Do not select new object OIDs that match recently-dead entries.

tglsfdc · tglsfdc · commit 5b3ed6b7880b · 2018-04-11T17:41:27.000-04:00
When selecting a new OID, we take care to avoid picking one that's already in use in the target table, so as not to create duplicates after the OID counter has wrapped around. However, up to now we used SnapshotDirty when scanning for pre-existing entries. That ignores committed-dead rows, so that 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 MVCC snapshots, creating a risk for duplicate OIDs when examining the catalogs within our own transaction. Such duplication couldn't be visible outside the object-creating transaction, though, and we've heard few if any field reports corresponding to such a symptom. * When selecting a TOAST OID, deleted toast rows definitely *are* visible to SnapshotToast, and will remain so until vacuumed away. This leads to a conflict that will manifest in errors like "unexpected chunk number 0 (expected 1) for toast value nnnnn". We've been seeing reports of such errors 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 be recycled, but that seems unlikely to create any large problems. Pavan Deolasee Discussion: https://postgr.es/m/CABOikdOgWT2hHkYG3Wwo2cyZJq2zfs1FH0FgX-=h4OLosXHf9w@mail.gmail.com
diff --git a/src/backend/access/heap/tuptoaster.c b/src/backend/access/heap/tuptoaster.c
@@ -1650,7 +1650,9 @@ toast_delete_datum(Relation rel, Datum value)
 /* ----------
  * toastrel_valueid_exists -
  *
- *	Test whether a toast value with the given ID exists in the toast relation
+ *	Test whether a toast value with the given ID exists in the toast relation.
+ *	For safety, we consider a value to exist if there are either live or dead
+ *	toast rows with that ID; see notes for GetNewOid().
  * ----------
  */
 static bool
@@ -1682,7 +1684,7 @@ toastrel_valueid_exists(Relation toastrel, Oid valueid)
 	 */
 	toastscan = systable_beginscan(toastrel,
 								   RelationGetRelid(toastidxs[validIndex]),
-								   true, SnapshotToast, 1, &toastkey);
+								   true, SnapshotAny, 1, &toastkey);
 
 	if (systable_getnext(toastscan) != NULL)
 		result = true;
diff --git a/src/backend/catalog/catalog.c b/src/backend/catalog/catalog.c
@@ -270,8 +270,12 @@ IsSharedRelation(Oid relationId)
  * managed to cycle through 2^32 OIDs and generate the same OID before we
  * finish inserting our row.  This seems unlikely to be a problem.  Note
  * that if we had to *commit* the row to end the race condition, the risk
- * would be rather higher; therefore we use SnapshotDirty in the test,
- * so that we will see uncommitted rows.
+ * would be rather higher; therefore we use SnapshotAny in the test, so that
+ * we will see uncommitted rows.  (We used to use SnapshotDirty, but that has
+ * the disadvantage that it ignores recently-deleted rows, creating a risk
+ * of transient conflicts for as long as our own MVCC snapshots think a
+ * recently-deleted row is live.  The risk is far higher when selecting TOAST
+ * OIDs, because SnapshotToast considers dead rows as active indefinitely.)
  */
 Oid
 GetNewOid(Relation relation)
@@ -324,13 +328,10 @@ Oid
 GetNewOidWithIndex(Relation relation, Oid indexId, AttrNumber oidcolumn)
 {
 	Oid			newOid;
-	SnapshotData SnapshotDirty;
 	SysScanDesc scan;
 	ScanKeyData key;
 	bool		collides;
 
-	InitDirtySnapshot(SnapshotDirty);
-
 	/* Generate new OIDs until we find one not in the table */
 	do
 	{
@@ -343,9 +344,9 @@ GetNewOidWithIndex(Relation relation, Oid indexId, AttrNumber oidcolumn)
 					BTEqualStrategyNumber, F_OIDEQ,
 					ObjectIdGetDatum(newOid));
 
-		/* see notes above about using SnapshotDirty */
+		/* see notes above about using SnapshotAny */
 		scan = systable_beginscan(relation, indexId, true,
-								  &SnapshotDirty, 1, &key);
+								  SnapshotAny, 1, &key);
 
 		collides = HeapTupleIsValid(systable_getnext(scan));
 
