relation toastrel, Oid valueid)
17561757
/*
17571758
* Is there any such chunk?
17581759
*/
1759-
init_toast_snapshot(&SnapshotToast);
17601760
toastscan=systable_beginscan(toastrel,
17611761
RelationGetRelid(toastidxs[validIndex]),
1762-
true,&SnapshotToast,1,&toastkey);
1762+
true,SnapshotAny,1,&toastkey);
17631763

17641764
if (systable_getnext(toastscan)!=NULL)
17651765
result= true;

‎src/backend/catalog/catalog.c

Lines changed: 8 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -278,8 +278,12 @@ IsSharedRelation(Oid relationId)
278278
* managed to cycle through 2^32 OIDs and generate the same OID before we
279279
* finish inserting our row. This seems unlikely to be a problem. Note
280280
* that if we had to *commit* the row to end the race condition, the risk
281-
* would be rather higher; therefore we use SnapshotDirty in the test,
282-
* so that we will see uncommitted rows.
281+
* would be rather higher; therefore we use SnapshotAny in the test, so that
282+
* we will see uncommitted rows. (We used to use SnapshotDirty, but that has
283+
* the disadvantage that it ignores recently-deleted rows, creating a risk
284+
* of transient conflicts for as long as our own MVCC snapshots think a
285+
* recently-deleted row is live. The risk is far higher when selecting TOAST
286+
* OIDs, because SnapshotToast considers dead rows as active indefinitely.)
283287
*/
284288
Oid
285289
GetNewOid(Relationrelation)
@@ -332,7 +336,6 @@ Oid
332336
GetNewOidWithIndex(Relationrelation,OidindexId,AttrNumberoidcolumn)
333337
{
334338
OidnewOid;
335-
SnapshotDataSnapshotDirty;
336339
SysScanDescscan;
337340
ScanKeyDatakey;
338341
boolcollides;
@@ -345,8 +348,6 @@ GetNewOidWithIndex(Relation relation, Oid indexId, AttrNumber oidcolumn)
345348
*/
346349
Assert(!IsBinaryUpgrade||RelationGetRelid(relation)!=TypeRelationId);
347350

348-
InitDirtySnapshot(SnapshotDirty);
349-
350351
/* Generate new OIDs until we find one not in the table */
351352
do
352353
{
@@ -359,9 +360,9 @@ GetNewOidWithIndex(Relation relation, Oid indexId, AttrNumber oidcolumn)
359360
BTEqualStrategyNumber,F_OIDEQ,
360361
ObjectIdGetDatum(newOid));
361362

362-
/* see notes above about usingSnapshotDirty */
363+
/* see notes above about usingSnapshotAny */
363364
scan=systable_beginscan(relation,indexId, true,
364-
&SnapshotDirty,1,&key);
365+
SnapshotAny,1,&key);
365366

366367
collides=HeapTupleIsValid(systable_getnext(scan));
367368

0 commit comments

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